2025
|
 | Rockenbach, Dinei A.; Araujo, Gabriell; Griebler, Dalvan; Fernandes, Luiz Gustavo GSParLib: A multi-level programming interface unifying OpenCL and CUDA for expressing stream and data parallelism Journal Article doi In: Computer Standards & Interfaces, vol. 92, pp. 103922, 2025. @article{ROCKENBACH:GSParLib:CSI:25,
title = {GSParLib: A multi-level programming interface unifying OpenCL and CUDA for expressing stream and data parallelism},
author = {Dinei A. Rockenbach and Gabriell Araujo and Dalvan Griebler and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1016/j.csi.2024.103922},
doi = {10.1016/j.csi.2024.103922},
year = {2025},
date = {2025-03-01},
urldate = {2025-03-01},
journal = {Computer Standards & Interfaces},
volume = {92},
pages = {103922},
publisher = {Elsevier},
abstract = {The evolution of Graphics Processing Units (GPUs) has allowed the industry to overcome long-lasting problems and challenges. Many belong to the stream processing domain, whose central aspect is continuously receiving and processing data from streaming data producers such as cameras and sensors. Nonetheless, programming GPUs is challenging because it requires deep knowledge of many-core programming, mechanisms and optimizations for GPUs. Current GPU programming standards do not target stream processing and present programmability and code portability limitations. Among our main scientific contributions resides GSParLib, a C++ multi-level programming interface unifying CUDA and OpenCL for GPU processing on stream and data parallelism with negligible performance losses compared to manual implementations; GSParLib is organized in two layers: one for general-purpose computing and another for high-level structured programming based on parallel patterns; a methodology to provide unified and driver agnostic interfaces minimizing performance losses; a set of parallelism strategies and optimizations for GPU processing targeting stream and data parallelism; and new experiments covering GPU performance on applications exposing stream and data parallelism.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The evolution of Graphics Processing Units (GPUs) has allowed the industry to overcome long-lasting problems and challenges. Many belong to the stream processing domain, whose central aspect is continuously receiving and processing data from streaming data producers such as cameras and sensors. Nonetheless, programming GPUs is challenging because it requires deep knowledge of many-core programming, mechanisms and optimizations for GPUs. Current GPU programming standards do not target stream processing and present programmability and code portability limitations. Among our main scientific contributions resides GSParLib, a C++ multi-level programming interface unifying CUDA and OpenCL for GPU processing on stream and data parallelism with negligible performance losses compared to manual implementations; GSParLib is organized in two layers: one for general-purpose computing and another for high-level structured programming based on parallel patterns; a methodology to provide unified and driver agnostic interfaces minimizing performance losses; a set of parallelism strategies and optimizations for GPU processing targeting stream and data parallelism; and new experiments covering GPU performance on applications exposing stream and data parallelism. |
2024
|
 | Hoffmann, Renato B.; Griebler, Dalvan; Righi, Rodrigo Rosa; Fernandes, Luiz G. Benchmarking parallel programming for single-board computers Journal Article doi In: Future Generation Computer Systems, vol. 161, pp. 119-134, 2024. @article{HOFFMANN:single-board-computers:FGCS:24,
title = {Benchmarking parallel programming for single-board computers},
author = {Renato B. Hoffmann and Dalvan Griebler and Rodrigo Rosa Righi and Luiz G. Fernandes},
url = {https://doi.org/10.1016/j.future.2024.07.003},
doi = {10.1016/j.future.2024.07.003},
year = {2024},
date = {2024-12-01},
urldate = {2024-12-01},
journal = {Future Generation Computer Systems},
volume = {161},
pages = {119-134},
publisher = {Elsevier},
abstract = {Within the computing continuum, SBCs (single-board computers) are essential in the Edge and Fog, with many featuring multiple processing cores and GPU accelerators. In this way, parallel computing plays a crucial role in enabling the full computational potential of SBCs. However, selecting the best-suited solution in this context is inherently complex due to the intricate interplay between PPI (parallel programming interface) strategies, SBC architectural characteristics, and application characteristics and constraints. To our knowledge, no solution presents a combined discussion of these three aspects. To tackle this problem, this article aims to provide a benchmark of the best-suited parallelism PPIs given a set of hardware and application characteristics and requirements. Compared to existing benchmarks, we introduce new metrics, additional applications, various parallelism interfaces, and extra hardware devices. Therefore, our contributions are the methodology to benchmark parallelism on SBCs and the characterization of the best-performing parallelism PPIs and strategies for given situations. We are confident that parallel computing will be mainstream to process edge and fog computing; thus, our solution provides the first insights regarding what kind of application and parallel programming interface is the most suited for a particular SBC hardware.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Within the computing continuum, SBCs (single-board computers) are essential in the Edge and Fog, with many featuring multiple processing cores and GPU accelerators. In this way, parallel computing plays a crucial role in enabling the full computational potential of SBCs. However, selecting the best-suited solution in this context is inherently complex due to the intricate interplay between PPI (parallel programming interface) strategies, SBC architectural characteristics, and application characteristics and constraints. To our knowledge, no solution presents a combined discussion of these three aspects. To tackle this problem, this article aims to provide a benchmark of the best-suited parallelism PPIs given a set of hardware and application characteristics and requirements. Compared to existing benchmarks, we introduce new metrics, additional applications, various parallelism interfaces, and extra hardware devices. Therefore, our contributions are the methodology to benchmark parallelism on SBCs and the characterization of the best-performing parallelism PPIs and strategies for given situations. We are confident that parallel computing will be mainstream to process edge and fog computing; thus, our solution provides the first insights regarding what kind of application and parallel programming interface is the most suited for a particular SBC hardware. |
 | Vogel, Adriano; Danelutto, Marco; Torquati, Massimo; Griebler, Dalvan; Fernandes, Luiz Gustavo Enhancing self-adaptation for efficient decision-making at run-time in streaming applications on multicores Journal Article doi In: The Journal of Supercomputing, vol. 80, no. 15, pp. 22213-22244, 2024. @article{VOGEL:Supercomputing:24,
title = {Enhancing self-adaptation for efficient decision-making at run-time in streaming applications on multicores},
author = {Adriano Vogel and Marco Danelutto and Massimo Torquati and Dalvan Griebler and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/s11227-024-06191-w},
doi = {10.1007/s11227-024-06191-w},
year = {2024},
date = {2024-10-01},
urldate = {2024-10-01},
journal = {The Journal of Supercomputing},
volume = {80},
number = {15},
pages = {22213-22244},
publisher = {Springer},
abstract = {Parallel computing is very important to accelerate the performance of computing applications. Moreover, parallel applications are expected to continue executing in more dynamic environments and react to changing conditions. In this context, applying self-adaptation is a potential solution to achieve a higher level of autonomic abstractions and runtime responsiveness. In our research, we aim to explore and assess the possible abstractions attainable through the transparent management of parallel executions by self-adaptation. Our primary objectives are to expand the adaptation space to better reflect real-world applications and assess the potential for self-adaptation to enhance efficiency. We provide the following scientific contributions: (I) A conceptual framework to improve the designing of self-adaptation; (II) A new decision-making strategy for applications with multiple parallel stages; (III) A comprehensive evaluation of the proposed decision-making strategy compared to the state-of-the-art. The results demonstrate that the proposed conceptual framework can help design and implement self-adaptive strategies that are more modular and reusable. The proposed decision-making strategy provides significant gains in accuracy compared to the state-of-the-art, increasing the parallel applications' performance and efficiency.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Parallel computing is very important to accelerate the performance of computing applications. Moreover, parallel applications are expected to continue executing in more dynamic environments and react to changing conditions. In this context, applying self-adaptation is a potential solution to achieve a higher level of autonomic abstractions and runtime responsiveness. In our research, we aim to explore and assess the possible abstractions attainable through the transparent management of parallel executions by self-adaptation. Our primary objectives are to expand the adaptation space to better reflect real-world applications and assess the potential for self-adaptation to enhance efficiency. We provide the following scientific contributions: (I) A conceptual framework to improve the designing of self-adaptation; (II) A new decision-making strategy for applications with multiple parallel stages; (III) A comprehensive evaluation of the proposed decision-making strategy compared to the state-of-the-art. The results demonstrate that the proposed conceptual framework can help design and implement self-adaptive strategies that are more modular and reusable. The proposed decision-making strategy provides significant gains in accuracy compared to the state-of-the-art, increasing the parallel applications' performance and efficiency. |
| Faé, Leonardo; Griebler, Dalvan An internal domain-specific language for expressing linear pipelines: a proof-of-concept with MPI in Rust Inproceedings doi In: Anais do XXVIII Simpósio Brasileiro de Linguagens de Programação, pp. 81-90, SBC, Curitiba/PR, 2024. @inproceedings{FAE:SBLP:24,
title = {An internal domain-specific language for expressing linear pipelines: a proof-of-concept with MPI in Rust},
author = {Leonardo Faé and Dalvan Griebler},
url = {https://doi.org/10.5753/sblp.2024.3691},
doi = {10.5753/sblp.2024.3691},
year = {2024},
date = {2024-09-01},
booktitle = {Anais do XXVIII Simpósio Brasileiro de Linguagens de Programação},
pages = {81-90},
publisher = {SBC},
address = {Curitiba/PR},
series = {SBLP'24},
abstract = {Parallel computation is necessary in order to process massive volumes of data in a timely manner. There are many parallel programming interfaces and environments, each with their own idiosyncrasies. This, alongside non-deterministic errors, make parallel programs notoriously challenging to write. Great effort has been put forth to make parallel programming for several environments easier. In this work, we propose a DSL for Rust, using the language’s source-to-source transformation facilities, that allows for automatic code generation for distributed environments that support the Message Passing Interface (MPI). Our DSL simplifies MPI’s quirks, allowing the programmer to focus almost exclusively on the computation at hand. Performance experiments show nearly or no runtime difference between our abstraction and manually written MPI code while resulting in less than half the lines of code. More elaborate code complexity metrics (Halstead) estimate from 4.5 to 14.7 times lower effort for expressing parallelism.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Parallel computation is necessary in order to process massive volumes of data in a timely manner. There are many parallel programming interfaces and environments, each with their own idiosyncrasies. This, alongside non-deterministic errors, make parallel programs notoriously challenging to write. Great effort has been put forth to make parallel programming for several environments easier. In this work, we propose a DSL for Rust, using the language’s source-to-source transformation facilities, that allows for automatic code generation for distributed environments that support the Message Passing Interface (MPI). Our DSL simplifies MPI’s quirks, allowing the programmer to focus almost exclusively on the computation at hand. Performance experiments show nearly or no runtime difference between our abstraction and manually written MPI code while resulting in less than half the lines of code. More elaborate code complexity metrics (Halstead) estimate from 4.5 to 14.7 times lower effort for expressing parallelism. |
| Löff, J'unior; Griebler, Dalvan; Fernandes, Luiz Gustavo; Binder, Walter MPR: An MPI Framework for Distributed Self-adaptive Stream Processing Inproceedings doi In: Euro-Par 2024: Parallel Processing, pp. 400-414, Springer, Madrid, Spain, 2024. @inproceedings{LOFF:Euro-Par:24,
title = {MPR: An MPI Framework for Distributed Self-adaptive Stream Processing},
author = {J'unior Löff and Dalvan Griebler and Luiz Gustavo Fernandes and Walter Binder},
url = {https://doi.org/10.1007/978-3-031-69583-4_28},
doi = {10.1007/978-3-031-69583-4_28},
year = {2024},
date = {2024-08-01},
booktitle = {Euro-Par 2024: Parallel Processing},
pages = {400-414},
publisher = {Springer},
address = {Madrid, Spain},
series = {Euro-Par'24},
abstract = {Stream processing systems must often cope with workloads varying in content, format, size, and input rate. The high variability and unpredictability make statically fine-tuning them very challenging. Our work addresses this limitation by providing a new framework and runtime system to simplify implementing and assessing new self-adaptive algorithms and optimizations. We implement a prototype on top of MPI called MPR and show its functionality. We focus on horizontal scaling by supporting the addition and removal of processes during execution time. Experiments reveal that MPR can achieve performance similar to that of a handwritten static MPI application. We also assess MPR's adaptation capabilities, showing that it can readily re-configure itself, with the help of a self-adaptive algorithm, in response to workload variations.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Stream processing systems must often cope with workloads varying in content, format, size, and input rate. The high variability and unpredictability make statically fine-tuning them very challenging. Our work addresses this limitation by providing a new framework and runtime system to simplify implementing and assessing new self-adaptive algorithms and optimizations. We implement a prototype on top of MPI called MPR and show its functionality. We focus on horizontal scaling by supporting the addition and removal of processes during execution time. Experiments reveal that MPR can achieve performance similar to that of a handwritten static MPI application. We also assess MPR's adaptation capabilities, showing that it can readily re-configure itself, with the help of a self-adaptive algorithm, in response to workload variations. |
| Guder, Larissa; Aires, João Paulo; Meneguzzi, Felipe; Griebler, Dalvan Dimensional Speech Emotion Recognition from Bimodal Features Inproceedings doi In: Anais do XXIV Simpósio Brasileiro de Computação Aplicada à Saúde, pp. 579-590, SBC, Goiânia, Brasil, 2024. @inproceedings{GUDER:SBCAS:24,
title = {Dimensional Speech Emotion Recognition from Bimodal Features},
author = {Larissa Guder and João Paulo Aires and Felipe Meneguzzi and Dalvan Griebler},
url = {https://doi.org/10.5753/sbcas.2024.2779},
doi = {10.5753/sbcas.2024.2779},
year = {2024},
date = {2024-07-01},
booktitle = {Anais do XXIV Simpósio Brasileiro de Computação Aplicada à Saúde},
pages = {579-590},
publisher = {SBC},
address = {Goiânia, Brasil},
abstract = {Considering the human-machine relationship, affective computing aims to allow computers to recognize or express emotions. Speech Emotion Recognition is a task from affective computing that aims to recognize emotions in an audio utterance. The most common way to predict emotions from the speech is using pre-determined classes in the offline mode. In that way, emotion recognition is restricted to the number of classes. To avoid this restriction, dimensional emotion recognition uses dimensions such as valence, arousal, and dominance to represent emotions with higher granularity. Existing approaches propose using textual information to improve results for the valence dimension. Although recent efforts have tried to improve results on speech emotion recognition to predict emotion dimensions, they do not consider real-world scenarios where processing the input quickly is necessary. Considering these aspects, we take the first step towards creating a bimodal approach for dimensional speech emotion recognition in streaming. Our approach combines sentence and audio representations as input to a recurrent neural network that performs speechemotion recognition. Our final architecture achieves a Concordance Correlation Coefficient of 0.5915 for arousal, 0.1431 for valence, and 0.5899 for dominance in the IEMOCAP dataset.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Considering the human-machine relationship, affective computing aims to allow computers to recognize or express emotions. Speech Emotion Recognition is a task from affective computing that aims to recognize emotions in an audio utterance. The most common way to predict emotions from the speech is using pre-determined classes in the offline mode. In that way, emotion recognition is restricted to the number of classes. To avoid this restriction, dimensional emotion recognition uses dimensions such as valence, arousal, and dominance to represent emotions with higher granularity. Existing approaches propose using textual information to improve results for the valence dimension. Although recent efforts have tried to improve results on speech emotion recognition to predict emotion dimensions, they do not consider real-world scenarios where processing the input quickly is necessary. Considering these aspects, we take the first step towards creating a bimodal approach for dimensional speech emotion recognition in streaming. Our approach combines sentence and audio representations as input to a recurrent neural network that performs speechemotion recognition. Our final architecture achieves a Concordance Correlation Coefficient of 0.5915 for arousal, 0.1431 for valence, and 0.5899 for dominance in the IEMOCAP dataset. |
| Gomes, Carlos Falcao Azevedo; Araujo, Adriel Silva; Ahmad, Sunna Imtiaz; Magnaguagno, Mauricio Cecilio; Teixeira, Vinicius Crisosthemos; Rajapuri, Anushri Singh; Roederer, Quinn; Griebler, Dalvan; Dutra, Vinicius; Turkkahraman, Hakan; Pinho, Marcio Sarroglia Multiview Machine Learning Classification of Tooth Extraction in Orthodontics Using Intraoral Scans Inproceedings doi In: 2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC), pp. 1977-1982, IEEE, Osaka, Japan, 2024. @inproceedings{GOMES:COMPSAC:24,
title = {Multiview Machine Learning Classification of Tooth Extraction in Orthodontics Using Intraoral Scans},
author = {Carlos Falcao Azevedo Gomes and Adriel Silva Araujo and Sunna Imtiaz Ahmad and Mauricio Cecilio Magnaguagno and Vinicius Crisosthemos Teixeira and Anushri Singh Rajapuri and Quinn Roederer and Dalvan Griebler and Vinicius Dutra and Hakan Turkkahraman and Marcio Sarroglia Pinho},
url = {https://doi.org/10.1109/COMPSAC61105.2024.00316},
doi = {10.1109/COMPSAC61105.2024.00316},
year = {2024},
date = {2024-07-01},
booktitle = {2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC)},
pages = {1977-1982},
publisher = {IEEE},
address = {Osaka, Japan},
abstract = {Orthodontic treatment planning often involves de-ciding whether to extract teeth, a critical and irreversible decision. Integrating machine learning (ML) can enhance decision-making. This study proposes using Intraoral Scans (IOS) 3D models to predict extraction/non-extraction binary decisions with ML models. We leverage a multiview approach, using images taken from multiple points of view of the 3D model. The methodology involved a dataset composed of preprocessed IOS from 181 subjects and an experimental procedure that evaluated multiple ML models in their ability to classify subjects using either grayscale pixel intensities or radiomic features. The results indicated that a logistic model applied to the radiomic features from the back and frontal views of the 3D models was one of the best model candidates, achieving a test accuracy of 70 % and F1 score of. 73 and. 65 for non-extraction and extraction cases, respectively. Overall, these findings indicate that a multiview approach to IOS 3D models can be used to predict extraction/non-extraction decisions. In addition, the results suggest that radiomic features provide useful information in the analysis of IOS data.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Orthodontic treatment planning often involves de-ciding whether to extract teeth, a critical and irreversible decision. Integrating machine learning (ML) can enhance decision-making. This study proposes using Intraoral Scans (IOS) 3D models to predict extraction/non-extraction binary decisions with ML models. We leverage a multiview approach, using images taken from multiple points of view of the 3D model. The methodology involved a dataset composed of preprocessed IOS from 181 subjects and an experimental procedure that evaluated multiple ML models in their ability to classify subjects using either grayscale pixel intensities or radiomic features. The results indicated that a logistic model applied to the radiomic features from the back and frontal views of the 3D models was one of the best model candidates, achieving a test accuracy of 70 % and F1 score of. 73 and. 65 for non-extraction and extraction cases, respectively. Overall, these findings indicate that a multiview approach to IOS 3D models can be used to predict extraction/non-extraction decisions. In addition, the results suggest that radiomic features provide useful information in the analysis of IOS data. |
| Leonarczyk, Ricardo; Griebler, Dalvan; Mencagli, Gabriele; Danelutto, Marco Evaluation of Adaptive Micro-batching Techniques for GPU-accelerated Stream Processing Inproceedings doi In: Euro-Par 2023: Parallel Processing Workshops, pp. 81-92, Springer, Limassol, Cyprus, 2024. @inproceedings{LEONARCZYK:Euro-ParW:23,
title = {Evaluation of Adaptive Micro-batching Techniques for GPU-accelerated Stream Processing},
author = {Ricardo Leonarczyk and Dalvan Griebler and Gabriele Mencagli and Marco Danelutto},
url = {https://doi.org/10.1007/978-3-031-50684-0_7},
doi = {10.1007/978-3-031-50684-0_7},
year = {2024},
date = {2024-04-01},
booktitle = {Euro-Par 2023: Parallel Processing Workshops},
pages = {81-92},
publisher = {Springer},
address = {Limassol, Cyprus},
series = {Euro-ParW'23},
abstract = {Stream processing plays a vital role in applications that require continuous, low-latency data processing. Thanks to their extensive parallel processing capabilities and relatively low cost, GPUs are well-suited to scenarios where such applications require substantial computational resources. However, micro-batching becomes essential for efficient GPU computation within stream processing systems. However, finding appropriate batch sizes to maintain an adequate level of service is often challenging, particularly in cases where applications experience fluctuations in input rate and workload. Addressing this challenge requires adjusting the optimal batch size at runtime. This study proposes a methodology for evaluating different self-adaptive micro-batching strategies in a real-world complex streaming application used as a benchmark.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Stream processing plays a vital role in applications that require continuous, low-latency data processing. Thanks to their extensive parallel processing capabilities and relatively low cost, GPUs are well-suited to scenarios where such applications require substantial computational resources. However, micro-batching becomes essential for efficient GPU computation within stream processing systems. However, finding appropriate batch sizes to maintain an adequate level of service is often challenging, particularly in cases where applications experience fluctuations in input rate and workload. Addressing this challenge requires adjusting the optimal batch size at runtime. This study proposes a methodology for evaluating different self-adaptive micro-batching strategies in a real-world complex streaming application used as a benchmark. |
 | Mencagli, Gabriele; Torquati, Massimo; Griebler, Dalvan; Fais, Alessandra; Danelutto, Marco General-purpose data stream processing on heterogeneous architectures with WindFlow Journal Article doi In: Journal of Parallel and Distributed Computing, vol. 184, pp. 104782, 2024. @article{MENCAGLI:JPDC:24,
title = {General-purpose data stream processing on heterogeneous architectures with WindFlow},
author = {Gabriele Mencagli and Massimo Torquati and Dalvan Griebler and Alessandra Fais and Marco Danelutto},
url = {https://doi.org/10.1016/j.jpdc.2023.104782},
doi = {10.1016/j.jpdc.2023.104782},
year = {2024},
date = {2024-02-01},
urldate = {2024-02-01},
journal = {Journal of Parallel and Distributed Computing},
volume = {184},
pages = {104782},
publisher = {Elsevier},
abstract = {Many emerging applications analyze data streams by running graphs of communicating tasks called operators. To develop and deploy such applications, Stream Processing Systems (SPSs) like Apache Storm and Flink have been made available to researchers and practitioners. They exhibit imperative or declarative programming interfaces to develop operators running arbitrary algorithms working on structured or unstructured data streams. In this context, the interest in leveraging hardware acceleration with GPUs has become more pronounced in high-throughput use cases. Unfortunately, GPU acceleration has been studied for relational operators working on structured streams only, while non-relational operators have often been overlooked. This paper presents WindFlow, a library supporting the seamless GPU offloading of general partitioned-stateful operators, extending the range of operators that benefit from hardware acceleration. Its design provides high throughput still exposing a high-level API to users compared with the raw utilization of GPUs in Apache Flink.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Many emerging applications analyze data streams by running graphs of communicating tasks called operators. To develop and deploy such applications, Stream Processing Systems (SPSs) like Apache Storm and Flink have been made available to researchers and practitioners. They exhibit imperative or declarative programming interfaces to develop operators running arbitrary algorithms working on structured or unstructured data streams. In this context, the interest in leveraging hardware acceleration with GPUs has become more pronounced in high-throughput use cases. Unfortunately, GPU acceleration has been studied for relational operators working on structured streams only, while non-relational operators have often been overlooked. This paper presents WindFlow, a library supporting the seamless GPU offloading of general partitioned-stateful operators, extending the range of operators that benefit from hardware acceleration. Its design provides high throughput still exposing a high-level API to users compared with the raw utilization of GPUs in Apache Flink. |
 | Garcia, Adriano Marques; Griebler, Dalvan; Schepke, Claudio; García, José Daniel; Muñoz, Javier Fernández; Fernandes, Luiz Gustavo Performance and programmability of GrPPI for parallel stream processing on multi-cores Journal Article doi In: The Journal of Supercomputing, vol. 80, no. 9, pp. 12966-13000, 2024. @article{GARCIA:JS:24,
title = {Performance and programmability of GrPPI for parallel stream processing on multi-cores},
author = {Adriano Marques Garcia and Dalvan Griebler and Claudio Schepke and José Daniel García and Javier Fernández Muñoz and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/s11227-024-05934-z},
doi = {10.1007/s11227-024-05934-z},
year = {2024},
date = {2024-02-01},
urldate = {2024-02-01},
journal = {The Journal of Supercomputing},
volume = {80},
number = {9},
pages = {12966-13000},
publisher = {Springer},
abstract = {GrPPI library aims to simplify the burdening task of parallel programming. It provides a unified, abstract, and generic layer while promising minimal overhead on performance. Although it supports stream parallelism, GrPPI lacks an evaluation regarding representative performance metrics for this domain, such as throughput and latency. This work evaluates GrPPI focused on parallel stream processing. We compare the throughput and latency performance, memory usage, and programmability of GrPPI against handwritten parallel code. For this, we use the benchmarking framework SPBench to build custom GrPPI benchmarks and benchmarks with handwritten parallel code using the same backends supported by GrPPI. The basis of the benchmarks is real applications, such as Lane Detection, Bzip2, Face Recognizer, and Ferret. Experiments show that while performance is often competitive with handwritten parallel code, the infeasibility of fine-tuning GrPPI is a crucial drawback for emerging applications. Despite this, programmability experiments estimate that GrPPI can potentially reduce the development time of parallel applications by about three times.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
GrPPI library aims to simplify the burdening task of parallel programming. It provides a unified, abstract, and generic layer while promising minimal overhead on performance. Although it supports stream parallelism, GrPPI lacks an evaluation regarding representative performance metrics for this domain, such as throughput and latency. This work evaluates GrPPI focused on parallel stream processing. We compare the throughput and latency performance, memory usage, and programmability of GrPPI against handwritten parallel code. For this, we use the benchmarking framework SPBench to build custom GrPPI benchmarks and benchmarks with handwritten parallel code using the same backends supported by GrPPI. The basis of the benchmarks is real applications, such as Lane Detection, Bzip2, Face Recognizer, and Ferret. Experiments show that while performance is often competitive with handwritten parallel code, the infeasibility of fine-tuning GrPPI is a crucial drawback for emerging applications. Despite this, programmability experiments estimate that GrPPI can potentially reduce the development time of parallel applications by about three times. |
 | Fischer, Gabriel Souto; Ramos, Gabriel Oliveira; Costa, Cristiano André; Alberti, Antonio Marcos; Griebler, Dalvan; Singh, Dhananjay; Righi, Rodrigo Rosa Multi-Hospital Management: Combining Vital Signs IoT Data and the Elasticity Technique to Support Healthcare 4.0 Journal Article doi In: IoT, vol. 5, no. 2, pp. 381-408, 2024. @article{FISCHER:IoT:24,
title = {Multi-Hospital Management: Combining Vital Signs IoT Data and the Elasticity Technique to Support Healthcare 4.0},
author = {Gabriel Souto Fischer and Gabriel Oliveira Ramos and Cristiano André Costa and Antonio Marcos Alberti and Dalvan Griebler and Dhananjay Singh and Rodrigo Rosa Righi},
url = {https://doi.org/10.3390/iot5020019},
doi = {10.3390/iot5020019},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
journal = {IoT},
volume = {5},
number = {2},
pages = {381-408},
publisher = {MDPI},
abstract = {Smart cities can improve the quality of life of citizens by optimizing the utilization of resources. In an IoT-connected environment, people's health can be constantly monitored, which can help identify medical problems before they become serious. However, overcrowded hospitals can lead to long waiting times for patients to receive treatment. The literature presents alternatives to address this problem by adjusting care capacity to demand. However, there is still a need for a solution that can adjust human resources in multiple healthcare settings, which is the reality of cities. This work introduces HealCity, a smart-city-focused model that can monitor patients’ use of healthcare settings and adapt the allocation of health professionals to meet their needs. HealCity uses vital signs (IoT) data in prediction techniques to anticipate when the demand for a given environment will exceed its capacity and suggests actions to allocate health professionals accordingly. Additionally, we introduce the concept of multilevel proactive human resources elasticity in smart cities, thus managing human resources at different levels of a smart city. An algorithm is also devised to automatically manage and identify the appropriate hospital for a possible future patient. Furthermore, some IoT deployment considerations are presented based on a hardware implementation for the proposed model. HealCity was evaluated with four hospital settings and obtained promising results: Compared to hospitals with rigid professional allocations, it reduced waiting time for care by up to 87.62%.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Smart cities can improve the quality of life of citizens by optimizing the utilization of resources. In an IoT-connected environment, people's health can be constantly monitored, which can help identify medical problems before they become serious. However, overcrowded hospitals can lead to long waiting times for patients to receive treatment. The literature presents alternatives to address this problem by adjusting care capacity to demand. However, there is still a need for a solution that can adjust human resources in multiple healthcare settings, which is the reality of cities. This work introduces HealCity, a smart-city-focused model that can monitor patients’ use of healthcare settings and adapt the allocation of health professionals to meet their needs. HealCity uses vital signs (IoT) data in prediction techniques to anticipate when the demand for a given environment will exceed its capacity and suggests actions to allocate health professionals accordingly. Additionally, we introduce the concept of multilevel proactive human resources elasticity in smart cities, thus managing human resources at different levels of a smart city. An algorithm is also devised to automatically manage and identify the appropriate hospital for a possible future patient. Furthermore, some IoT deployment considerations are presented based on a hardware implementation for the proposed model. HealCity was evaluated with four hospital settings and obtained promising results: Compared to hospitals with rigid professional allocations, it reduced waiting time for care by up to 87.62%. |
2023
|
| Bianchessi, Arthur S.; Mallmann, Leonardo; Hoffmann, Renato Barreto; Griebler, Dalvan Conversão do NAS Parallel Benchmarks para C++ Standard Inproceedings doi In: Anais do XXIII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD), pp. 313-324, SBC, Porto Alegre, Brasil, 2023. @inproceedings{BIANCHESSI:WSCAD:23,
title = {Conversão do NAS Parallel Benchmarks para C++ Standard},
author = {Arthur S. Bianchessi and Leonardo Mallmann and Renato Barreto Hoffmann and Dalvan Griebler},
url = {https://doi.org/10.5753/wscad.2023.235913},
doi = {10.5753/wscad.2023.235913},
year = {2023},
date = {2023-10-01},
booktitle = {Anais do XXIII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD)},
pages = {313-324},
publisher = {SBC},
address = {Porto Alegre, Brasil},
abstract = {A linguagem C++ recebeu novas abstrações de paralelismo com a definição das políticas de execução dos algoritmos da biblioteca padrão. Entretanto, a adequabilidade e o desempenho dessa alternativa ainda necessita ser estudado em comparação com outras alternativas bem estabelecidas. Portanto, o objetivo deste trabalho foi explorar a vasta gama de opções de recursos da biblioteca padrão C++ para avaliar a aplicabilidade e desempenho a partir de cinco kernels do NPB. Através dos experimentos em um ambiente multithreaded, foi constatado que a incorporação de estruturas de dados da biblioteca padrão, assim como a abstração para acesso multidimensional criada, não apresentam impacto notável no tempo de execução. Já os algoritmos com políticas de execução paralela demonstraram uma perda de desempenho estatisticamente significativa.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
A linguagem C++ recebeu novas abstrações de paralelismo com a definição das políticas de execução dos algoritmos da biblioteca padrão. Entretanto, a adequabilidade e o desempenho dessa alternativa ainda necessita ser estudado em comparação com outras alternativas bem estabelecidas. Portanto, o objetivo deste trabalho foi explorar a vasta gama de opções de recursos da biblioteca padrão C++ para avaliar a aplicabilidade e desempenho a partir de cinco kernels do NPB. Através dos experimentos em um ambiente multithreaded, foi constatado que a incorporação de estruturas de dados da biblioteca padrão, assim como a abstração para acesso multidimensional criada, não apresentam impacto notável no tempo de execução. Já os algoritmos com políticas de execução paralela demonstraram uma perda de desempenho estatisticamente significativa. |
| Alf, Lucas; Hoffmann, Renato Barreto; Müller, Caetano; Griebler, Dalvan Análise da Execução de Algoritmos de Aprendizado de Máquina em Dispositivos Embarcados Inproceedings doi In: Anais do XXIII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD), pp. 61-72, SBC, Porto Alegre, Brasil, 2023. @inproceedings{ALF:WSCAD:23,
title = {Análise da Execução de Algoritmos de Aprendizado de Máquina em Dispositivos Embarcados},
author = {Lucas Alf and Renato Barreto Hoffmann and Caetano Müller and Dalvan Griebler},
url = {https://doi.org/10.5753/wscad.2023.235915},
doi = {10.5753/wscad.2023.235915},
year = {2023},
date = {2023-10-01},
booktitle = {Anais do XXIII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD)},
pages = {61-72},
publisher = {SBC},
address = {Porto Alegre, Brasil},
abstract = {Os avanços na área de IoT motivam a utilização de algoritmos de aprendizado de máquina em dispositivos embarcados. Entretanto, esses algoritmos exigem uma quantidade considerável de recursos computacionais. O objetivo deste trabalho consistiu em analisar algoritmos de aprendizado de máquina em dispositivos embarcados utilizando paralelismo em CPU e GPU com o intuito de compreender quais características de hardware e software desempenham melhor em relação ao consumo energético, inferências por segundo e acurácia. Foram avaliados três modelos de Convolutional Neural Network, bem como algoritmos tradicionais e redes neurais de classificação e regressão. Os experimentos demonstraram que o PyTorch obteve o melhor desempenho nos modelos de CNN e nas redes neurais de classificação e regressão usando GPU, enquanto o Keras obteve um melhor desempenho ao utilizar somente CPU.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Os avanços na área de IoT motivam a utilização de algoritmos de aprendizado de máquina em dispositivos embarcados. Entretanto, esses algoritmos exigem uma quantidade considerável de recursos computacionais. O objetivo deste trabalho consistiu em analisar algoritmos de aprendizado de máquina em dispositivos embarcados utilizando paralelismo em CPU e GPU com o intuito de compreender quais características de hardware e software desempenham melhor em relação ao consumo energético, inferências por segundo e acurácia. Foram avaliados três modelos de Convolutional Neural Network, bem como algoritmos tradicionais e redes neurais de classificação e regressão. Os experimentos demonstraram que o PyTorch obteve o melhor desempenho nos modelos de CNN e nas redes neurais de classificação e regressão usando GPU, enquanto o Keras obteve um melhor desempenho ao utilizar somente CPU. |
| Andrade, Gabriella; Griebler, Dalvan; Santos, Rodrigo; Fernandes, Luiz Gustavo Extending the Planning Poker Method to Estimate the Development Effort of Parallel Applications Inproceedings doi In: Anais do XXIII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD), pp. 181-192, SBC, Porto Alegre, Brasil, 2023. @inproceedings{ANDRADE:WSCAD:23,
title = {Extending the Planning Poker Method to Estimate the Development Effort of Parallel Applications},
author = {Gabriella Andrade and Dalvan Griebler and Rodrigo Santos and Luiz Gustavo Fernandes},
url = {https://doi.org/10.5753/wscad.2023.235925},
doi = {10.5753/wscad.2023.235925},
year = {2023},
date = {2023-10-01},
booktitle = {Anais do XXIII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD)},
pages = {181-192},
publisher = {SBC},
address = {Porto Alegre, Brasil},
abstract = {Since different Parallel Programming Interfaces (PPIs) are available to programmers, evaluating them to identify the most suitable PPI also became necessary. Recently, in addition to the performance of PPIs, developers’ productivity has also been evaluated by researchers in parallel processing. Some researchers conduct empirical studies involving people for productivity evaluation, which is time-consuming. Aiming to propose a less costly method for evaluating the development effort of parallel applications, we proposed modifying the Planning Poker method in this paper. We consider a representative set of parallel stream processing applications to evaluate the proposed modification. Our results showed that the proposed method required less effort for practical use than the controlled experiments with students.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Since different Parallel Programming Interfaces (PPIs) are available to programmers, evaluating them to identify the most suitable PPI also became necessary. Recently, in addition to the performance of PPIs, developers’ productivity has also been evaluated by researchers in parallel processing. Some researchers conduct empirical studies involving people for productivity evaluation, which is time-consuming. Aiming to propose a less costly method for evaluating the development effort of parallel applications, we proposed modifying the Planning Poker method in this paper. We consider a representative set of parallel stream processing applications to evaluate the proposed modification. Our results showed that the proposed method required less effort for practical use than the controlled experiments with students. |
| Hoffmann, Renato Barreto; Faé, Leonardo; Manssour, Isabel; Griebler, Dalvan Analyzing C++ Stream Parallelism in Shared-Memory when Porting to Flink and Storm Inproceedings doi In: International Symposium on Computer Architecture and High Performance Computing Workshops (SBAC-PADW), pp. 1-8, IEEE, Porto Alegre, Brazil, 2023. @inproceedings{HOFFMANN:SBAC-PADW:23,
title = {Analyzing C++ Stream Parallelism in Shared-Memory when Porting to Flink and Storm},
author = {Renato Barreto Hoffmann and Leonardo Faé and Isabel Manssour and Dalvan Griebler},
url = {https://doi.org/10.1109/SBAC-PADW60351.2023.00017},
doi = {10.1109/SBAC-PADW60351.2023.00017},
year = {2023},
date = {2023-10-01},
booktitle = {International Symposium on Computer Architecture and High Performance Computing Workshops (SBAC-PADW)},
pages = {1-8},
publisher = {IEEE},
address = {Porto Alegre, Brazil},
series = {SBAC-PADW'23},
abstract = {Stream processing plays a crucial role in various information-oriented digital systems. Two popular frameworks for real-time data processing, Flink and Storm, provide solutions for effective parallel stream processing in Java. An option to leverage Java's mature ecosystem for distributed stream processing involves porting legacy C++ applications to Java. However, this raises considerations on the adequacy of the equivalent Java mechanisms and potential degradation in throughput. Therefore, our objective is to evaluate programmability and performance when converting stream processing applications from C++ to Java while also exploring the parallelization capabilities offered by Flink and Storm. Furthermore, we aim to assess the throughput of Flink and Storm on shared-memory manycore machines, a hardware architecture commonly found in cloud environments. To achieve this, we conduct experiments involving four different stream processing applications. We highlight challenges encountered when porting C++ to Java and working with Flink and Storm. Furthermore, we discuss throughput, latency, CPU, and memory usage results.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Stream processing plays a crucial role in various information-oriented digital systems. Two popular frameworks for real-time data processing, Flink and Storm, provide solutions for effective parallel stream processing in Java. An option to leverage Java's mature ecosystem for distributed stream processing involves porting legacy C++ applications to Java. However, this raises considerations on the adequacy of the equivalent Java mechanisms and potential degradation in throughput. Therefore, our objective is to evaluate programmability and performance when converting stream processing applications from C++ to Java while also exploring the parallelization capabilities offered by Flink and Storm. Furthermore, we aim to assess the throughput of Flink and Storm on shared-memory manycore machines, a hardware architecture commonly found in cloud environments. To achieve this, we conduct experiments involving four different stream processing applications. We highlight challenges encountered when porting C++ to Java and working with Flink and Storm. Furthermore, we discuss throughput, latency, CPU, and memory usage results. |
| Faé, Leonardo; Hoffmann, Renato Barreto; Griebler, Dalvan Source-to-Source Code Transformation on Rust for High-Level Stream Parallelism Inproceedings doi In: XXVII Brazilian Symposium on Programming Languages (SBLP), pp. 41-49, ACM, Campo Grande, Brazil, 2023. @inproceedings{FAE:SBLP:23,
title = {Source-to-Source Code Transformation on Rust for High-Level Stream Parallelism},
author = {Leonardo Faé and Renato Barreto Hoffmann and Dalvan Griebler},
url = {https://doi.org/10.1145/3624309.3624320},
doi = {10.1145/3624309.3624320},
year = {2023},
date = {2023-09-01},
booktitle = {XXVII Brazilian Symposium on Programming Languages (SBLP)},
pages = {41-49},
publisher = {ACM},
address = {Campo Grande, Brazil},
series = {SBLP'23},
abstract = {Utilizing parallel systems to their full potential can be challenging for general-purpose developers. A solution to this problem is to create high-level abstractions using Domain-Specific Languages (DSL). We create a stream-processing DSL for Rust, a growing programming language focusing on performance and safety. To that end, we explore Rust’s macros as a high-level abstraction tool to support an existing DSL language named SPar and perform source-to-source code transformations in the abstract syntax tree. We aim to assess the Rust source-to-source code transformations toolset and its implications. We highlight that Rust macros are powerful tools for performing source-to-source code transformations for abstracting structured stream processing. In addition, execution time and programmability results are comparable to other solutions.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Utilizing parallel systems to their full potential can be challenging for general-purpose developers. A solution to this problem is to create high-level abstractions using Domain-Specific Languages (DSL). We create a stream-processing DSL for Rust, a growing programming language focusing on performance and safety. To that end, we explore Rust’s macros as a high-level abstraction tool to support an existing DSL language named SPar and perform source-to-source code transformations in the abstract syntax tree. We aim to assess the Rust source-to-source code transformations toolset and its implications. We highlight that Rust macros are powerful tools for performing source-to-source code transformations for abstracting structured stream processing. In addition, execution time and programmability results are comparable to other solutions. |
 | Andrade, Gabriella; Griebler, Dalvan; Santos, Rodrigo; Fernandes, Luiz Gustavo A parallel programming assessment for stream processing applications on multi-core systems Journal Article doi In: Computer Standards & Interfaces, vol. 84, pp. 103691, 2023. @article{ANDRADE:CSI:2023,
title = {A parallel programming assessment for stream processing applications on multi-core systems},
author = {Gabriella Andrade and Dalvan Griebler and Rodrigo Santos and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1016/j.csi.2022.103691},
doi = {10.1016/j.csi.2022.103691},
year = {2023},
date = {2023-03-01},
journal = {Computer Standards & Interfaces},
volume = {84},
pages = {103691},
publisher = {Elsevier},
abstract = {Multi-core systems are any computing device nowadays and stream processing applications are becoming recurrent workloads, demanding parallelism to achieve the desired quality of service. As soon as data, tasks, or requests arrive, they must be computed, analyzed, or processed. Since building such applications is not a trivial task, the software industry must adopt parallel APIs (Application Programming Interfaces) that simplify the exploitation of parallelism in hardware for accelerating time-to-market. In the last years, research efforts in academia and industry provided a set of parallel APIs, increasing productivity to software developers. However, a few studies are seeking to prove the usability of these interfaces. In this work, we aim to present a parallel programming assessment regarding the usability of parallel API for expressing parallelism on the stream processing application domain and multi-core systems. To this end, we conducted an empirical study with beginners in parallel application development. The study covered three parallel APIs, reporting several quantitative and qualitative indicators involving developers. Our contribution also comprises a parallel programming assessment methodology, which can be replicated in future assessments. This study revealed important insights such as recurrent compile-time and programming logic errors performed by beginners in parallel programming, as well as the programming effort, challenges, and learning curve. Moreover, we collected the participants’ opinions about their experience in this study to understand deeply the results achieved.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Multi-core systems are any computing device nowadays and stream processing applications are becoming recurrent workloads, demanding parallelism to achieve the desired quality of service. As soon as data, tasks, or requests arrive, they must be computed, analyzed, or processed. Since building such applications is not a trivial task, the software industry must adopt parallel APIs (Application Programming Interfaces) that simplify the exploitation of parallelism in hardware for accelerating time-to-market. In the last years, research efforts in academia and industry provided a set of parallel APIs, increasing productivity to software developers. However, a few studies are seeking to prove the usability of these interfaces. In this work, we aim to present a parallel programming assessment regarding the usability of parallel API for expressing parallelism on the stream processing application domain and multi-core systems. To this end, we conducted an empirical study with beginners in parallel application development. The study covered three parallel APIs, reporting several quantitative and qualitative indicators involving developers. Our contribution also comprises a parallel programming assessment methodology, which can be replicated in future assessments. This study revealed important insights such as recurrent compile-time and programming logic errors performed by beginners in parallel programming, as well as the programming effort, challenges, and learning curve. Moreover, we collected the participants’ opinions about their experience in this study to understand deeply the results achieved. |
| Garcia, Adriano Marques; Griebler, Dalvan; Schepke, Claudio; García, José Daniel; Muñoz, Javier Fernández; Fernandes, Luiz Gustavo A Latency, Throughput, and Programmability Perspective of GrPPI for Streaming on Multi-cores Inproceedings doi In: 31st Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP), pp. 164-168, IEEE, Naples, Italy, 2023. @inproceedings{GARCIA:PDP:23,
title = {A Latency, Throughput, and Programmability Perspective of GrPPI for Streaming on Multi-cores},
author = {Adriano Marques Garcia and Dalvan Griebler and Claudio Schepke and José Daniel García and Javier Fernández Muñoz and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/PDP59025.2023.00033},
doi = {10.1109/PDP59025.2023.00033},
year = {2023},
date = {2023-03-01},
booktitle = {31st Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP)},
pages = {164-168},
publisher = {IEEE},
address = {Naples, Italy},
series = {PDP'23},
abstract = {Several solutions aim to simplify the burdening task of parallel programming. The GrPPI library is one of them. It allows users to implement parallel code for multiple backends through a unified, abstract, and generic layer while promising minimal overhead on performance. An outspread evaluation of GrPPI regarding stream parallelism with representative metrics for this domain, such as throughput and latency, was not yet done. In this work, we evaluate GrPPI focused on stream processing. We evaluate performance, memory usage, and programming effort and compare them against handwritten parallel code. For this, we use the benchmarking framework SPBench to build custom GrPPI benchmarks. The basis of the benchmarks is real applications, such as Lane Detection, Bzip2, Face Recognizer, and Ferret. Experiments show that while performance is competitive with handwritten code in some cases, in other cases, the infeasibility of fine-tuning GrPPI is a crucial drawback. Despite this, programmability experiments estimate that GrPPI has the potential to reduce by about three times the development time of parallel applications.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Several solutions aim to simplify the burdening task of parallel programming. The GrPPI library is one of them. It allows users to implement parallel code for multiple backends through a unified, abstract, and generic layer while promising minimal overhead on performance. An outspread evaluation of GrPPI regarding stream parallelism with representative metrics for this domain, such as throughput and latency, was not yet done. In this work, we evaluate GrPPI focused on stream processing. We evaluate performance, memory usage, and programming effort and compare them against handwritten parallel code. For this, we use the benchmarking framework SPBench to build custom GrPPI benchmarks. The basis of the benchmarks is real applications, such as Lane Detection, Bzip2, Face Recognizer, and Ferret. Experiments show that while performance is competitive with handwritten code in some cases, in other cases, the infeasibility of fine-tuning GrPPI is a crucial drawback. Despite this, programmability experiments estimate that GrPPI has the potential to reduce by about three times the development time of parallel applications. |
 | Garcia, Adriano Marques; Griebler, Dalvan; Schepke, Claudio; Fernandes, Luiz Gustavo Micro-batch and data frequency for stream processing on multi-cores Journal Article doi In: The Journal of Supercomputing, vol. 79, no. 8, pp. 9206-9244, 2023. @article{GARCIA:JS:23,
title = {Micro-batch and data frequency for stream processing on multi-cores},
author = {Adriano Marques Garcia and Dalvan Griebler and Claudio Schepke and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/s11227-022-05024-y},
doi = {10.1007/s11227-022-05024-y},
year = {2023},
date = {2023-01-01},
journal = {The Journal of Supercomputing},
volume = {79},
number = {8},
pages = {9206-9244},
publisher = {Springer},
abstract = {Latency or throughput is often critical performance metrics in stream processing. Applications’ performance can fluctuate depending on the input stream. This unpredictability is due to the variety in data arrival frequency and size, complexity, and other factors. Researchers are constantly investigating new ways to mitigate the impact of these variations on performance with self-adaptive techniques involving elasticity or micro-batching. However, there is a lack of benchmarks capable of creating test scenarios to further evaluate these techniques. This work extends and improves the SPBench benchmarking framework to support dynamic micro-batching and data stream frequency management. We also propose a set of algorithms that generates the most commonly used frequency patterns for benchmarking stream processing in related work. It allows the creation of a wide variety of test scenarios. To validate our solution, we use SPBench to create custom benchmarks and evaluate the impact of micro-batching and data stream frequency on the performance of Intel TBB and FastFlow. These are two libraries that leverage stream parallelism for multi-core architectures. Our results demonstrated that our test cases did not benefit from micro-batches on multi-cores. For different data stream frequency configurations, TBB ensured the lowest latency, while FastFlow assured higher throughput in shorter pipelines.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Latency or throughput is often critical performance metrics in stream processing. Applications’ performance can fluctuate depending on the input stream. This unpredictability is due to the variety in data arrival frequency and size, complexity, and other factors. Researchers are constantly investigating new ways to mitigate the impact of these variations on performance with self-adaptive techniques involving elasticity or micro-batching. However, there is a lack of benchmarks capable of creating test scenarios to further evaluate these techniques. This work extends and improves the SPBench benchmarking framework to support dynamic micro-batching and data stream frequency management. We also propose a set of algorithms that generates the most commonly used frequency patterns for benchmarking stream processing in related work. It allows the creation of a wide variety of test scenarios. To validate our solution, we use SPBench to create custom benchmarks and evaluate the impact of micro-batching and data stream frequency on the performance of Intel TBB and FastFlow. These are two libraries that leverage stream parallelism for multi-core architectures. Our results demonstrated that our test cases did not benefit from micro-batches on multi-cores. For different data stream frequency configurations, TBB ensured the lowest latency, while FastFlow assured higher throughput in shorter pipelines. |
 | Araujo, Gabriell; Griebler, Dalvan; Rockenbach, Dinei A.; Danelutto, Marco; Fernandes, Luiz Gustavo NAS Parallel Benchmarks with CUDA and Beyond Journal Article doi In: Software: Practice and Experience, vol. 53, no. 1, pp. 53-80, 2023. @article{ARAUJO:SPE:23,
title = {NAS Parallel Benchmarks with CUDA and Beyond},
author = {Gabriell Araujo and Dalvan Griebler and Dinei A. Rockenbach and Marco Danelutto and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1002/spe.3056},
doi = {10.1002/spe.3056},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {Software: Practice and Experience},
volume = {53},
number = {1},
pages = {53-80},
publisher = {Wiley},
abstract = {NAS Parallel Benchmarks (NPB) is a standard benchmark suite used in the evaluation of parallel hardware and software. Several research efforts from academia have made these benchmarks available with different parallel programming models beyond the original versions with OpenMP and MPI. This work joins these research efforts by providing a new CUDA implementation for NPB. Our contribution covers different aspects beyond the implementation. First, we define design principles based on the best programming practices for GPUs and apply them to each benchmark using CUDA. Second, we provide ease of use parametrization support for configuring the number of threads per block in our version. Third, we conduct a broad study on the impact of the number of threads per block in the benchmarks. Fourth, we propose and evaluate five strategies for helping to find a better number of threads per block configuration. The results have revealed relevant performance improvement solely by changing the number of threads per block, showing performance improvements from 8% up to 717% among the benchmarks. Fifth, we conduct a comparative analysis with the literature, evaluating performance, memory consumption, code refactoring required, and parallelism implementations. The performance results have shown up to 267% improvements over the best benchmarks versions available. We also observe the best and worst design choices, concerning code size and the performance trade-off. Lastly, we highlight the challenges of implementing parallel CFD applications for GPUs and how the computations impact the GPU's behavior.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
NAS Parallel Benchmarks (NPB) is a standard benchmark suite used in the evaluation of parallel hardware and software. Several research efforts from academia have made these benchmarks available with different parallel programming models beyond the original versions with OpenMP and MPI. This work joins these research efforts by providing a new CUDA implementation for NPB. Our contribution covers different aspects beyond the implementation. First, we define design principles based on the best programming practices for GPUs and apply them to each benchmark using CUDA. Second, we provide ease of use parametrization support for configuring the number of threads per block in our version. Third, we conduct a broad study on the impact of the number of threads per block in the benchmarks. Fourth, we propose and evaluate five strategies for helping to find a better number of threads per block configuration. The results have revealed relevant performance improvement solely by changing the number of threads per block, showing performance improvements from 8% up to 717% among the benchmarks. Fifth, we conduct a comparative analysis with the literature, evaluating performance, memory consumption, code refactoring required, and parallelism implementations. The performance results have shown up to 267% improvements over the best benchmarks versions available. We also observe the best and worst design choices, concerning code size and the performance trade-off. Lastly, we highlight the challenges of implementing parallel CFD applications for GPUs and how the computations impact the GPU's behavior. |
 | Garcia, Adriano Marques; Griebler, Dalvan; Schepke, Claudio; Fernandes, Luiz Gustavo SPBench: a framework for creating benchmarks of stream processing applications Journal Article doi In: Computing, vol. 105, no. 5, pp. 1077-1099, 2023. @article{GARCIA:Computing:23,
title = {SPBench: a framework for creating benchmarks of stream processing applications},
author = {Adriano Marques Garcia and Dalvan Griebler and Claudio Schepke and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/s00607-021-01025-6},
doi = {10.1007/s00607-021-01025-6},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {Computing},
volume = {105},
number = {5},
pages = {1077-1099},
publisher = {Springer},
abstract = {In a fast-changing data-driven world, real-time data processing systems are becoming ubiquitous in everyday applications. The increasing data we produce, such as audio, video, image, and, text are demanding quickly and efficiently computation. Stream Parallelism allows accelerating this computation for real-time processing. But it is still a challenging task and most reserved for experts. In this paper, we present SPBench, a framework for benchmarking stream processing applications. It aims to support users with a set of real-world stream processing applications, which are made accessible through an Application Programming Interface (API) and executable via Command Line Interface (CLI) to create custom benchmarks. We tested SPBench by implementing parallel benchmarks with Intel Threading Building Blocks (TBB), FastFlow, and SPar. This evaluation provided useful insights and revealed the feasibility of the proposed framework in terms of usage, customization, and performance analysis. SPBench demonstrated to be a high-level, reusable, extensible, and easy of use abstraction to build parallel stream processing benchmarks on multi-core architectures.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In a fast-changing data-driven world, real-time data processing systems are becoming ubiquitous in everyday applications. The increasing data we produce, such as audio, video, image, and, text are demanding quickly and efficiently computation. Stream Parallelism allows accelerating this computation for real-time processing. But it is still a challenging task and most reserved for experts. In this paper, we present SPBench, a framework for benchmarking stream processing applications. It aims to support users with a set of real-world stream processing applications, which are made accessible through an Application Programming Interface (API) and executable via Command Line Interface (CLI) to create custom benchmarks. We tested SPBench by implementing parallel benchmarks with Intel Threading Building Blocks (TBB), FastFlow, and SPar. This evaluation provided useful insights and revealed the feasibility of the proposed framework in terms of usage, customization, and performance analysis. SPBench demonstrated to be a high-level, reusable, extensible, and easy of use abstraction to build parallel stream processing benchmarks on multi-core architectures. |
2022
|
 | Löff, Júnior; Hoffmann, Renato Barreto; Griebler, Dalvan; Fernandes, Luiz Gustavo Combining stream with data parallelism abstractions for multi-cores Journal Article doi In: Journal of Computer Languages, vol. 73, pp. 101160, 2022. @article{LOFF:COLA:22,
title = {Combining stream with data parallelism abstractions for multi-cores},
author = {Júnior Löff and Renato Barreto Hoffmann and Dalvan Griebler and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1016/j.cola.2022.101160},
doi = {10.1016/j.cola.2022.101160},
year = {2022},
date = {2022-12-01},
urldate = {2022-12-01},
journal = {Journal of Computer Languages},
volume = {73},
pages = {101160},
publisher = {Elsevier},
abstract = {Stream processing applications have seen an increasing demand with the raised availability of sensors, IoT devices, and user data. Modern systems can generate millions of data items per day that require to be processed timely. To deal with this demand, application programmers must consider parallelism to exploit the maximum performance of the underlying hardware resources. In this work, we introduce improvements to stream processing applications by exploiting fine-grained data parallelism (via Map and MapReduce) inside coarse-grained stream parallelism stages. The improvements are including techniques for identifying data parallelism in sequential codes, a new language, semantic analysis, and a set of definition and transformation rules to perform source-to-source parallel code generation. Moreover, we investigate the feasibility of employing higher-level programming abstractions to support the proposed optimizations. For that, we elect SPar programming model as a use case, and extend it by adding two new attributes to its language and implementing our optimizations as a new algorithm in the SPar compiler. We conduct a set of experiments in representative stream processing and data-parallel applications. The results showed that our new compiler algorithm is efficient and that performance improved by up to 108.4x in data-parallel applications. Furthermore, experiments evaluating stream processing applications towards the composition of stream and data parallelism revealed new insights. The results showed that such composition may improve latencies by up to an order of magnitude. Also, it enables programmers to exploit different degrees of stream and data parallelism to accomplish a balance between throughput and latency according to their necessity.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Stream processing applications have seen an increasing demand with the raised availability of sensors, IoT devices, and user data. Modern systems can generate millions of data items per day that require to be processed timely. To deal with this demand, application programmers must consider parallelism to exploit the maximum performance of the underlying hardware resources. In this work, we introduce improvements to stream processing applications by exploiting fine-grained data parallelism (via Map and MapReduce) inside coarse-grained stream parallelism stages. The improvements are including techniques for identifying data parallelism in sequential codes, a new language, semantic analysis, and a set of definition and transformation rules to perform source-to-source parallel code generation. Moreover, we investigate the feasibility of employing higher-level programming abstractions to support the proposed optimizations. For that, we elect SPar programming model as a use case, and extend it by adding two new attributes to its language and implementing our optimizations as a new algorithm in the SPar compiler. We conduct a set of experiments in representative stream processing and data-parallel applications. The results showed that our new compiler algorithm is efficient and that performance improved by up to 108.4x in data-parallel applications. Furthermore, experiments evaluating stream processing applications towards the composition of stream and data parallelism revealed new insights. The results showed that such composition may improve latencies by up to an order of magnitude. Also, it enables programmers to exploit different degrees of stream and data parallelism to accomplish a balance between throughput and latency according to their necessity. |
 | Ernstsson, August; Griebler, Dalvan; Kessler, Christoph Assessing Application Efficiency and Performance Portability in Single-Source Programming for Heterogeneous Parallel Systems Journal Article doi In: International Journal of Parallel Programming, vol. 51, no. 5, pp. 61-82, 2022. @article{Ernstsson:IJPP:22,
title = {Assessing Application Efficiency and Performance Portability in Single-Source Programming for Heterogeneous Parallel Systems},
author = {August Ernstsson and Dalvan Griebler and Christoph Kessler},
url = {https://doi.org/10.1007/s10766-022-00746-1},
doi = {10.1007/s10766-022-00746-1},
year = {2022},
date = {2022-12-01},
urldate = {2022-12-01},
journal = {International Journal of Parallel Programming},
volume = {51},
number = {5},
pages = {61-82},
publisher = {Springer},
abstract = {We analyze the performance portability of the skeleton-based, single-source multi-backend high-level programming framework SkePU across multiple different CPU–GPU heterogeneous systems. Thereby, we provide a systematic application efficiency characterization of SkePU-generated code in comparison to equivalent hand-written code in more low-level parallel programming models such as OpenMP and CUDA. For this purpose, we contribute ports of the STREAM benchmark suite and of a part of the NAS Parallel Benchmark suite to SkePU. We show that for STREAM and the EP benchmark, SkePU regularly scores efficiency values above 80% and in particular for CPU systems, SkePU can outperform hand-written code..},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
We analyze the performance portability of the skeleton-based, single-source multi-backend high-level programming framework SkePU across multiple different CPU–GPU heterogeneous systems. Thereby, we provide a systematic application efficiency characterization of SkePU-generated code in comparison to equivalent hand-written code in more low-level parallel programming models such as OpenMP and CUDA. For this purpose, we contribute ports of the STREAM benchmark suite and of a part of the NAS Parallel Benchmark suite to SkePU. We show that for STREAM and the EP benchmark, SkePU regularly scores efficiency values above 80% and in particular for CPU systems, SkePU can outperform hand-written code.. |
| Andrade, Gabriella; Griebler, Dalvan; Santos, Rodrigo; Fernandes, Luiz Gustavo Opinião de Brasileiros Sobre a Produtividade no Desenvolvimento de Aplicações Paralelas Inproceedings doi In: Anais do XXII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD), pp. 276-287, SBC, Florianópolis, Brasil, 2022. @inproceedings{ANDRADE:WSCAD:22,
title = {Opinião de Brasileiros Sobre a Produtividade no Desenvolvimento de Aplicações Paralelas},
author = {Gabriella Andrade and Dalvan Griebler and Rodrigo Santos and Luiz Gustavo Fernandes},
url = {https://doi.org/10.5753/wscad.2022.226392},
doi = {10.5753/wscad.2022.226392},
year = {2022},
date = {2022-10-01},
booktitle = {Anais do XXII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD)},
pages = {276-287},
publisher = {SBC},
address = {Florianópolis, Brasil},
abstract = {A partir da popularização das arquiteturas paralelas, surgiram várias interfaces de programação a fim de facilitar a exploração de tais arquiteturas e aumentar a produtividade dos desenvolvedores. Entretanto, desenvolver aplicações paralelas ainda é uma tarefa complexa para desenvolvedores com pouca experiência. Neste trabalho, realizamos uma pesquisa para descobrir a opinião de desenvolvedores de aplicações paralelas sobre os fatores que impedem a produtividade. Nossos resultados mostraram que a experiência dos desenvolvedores é uma das principais razões para a baixa produtividade. Além disso, os resultados indicaram formas para contornar este problema, como melhorar e incentivar o ensino de programação paralela em cursos de graduação.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
A partir da popularização das arquiteturas paralelas, surgiram várias interfaces de programação a fim de facilitar a exploração de tais arquiteturas e aumentar a produtividade dos desenvolvedores. Entretanto, desenvolver aplicações paralelas ainda é uma tarefa complexa para desenvolvedores com pouca experiência. Neste trabalho, realizamos uma pesquisa para descobrir a opinião de desenvolvedores de aplicações paralelas sobre os fatores que impedem a produtividade. Nossos resultados mostraram que a experiência dos desenvolvedores é uma das principais razões para a baixa produtividade. Além disso, os resultados indicaram formas para contornar este problema, como melhorar e incentivar o ensino de programação paralela em cursos de graduação. |
| Rockenbach, Dinei A.; Löff, Júnior; Araujo, Gabriell; Griebler, Dalvan; Fernandes, Luiz G. High-Level Stream and Data Parallelism in C++ for GPUs Inproceedings doi In: XXVI Brazilian Symposium on Programming Languages (SBLP), pp. 41-49, ACM, Uberlândia, Brazil, 2022. @inproceedings{ROCKENBACH:SBLP:22,
title = {High-Level Stream and Data Parallelism in C++ for GPUs},
author = {Dinei A. Rockenbach and Júnior Löff and Gabriell Araujo and Dalvan Griebler and Luiz G. Fernandes},
url = {https://doi.org/10.1145/3561320.3561327},
doi = {10.1145/3561320.3561327},
year = {2022},
date = {2022-10-01},
booktitle = {XXVI Brazilian Symposium on Programming Languages (SBLP)},
pages = {41-49},
publisher = {ACM},
address = {Uberlândia, Brazil},
series = {SBLP'22},
abstract = {GPUs are massively parallel processors that allow solving problems that are not viable to traditional processors like CPUs. However, implementing applications for GPUs is challenging to programmers as it requires parallel programming to efficiently exploit the GPU resources. In this sense, parallel programming abstractions, notably domain-specific languages, are fundamental for improving programmability. SPar is a high-level Domain-Specific Language (DSL) that allows expressing stream and data parallelism in the serial code through annotations using C++ attributes. This work elaborates on a methodology and tool for GPU code generation by introducing new attributes to SPar language and transformation rules to SPar compiler. These new contributions, besides the gains in simplicity and code reduction compared to CUDA and OpenCL, enabled SPar achieve of higher throughput when exploring combined CPU and GPU parallelism, and when using batching.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
GPUs are massively parallel processors that allow solving problems that are not viable to traditional processors like CPUs. However, implementing applications for GPUs is challenging to programmers as it requires parallel programming to efficiently exploit the GPU resources. In this sense, parallel programming abstractions, notably domain-specific languages, are fundamental for improving programmability. SPar is a high-level Domain-Specific Language (DSL) that allows expressing stream and data parallelism in the serial code through annotations using C++ attributes. This work elaborates on a methodology and tool for GPU code generation by introducing new attributes to SPar language and transformation rules to SPar compiler. These new contributions, besides the gains in simplicity and code reduction compared to CUDA and OpenCL, enabled SPar achieve of higher throughput when exploring combined CPU and GPU parallelism, and when using batching. |
| Andrade, Gabriella; Griebler, Dalvan; Santos, Rodrigo; Kessler, Christoph; Ernstsson, August; Fernandes, Luiz Gustavo Analyzing Programming Effort Model Accuracy of High-Level Parallel Programs for Stream Processing Inproceedings doi In: 48th Euromicro Conference on Software Engineering and Advanced Applications (SEAA 2022), pp. 229-232, IEEE, Gran Canaria, Spain, 2022. @inproceedings{ANDRADE:SEAA:22,
title = {Analyzing Programming Effort Model Accuracy of High-Level Parallel Programs for Stream Processing},
author = {Gabriella Andrade and Dalvan Griebler and Rodrigo Santos and Christoph Kessler and August Ernstsson and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/SEAA56994.2022.00043},
doi = {10.1109/SEAA56994.2022.00043},
year = {2022},
date = {2022-09-01},
booktitle = {48th Euromicro Conference on Software Engineering and Advanced Applications (SEAA 2022)},
pages = {229-232},
publisher = {IEEE},
address = {Gran Canaria, Spain},
series = {SEAA'22},
abstract = {Over the years, several Parallel Programming Models (PPMs) have supported the abstraction of programming complexity for parallel computer systems. However, few studies aim to evaluate the productivity reached by such abstractions since this is a complex task that involves human beings. There are several studies to develop predictive methods to estimate the effort required to program applications in software engineering. In order to evaluate the reliability of such metrics, it is necessary to assess the accuracy in different programming domains. In this work, we used the data of an experiment conducted with beginners in parallel programming to determine the effort required for implementing stream parallelism using FastFlow, SPar, and TBB. Our results show that some traditional software effort estimation models, such as COCOMO II, fall short, while Putnam's model could be an alternative for high-level PPMs evaluation. To overcome the limitations of existing models, we plan to create a parallelism-aware model to evaluate applications in this domain in future work.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Over the years, several Parallel Programming Models (PPMs) have supported the abstraction of programming complexity for parallel computer systems. However, few studies aim to evaluate the productivity reached by such abstractions since this is a complex task that involves human beings. There are several studies to develop predictive methods to estimate the effort required to program applications in software engineering. In order to evaluate the reliability of such metrics, it is necessary to assess the accuracy in different programming domains. In this work, we used the data of an experiment conducted with beginners in parallel programming to determine the effort required for implementing stream parallelism using FastFlow, SPar, and TBB. Our results show that some traditional software effort estimation models, such as COCOMO II, fall short, while Putnam's model could be an alternative for high-level PPMs evaluation. To overcome the limitations of existing models, we plan to create a parallelism-aware model to evaluate applications in this domain in future work. |
| Garcia, Adriano Marques; Griebler, Dalvan; Schepke, Claudio; Fernandes, Luiz Gustavo Evaluating Micro-batch and Data Frequency for Stream Processing Applications on Multi-cores Inproceedings doi In: 30th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP), pp. 10-17, IEEE, Valladolid, Spain, 2022. @inproceedings{GARCIA:PDP:22,
title = {Evaluating Micro-batch and Data Frequency for Stream Processing Applications on Multi-cores},
author = {Adriano Marques Garcia and Dalvan Griebler and Claudio Schepke and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/PDP55904.2022.00011},
doi = {10.1109/PDP55904.2022.00011},
year = {2022},
date = {2022-04-01},
booktitle = {30th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP)},
pages = {10-17},
publisher = {IEEE},
address = {Valladolid, Spain},
series = {PDP'22},
abstract = {In stream processing, data arrives constantly and is often unpredictable. It can show large fluctuations in arrival frequency, size, complexity, and other factors. These fluctuations can strongly impact application latency and throughput, which are critical factors in this domain. Therefore, there is a significant amount of research on self-adaptive techniques involving elasticity or micro-batching as a way to mitigate this impact. However, there is a lack of benchmarks and tools for helping researchers to investigate micro-batching and data stream frequency implications. In this paper, we extend a benchmarking framework to support dynamic micro-batching and data stream frequency management. We used it to create custom benchmarks and compare latency and throughput aspects from two different parallel libraries. We validate our solution through an extensive analysis of the impact of micro-batching and data stream frequency on stream processing applications using Intel TBB and FastFlow, which are two libraries that leverage stream parallelism on multi-core architectures. Our results demonstrated up to 33% throughput gain over latency using micro-batches. Additionally, while TBB ensures lower latency, FastFlow ensures higher throughput in the parallel applications for different data stream frequency configurations.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In stream processing, data arrives constantly and is often unpredictable. It can show large fluctuations in arrival frequency, size, complexity, and other factors. These fluctuations can strongly impact application latency and throughput, which are critical factors in this domain. Therefore, there is a significant amount of research on self-adaptive techniques involving elasticity or micro-batching as a way to mitigate this impact. However, there is a lack of benchmarks and tools for helping researchers to investigate micro-batching and data stream frequency implications. In this paper, we extend a benchmarking framework to support dynamic micro-batching and data stream frequency management. We used it to create custom benchmarks and compare latency and throughput aspects from two different parallel libraries. We validate our solution through an extensive analysis of the impact of micro-batching and data stream frequency on stream processing applications using Intel TBB and FastFlow, which are two libraries that leverage stream parallelism on multi-core architectures. Our results demonstrated up to 33% throughput gain over latency using micro-batches. Additionally, while TBB ensures lower latency, FastFlow ensures higher throughput in the parallel applications for different data stream frequency configurations. |
| Mencagli, Gabriele; Griebler, Dalvan; Danelutto, Marco Towards Parallel Data Stream Processing on System-on-Chip CPU+GPU Devices Inproceedings doi In: 30th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP), pp. 34-38, IEEE, Valladolid, Spain, 2022. @inproceedings{MENCAGLI:PDP:22,
title = {Towards Parallel Data Stream Processing on System-on-Chip CPU+GPU Devices},
author = {Gabriele Mencagli and Dalvan Griebler and Marco Danelutto},
url = {https://doi.org/10.1109/PDP55904.2022.00014},
doi = {10.1109/PDP55904.2022.00014},
year = {2022},
date = {2022-04-01},
booktitle = {30th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP)},
pages = {34-38},
publisher = {IEEE},
address = {Valladolid, Spain},
series = {PDP'22},
abstract = {Data Stream Processing is a pervasive computing paradigm with a wide spectrum of applications. Traditional streaming systems exploit the processing capabilities provided by homogeneous Clusters and Clouds. Due to the transition to streaming systems suitable for IoT/Edge environments, there has been the urgent need of new streaming frameworks and tools tailored for embedded platforms, often available as System-onChips composed of a small multicore CPU and an integrated onchip GPU. Exploiting this hybrid hardware requires special care in the runtime system design. In this paper, we discuss the support provided by the WindFlow library, showing its design principles and its effectiveness on the NVIDIA Jetson Nano board.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Data Stream Processing is a pervasive computing paradigm with a wide spectrum of applications. Traditional streaming systems exploit the processing capabilities provided by homogeneous Clusters and Clouds. Due to the transition to streaming systems suitable for IoT/Edge environments, there has been the urgent need of new streaming frameworks and tools tailored for embedded platforms, often available as System-onChips composed of a small multicore CPU and an integrated onchip GPU. Exploiting this hybrid hardware requires special care in the runtime system design. In this paper, we discuss the support provided by the WindFlow library, showing its design principles and its effectiveness on the NVIDIA Jetson Nano board. |
 | Gomes, Márcio Miguel; Righi, Rodrigo Rosa; Costa, Cristiano André; Griebler, Dalvan Steam++: An Extensible End-to-end Framework for Developing IoT Data Processing Applications in the Fog Journal Article doi In: International Journal of Computer Science & Information Technology, vol. 14, no. 1, pp. 31-51, 2022. @article{GOMES:IJCSIT:22,
title = {Steam++: An Extensible End-to-end Framework for Developing IoT Data Processing Applications in the Fog},
author = {Márcio Miguel Gomes and Rodrigo Rosa Righi and Cristiano André Costa and Dalvan Griebler},
url = {http://dx.doi.org/10.5121/ijcsit.2022.14103},
doi = {10.5121/ijcsit.2022.14103},
year = {2022},
date = {2022-02-01},
urldate = {2022-02-01},
journal = {International Journal of Computer Science & Information Technology},
volume = {14},
number = {1},
pages = {31-51},
publisher = {AIRCC},
abstract = {IoT applications usually rely on cloud computing services to perform data analysis such as filtering, aggregation, classification, pattern detection, and prediction. When applied to specific domains, the IoT needs to deal with unique constraints. Besides the hostile environment such as vibration and electricmagnetic interference, resulting in malfunction, noise, and data loss, industrial plants often have Internet access restricted or unavailable, forcing us to design stand-alone fog and edge computing solutions. In this context, we present STEAM++, a lightweight and extensible framework for real-time data stream processing and decision-making in the network edge, targeting hardware-limited devices, besides proposing a micro-benchmark methodology for assessing embedded IoT applications. In real-case experiments in a semiconductor industry, we processed an entire data flow, from values sensing, processing and analysing data, detecting relevant events, and finally, publishing results to a dashboard. On average, the application consumed less than 500kb RAM and 1.0% of CPU usage, processing up to 239 data packets per second and reducing the output data size to 14% of the input raw data size when notifying events.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
IoT applications usually rely on cloud computing services to perform data analysis such as filtering, aggregation, classification, pattern detection, and prediction. When applied to specific domains, the IoT needs to deal with unique constraints. Besides the hostile environment such as vibration and electricmagnetic interference, resulting in malfunction, noise, and data loss, industrial plants often have Internet access restricted or unavailable, forcing us to design stand-alone fog and edge computing solutions. In this context, we present STEAM++, a lightweight and extensible framework for real-time data stream processing and decision-making in the network edge, targeting hardware-limited devices, besides proposing a micro-benchmark methodology for assessing embedded IoT applications. In real-case experiments in a semiconductor industry, we processed an entire data flow, from values sensing, processing and analysing data, detecting relevant events, and finally, publishing results to a dashboard. On average, the application consumed less than 500kb RAM and 1.0% of CPU usage, processing up to 239 data packets per second and reducing the output data size to 14% of the input raw data size when notifying events. |
 | Löff, Júnior; Hoffmann, Renato Barreto; Pieper, Ricardo; Griebler, Dalvan; Fernandes, Luiz Gustavo DSParLib: A C++ Template Library for Distributed Stream Parallelism Journal Article doi In: International Journal of Parallel Programming, vol. 50, no. 5, pp. 454-485, 2022. @article{LOFF:IJPP:22,
title = {DSParLib: A C++ Template Library for Distributed Stream Parallelism},
author = {Júnior Löff and Renato Barreto Hoffmann and Ricardo Pieper and Dalvan Griebler and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/s10766-022-00737-2},
doi = {10.1007/s10766-022-00737-2},
year = {2022},
date = {2022-01-01},
journal = {International Journal of Parallel Programming},
volume = {50},
number = {5},
pages = {454-485},
publisher = {Springer},
abstract = {Stream processing applications deal with millions of data items continuously generated over time. Often, they must be processed in real-time and scale performance, which requires the use of distributed parallel computing resources. In C/C++, the current state-of-the-art for distributed architectures and High-Performance Computing is Message Passing Interface (MPI). However, exploiting stream parallelism using MPI is complex and error-prone because it exposes many low-level details to the programmer. In this work, we introduce a new parallel programming abstraction for implementing distributed stream parallelism named DSParLib. Our abstraction of MPI simplifies parallel programming by providing a pattern-based and building block-oriented development to inter-connect, model, and parallelize data streams found in modern applications. Experiments conducted with five different stream processing applications and the representative PARSEC Ferret benchmark revealed that DSParLib is efficient and flexible. Also, DSParLib achieved similar or better performance, required less coding, and provided simpler abstractions to express parallelism with respect to handwritten MPI programs.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Stream processing applications deal with millions of data items continuously generated over time. Often, they must be processed in real-time and scale performance, which requires the use of distributed parallel computing resources. In C/C++, the current state-of-the-art for distributed architectures and High-Performance Computing is Message Passing Interface (MPI). However, exploiting stream parallelism using MPI is complex and error-prone because it exposes many low-level details to the programmer. In this work, we introduce a new parallel programming abstraction for implementing distributed stream parallelism named DSParLib. Our abstraction of MPI simplifies parallel programming by providing a pattern-based and building block-oriented development to inter-connect, model, and parallelize data streams found in modern applications. Experiments conducted with five different stream processing applications and the representative PARSEC Ferret benchmark revealed that DSParLib is efficient and flexible. Also, DSParLib achieved similar or better performance, required less coding, and provided simpler abstractions to express parallelism with respect to handwritten MPI programs. |
 | Hoffmann, Renato Barreto; Löff, Júnior; Griebler, Dalvan; Fernandes, Luiz Gustavo OpenMP as runtime for providing high-level stream parallelism on multi-cores Journal Article doi In: The Journal of Supercomputing, vol. 78, no. 1, pp. 7655-7676, 2022. @article{HOFFMANN:Jsuper:2022,
title = {OpenMP as runtime for providing high-level stream parallelism on multi-cores},
author = {Renato Barreto Hoffmann and Júnior Löff and Dalvan Griebler and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/s11227-021-04182-9},
doi = {10.1007/s11227-021-04182-9},
year = {2022},
date = {2022-01-01},
journal = {The Journal of Supercomputing},
volume = {78},
number = {1},
pages = {7655-7676},
publisher = {Springer},
address = {New York, United States},
abstract = {OpenMP is an industry and academic standard for parallel programming. However, using it for developing parallel stream processing applications is complex and challenging. OpenMP lacks key programming mechanisms and abstractions for this particular domain. To tackle this problem, we used a high-level parallel programming framework (named SPar) for automatically generating parallel OpenMP code. We achieved this by leveraging SPar’s language and its domain-specific code annotations for simplifying the complexity and verbosity added by OpenMP in this application domain. Consequently, we implemented a new compiler algorithm in SPar for automatically generating parallel code targeting the OpenMP runtime using source-to-source code transformations. The experiments in four different stream processing applications demonstrated that the execution time of SPar was improved up to 25.42% when using the OpenMP runtime. Additionally, our abstraction over OpenMP introduced at most 1.72% execution time overhead when compared to handwritten parallel codes. Furthermore, SPar significantly reduces the total source lines of code required to express parallelism with respect to plain OpenMP parallel codes.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
OpenMP is an industry and academic standard for parallel programming. However, using it for developing parallel stream processing applications is complex and challenging. OpenMP lacks key programming mechanisms and abstractions for this particular domain. To tackle this problem, we used a high-level parallel programming framework (named SPar) for automatically generating parallel OpenMP code. We achieved this by leveraging SPar’s language and its domain-specific code annotations for simplifying the complexity and verbosity added by OpenMP in this application domain. Consequently, we implemented a new compiler algorithm in SPar for automatically generating parallel code targeting the OpenMP runtime using source-to-source code transformations. The experiments in four different stream processing applications demonstrated that the execution time of SPar was improved up to 25.42% when using the OpenMP runtime. Additionally, our abstraction over OpenMP introduced at most 1.72% execution time overhead when compared to handwritten parallel codes. Furthermore, SPar significantly reduces the total source lines of code required to express parallelism with respect to plain OpenMP parallel codes. |
2021
|
| Löff, Júnior; Hoffmann, Renato Barreto; Griebler, Dalvan; Fernandes, Luiz G. High-Level Stream and Data Parallelism in C++ for Multi-Cores Inproceedings doi In: XXV Brazilian Symposium on Programming Languages (SBLP), pp. 41-48, ACM, Joinville, Brazil, 2021. @inproceedings{LOFF:SBLP:21,
title = {High-Level Stream and Data Parallelism in C++ for Multi-Cores},
author = {Júnior Löff and Renato Barreto Hoffmann and Dalvan Griebler and Luiz G. Fernandes},
url = {https://doi.org/10.1145/3475061.3475078},
doi = {10.1145/3475061.3475078},
year = {2021},
date = {2021-10-01},
booktitle = {XXV Brazilian Symposium on Programming Languages (SBLP)},
pages = {41-48},
publisher = {ACM},
address = {Joinville, Brazil},
series = {SBLP'21},
abstract = {Stream processing applications have seen an increasing demand with the increased availability of sensors, IoT devices, and user data. Modern systems can generate millions of data items per day that require to be processed timely. To deal with this demand, application programmers must consider parallelism to exploit the maximum performance of the underlying hardware resources. However, parallel programming is often difficult and error-prone, because programmers must deal with low-level system and architecture details. In this work, we introduce a new strategy for automatic data-parallel code generation in C++ targeting multi-core architectures. This strategy was integrated with an annotation-based parallel programming abstraction named SPar. We have increased SPar’s expressiveness for supporting stream and data parallelism, and their arbitrary composition. Therefore, we added two new attributes to its language and improved the compiler parallel code generation. We conducted a set of experiments on different stream and data-parallel applications to assess the efficiency of our solution. The results showed that the new SPar version obtained similar performance with respect to handwritten parallelizations. Moreover, the new SPar version is able to achieve up to 74.9x better performance with respect to the original ones due to this work.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Stream processing applications have seen an increasing demand with the increased availability of sensors, IoT devices, and user data. Modern systems can generate millions of data items per day that require to be processed timely. To deal with this demand, application programmers must consider parallelism to exploit the maximum performance of the underlying hardware resources. However, parallel programming is often difficult and error-prone, because programmers must deal with low-level system and architecture details. In this work, we introduce a new strategy for automatic data-parallel code generation in C++ targeting multi-core architectures. This strategy was integrated with an annotation-based parallel programming abstraction named SPar. We have increased SPar’s expressiveness for supporting stream and data parallelism, and their arbitrary composition. Therefore, we added two new attributes to its language and improved the compiler parallel code generation. We conducted a set of experiments on different stream and data-parallel applications to assess the efficiency of our solution. The results showed that the new SPar version obtained similar performance with respect to handwritten parallelizations. Moreover, the new SPar version is able to achieve up to 74.9x better performance with respect to the original ones due to this work. |
| Andrade, Gabriella; Griebler, Dalvan; Santos, Rodrigo; Danelutto, Marco; Fernandes, Luiz Gustavo Assessing Coding Metrics for Parallel Programming of Stream Processing Programs on Multi-cores Inproceedings doi In: 47th Euromicro Conference on Software Engineering and Advanced Applications (SEAA 2021), pp. 291-295, IEEE, Pavia, Italy, 2021. @inproceedings{ANDRADE:SEAA:21,
title = {Assessing Coding Metrics for Parallel Programming of Stream Processing Programs on Multi-cores},
author = {Gabriella Andrade and Dalvan Griebler and Rodrigo Santos and Marco Danelutto and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/SEAA53835.2021.00044},
doi = {10.1109/SEAA53835.2021.00044},
year = {2021},
date = {2021-09-01},
booktitle = {47th Euromicro Conference on Software Engineering and Advanced Applications (SEAA 2021)},
pages = {291-295},
publisher = {IEEE},
address = {Pavia, Italy},
series = {SEAA'21},
abstract = {From the popularization of multi-core architectures, several parallel APIs have emerged, helping to abstract the programming complexity and increasing productivity in application development. Unfortunately, only a few research efforts in this direction managed to show the usability pay-back of the programming abstraction created, because it is not easy and poses many challenges for conducting empirical software engineering. We believe that coding metrics commonly used in software engineering code measurements can give useful indicators on the programming effort of parallel applications and APIs. These metrics were designed for general purposes without considering the evaluation of applications from a specific domain. In this study, we aim to evaluate the feasibility of seven coding metrics to be used in the parallel programming domain. To do so, five stream processing applications implemented with different parallel APIs for multi-cores were considered. Our experiments have shown COCOMO II is a suitable model for evaluating the productivity of different parallel APIs targeting multi-cores on stream processing applications while other metrics are restricted to the code size.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
From the popularization of multi-core architectures, several parallel APIs have emerged, helping to abstract the programming complexity and increasing productivity in application development. Unfortunately, only a few research efforts in this direction managed to show the usability pay-back of the programming abstraction created, because it is not easy and poses many challenges for conducting empirical software engineering. We believe that coding metrics commonly used in software engineering code measurements can give useful indicators on the programming effort of parallel applications and APIs. These metrics were designed for general purposes without considering the evaluation of applications from a specific domain. In this study, we aim to evaluate the feasibility of seven coding metrics to be used in the parallel programming domain. To do so, five stream processing applications implemented with different parallel APIs for multi-cores were considered. Our experiments have shown COCOMO II is a suitable model for evaluating the productivity of different parallel APIs targeting multi-cores on stream processing applications while other metrics are restricted to the code size. |
 | Löff, Júnior; Griebler, Dalvan; Mencagli, Gabriele; Araujo, Gabriell; Torquati, Massimo; Danelutto, Marco; Fernandes, Luiz Gustavo The NAS parallel benchmarks for evaluating C++ parallel programming frameworks on shared-memory architectures Journal Article doi In: Future Generation Computer Systems, vol. 125, pp. 743-757, 2021. @article{LOFF:FGCS:21,
title = {The NAS parallel benchmarks for evaluating C++ parallel programming frameworks on shared-memory architectures},
author = {Júnior Löff and Dalvan Griebler and Gabriele Mencagli and Gabriell Araujo and Massimo Torquati and Marco Danelutto and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1016/j.future.2021.07.021},
doi = {10.1016/j.future.2021.07.021},
year = {2021},
date = {2021-07-01},
journal = {Future Generation Computer Systems},
volume = {125},
pages = {743-757},
publisher = {Elsevier},
abstract = {The NAS Parallel Benchmarks (NPB), originally implemented mostly in Fortran, is a consolidated suite containing several benchmarks extracted from Computational Fluid Dynamics (CFD) models. The benchmark suite has important characteristics such as intensive memory communications, complex data dependencies, different memory access patterns, and hardware components/sub-systems overload. Parallel programming APIs, libraries, and frameworks that are written in C++ as well as new optimizations and parallel processing techniques can benefit if NPB is made fully available in this programming language. In this paper we present NPB-CPP, a fully C++ translated version of NPB consisting of all the NPB kernels and pseudo-applications developed using OpenMP, Intel TBB, and FastFlow parallel frameworks for multicores. The design of NPB-CPP leverages the Structured Parallel Programming methodology (essentially based on parallel design patterns). We show the structure of each benchmark application in terms of composition of few patterns (notably Map and MapReduce constructs) provided by the selected C++ frameworks. The experimental evaluation shows the accuracy of NPB-CPP with respect to the original NPB source code. Furthermore, we carefully evaluate the parallel performance on three multi-core systems (Intel, IBM Power and AMD) with different C++ compilers (gcc, icc and clang) by discussing the performance differences in order to give to the researchers useful insights to choose the best parallel programming framework for a given type of problem.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The NAS Parallel Benchmarks (NPB), originally implemented mostly in Fortran, is a consolidated suite containing several benchmarks extracted from Computational Fluid Dynamics (CFD) models. The benchmark suite has important characteristics such as intensive memory communications, complex data dependencies, different memory access patterns, and hardware components/sub-systems overload. Parallel programming APIs, libraries, and frameworks that are written in C++ as well as new optimizations and parallel processing techniques can benefit if NPB is made fully available in this programming language. In this paper we present NPB-CPP, a fully C++ translated version of NPB consisting of all the NPB kernels and pseudo-applications developed using OpenMP, Intel TBB, and FastFlow parallel frameworks for multicores. The design of NPB-CPP leverages the Structured Parallel Programming methodology (essentially based on parallel design patterns). We show the structure of each benchmark application in terms of composition of few patterns (notably Map and MapReduce constructs) provided by the selected C++ frameworks. The experimental evaluation shows the accuracy of NPB-CPP with respect to the original NPB source code. Furthermore, we carefully evaluate the parallel performance on three multi-core systems (Intel, IBM Power and AMD) with different C++ compilers (gcc, icc and clang) by discussing the performance differences in order to give to the researchers useful insights to choose the best parallel programming framework for a given type of problem. |
 | Pieper, Ricardo; Löff, Júnior; Hoffmann, Renato Berreto; Griebler, Dalvan; Fernandes, Luiz Gustavo High-level and Efficient Structured Stream Parallelism for Rust on Multi-cores Journal Article doi In: Journal of Computer Languages, vol. 65, pp. 101054, 2021. @article{PIEPER:COLA:21,
title = {High-level and Efficient Structured Stream Parallelism for Rust on Multi-cores},
author = {Ricardo Pieper and Júnior Löff and Renato Berreto Hoffmann and Dalvan Griebler and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1016/j.cola.2021.101054},
doi = {10.1016/j.cola.2021.101054},
year = {2021},
date = {2021-07-01},
journal = {Journal of Computer Languages},
volume = {65},
pages = {101054},
publisher = {Elsevier},
abstract = {This work aims at contributing with a structured parallel programming abstraction for Rust in order to provide ready-to-use parallel patterns that abstract low-level and architecture-dependent details from application programmers. We focus on stream processing applications running on shared-memory multi-core architectures (i.e, video processing, compression, and others). Therefore, we provide a new high-level and efficient parallel programming abstraction for expressing stream parallelism, named Rust-SSP. We also created a new stream benchmark suite for Rust that represents real-world scenarios and has different application characteristics and workloads. Our benchmark suite is an initiative to assess existing parallelism abstraction for this domain, as parallel implementations using these abstractions were provided. The results revealed that Rust-SSP achieved up to 41.1% better performance than other solutions. In terms of programmability, the results revealed that Rust-SSP requires the smallest number of extra lines of code to enable stream parallelism..},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This work aims at contributing with a structured parallel programming abstraction for Rust in order to provide ready-to-use parallel patterns that abstract low-level and architecture-dependent details from application programmers. We focus on stream processing applications running on shared-memory multi-core architectures (i.e, video processing, compression, and others). Therefore, we provide a new high-level and efficient parallel programming abstraction for expressing stream parallelism, named Rust-SSP. We also created a new stream benchmark suite for Rust that represents real-world scenarios and has different application characteristics and workloads. Our benchmark suite is an initiative to assess existing parallelism abstraction for this domain, as parallel implementations using these abstractions were provided. The results revealed that Rust-SSP achieved up to 41.1% better performance than other solutions. In terms of programmability, the results revealed that Rust-SSP requires the smallest number of extra lines of code to enable stream parallelism.. |
 | Gomes, Márcio Miguel; Righi, Rodrigo Rosa; Costa, Cristiano André; Griebler, Dalvan Simplifying IoT data stream enrichment and analytics in the edge Journal Article doi In: Computers & Electrical Engineering, vol. 92, pp. 107110, 2021. @article{GOMES:CEE:21,
title = {Simplifying IoT data stream enrichment and analytics in the edge},
author = {Márcio Miguel Gomes and Rodrigo Rosa Righi and Cristiano André Costa and Dalvan Griebler},
url = {https://doi.org/10.1016/j.compeleceng.2021.107110},
doi = {10.1016/j.compeleceng.2021.107110},
year = {2021},
date = {2021-06-01},
urldate = {2021-06-01},
journal = {Computers & Electrical Engineering},
volume = {92},
pages = {107110},
publisher = {Elsevier},
abstract = {Edge devices are usually limited in resources. They often send data to the cloud, where techniques such as filtering, aggregation, classification, pattern detection, and prediction are performed. This process results in critical issues such as data loss, high response time, and overhead. On the other hand, processing data in the edge is not a simple task due to devices’ heterogeneity, resource limitations, a variety of programming languages and standards. In this context, this work proposes STEAM, a framework for developing data stream processing applications in the edge targeting hardware-limited devices. As the main contribution, STEAM enables the development of applications for different platforms, with standardized functions and class structures that use consolidated IoT data formats and communication protocols. Moreover, the experiments revealed the viability of stream processing in the edge resulting in the reduction of response time without compromising the quality of results.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Edge devices are usually limited in resources. They often send data to the cloud, where techniques such as filtering, aggregation, classification, pattern detection, and prediction are performed. This process results in critical issues such as data loss, high response time, and overhead. On the other hand, processing data in the edge is not a simple task due to devices’ heterogeneity, resource limitations, a variety of programming languages and standards. In this context, this work proposes STEAM, a framework for developing data stream processing applications in the edge targeting hardware-limited devices. As the main contribution, STEAM enables the development of applications for different platforms, with standardized functions and class structures that use consolidated IoT data formats and communication protocols. Moreover, the experiments revealed the viability of stream processing in the edge resulting in the reduction of response time without compromising the quality of results. |
 | Vogel, Adriano; Mencagli, Gabriele; Griebler, Dalvan; Danelutto, Marco; Fernandes, Luiz Gustavo Online and Transparent Self-adaptation of Stream Parallel Patterns Journal Article doi In: Computing, vol. 105, no. 5, pp. 1039-1057, 2021. @article{VOGEL:Computing:23,
title = {Online and Transparent Self-adaptation of Stream Parallel Patterns},
author = {Adriano Vogel and Gabriele Mencagli and Dalvan Griebler and Marco Danelutto and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/s00607-021-00998-8},
doi = {10.1007/s00607-021-00998-8},
year = {2021},
date = {2021-05-01},
journal = {Computing},
volume = {105},
number = {5},
pages = {1039-1057},
publisher = {Springer},
abstract = {Several real-world parallel applications are becoming more dynamic and long-running, demanding online (at run-time) adaptations. Stream processing is a representative scenario that computes data items arriving in real-time and where parallel executions are necessary. However, it is challenging for humans to monitor and manually self-optimize complex and long-running parallel executions continuously. Moreover, although high-level and structured parallel programming aims to facilitate parallelism, several issues still need to be addressed for improving the existing abstractions. In this paper, we extend self-adaptiveness for supporting autonomous and online changes of the parallel pattern compositions. Online self-adaptation is achieved with an online profiler that characterizes the applications, which is combined with a new self-adaptive strategy and a model for smooth transitions on reconfigurations. The solution provides a new abstraction layer that enables application programmers to define non-functional requirements instead of hand-tuning complex configurations. Hence, we contribute with additional abstractions and flexible self-adaptation for responsiveness at run-time. The proposed solution is evaluated with applications having different processing characteristics, workloads, and configurations. The results show that it is possible to provide additional abstractions, flexibility, and responsiveness while achieving performance comparable to the best static configuration executions.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Several real-world parallel applications are becoming more dynamic and long-running, demanding online (at run-time) adaptations. Stream processing is a representative scenario that computes data items arriving in real-time and where parallel executions are necessary. However, it is challenging for humans to monitor and manually self-optimize complex and long-running parallel executions continuously. Moreover, although high-level and structured parallel programming aims to facilitate parallelism, several issues still need to be addressed for improving the existing abstractions. In this paper, we extend self-adaptiveness for supporting autonomous and online changes of the parallel pattern compositions. Online self-adaptation is achieved with an online profiler that characterizes the applications, which is combined with a new self-adaptive strategy and a model for smooth transitions on reconfigurations. The solution provides a new abstraction layer that enables application programmers to define non-functional requirements instead of hand-tuning complex configurations. Hence, we contribute with additional abstractions and flexible self-adaptation for responsiveness at run-time. The proposed solution is evaluated with applications having different processing characteristics, workloads, and configurations. The results show that it is possible to provide additional abstractions, flexibility, and responsiveness while achieving performance comparable to the best static configuration executions. |
 | Vogel, Adriano; Griebler, Dalvan; Danelutto, Marco; Fernandes, Luiz Gustavo Self-adaptation on Parallel Stream Processing: A Systematic Review Journal Article doi In: Concurrency and Computation: Practice and Experience, vol. 34, no. 6, pp. e6759, 2021. @article{VOGEL:Survey:CCPE:2021,
title = {Self-adaptation on Parallel Stream Processing: A Systematic Review},
author = {Adriano Vogel and Dalvan Griebler and Marco Danelutto and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1002/cpe.6759},
doi = {10.1002/cpe.6759},
year = {2021},
date = {2021-03-01},
urldate = {2021-03-01},
journal = {Concurrency and Computation: Practice and Experience},
volume = {34},
number = {6},
pages = {e6759},
publisher = {Wiley},
abstract = {A recurrent challenge in real-world applications is autonomous management of the executions at run-time. In this vein, stream processing is a class of applications that compute data flowing in the form of streams (e.g., video feeds, images, and data analytics), where parallel computing can help accelerate the executions. On the one hand, stream processing applications are becoming more complex, dynamic, and long-running. On the other hand, it is unfeasible for humans to monitor and manually change the executions continuously. Hence, self-adaptation can reduce costs and human efforts by providing a higher-level abstraction with an autonomic/seamless management of executions. In this work, we aim at providing a literature review regarding self-adaptation applied to the parallel stream processing domain. We present a comprehensive revision using a systematic literature review method. Moreover, we propose a taxonomy to categorize and classify the existing self-adaptive approaches. Finally, applying the taxonomy made it possible to characterize the state-of-the-art, identify trends, and discuss open research challenges and future opportunities.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
A recurrent challenge in real-world applications is autonomous management of the executions at run-time. In this vein, stream processing is a class of applications that compute data flowing in the form of streams (e.g., video feeds, images, and data analytics), where parallel computing can help accelerate the executions. On the one hand, stream processing applications are becoming more complex, dynamic, and long-running. On the other hand, it is unfeasible for humans to monitor and manually change the executions continuously. Hence, self-adaptation can reduce costs and human efforts by providing a higher-level abstraction with an autonomic/seamless management of executions. In this work, we aim at providing a literature review regarding self-adaptation applied to the parallel stream processing domain. We present a comprehensive revision using a systematic literature review method. Moreover, we propose a taxonomy to categorize and classify the existing self-adaptive approaches. Finally, applying the taxonomy made it possible to characterize the state-of-the-art, identify trends, and discuss open research challenges and future opportunities. |
| Vogel, Adriano; Mencagli, Gabriele; Griebler, Dalvan; Danelutto, Marco; Fernandes, Luiz Gustavo Towards On-the-fly Self-Adaptation of Stream Parallel Patterns Inproceedings doi In: 29th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP), pp. 889-93, IEEE, Valladolid, Spain, 2021. @inproceedings{VOGEL:PDP:21,
title = {Towards On-the-fly Self-Adaptation of Stream Parallel Patterns},
author = {Adriano Vogel and Gabriele Mencagli and Dalvan Griebler and Marco Danelutto and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/PDP52278.2021.00022},
doi = {10.1109/PDP52278.2021.00022},
year = {2021},
date = {2021-03-01},
booktitle = {29th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP)},
pages = {889-93},
publisher = {IEEE},
address = {Valladolid, Spain},
series = {PDP'21},
abstract = {Stream processing applications compute streams of data and provide insightful results in a timely manner, where parallel computing is necessary for accelerating the application executions. Considering that these applications are becoming increasingly dynamic and long-running, a potential solution is to apply dynamic runtime changes. However, it is challenging for humans to continuously monitor and manually self-optimize the executions. In this paper, we propose self-adaptiveness of the parallel patterns used, enabling flexible on-the-fly adaptations. The proposed solution is evaluated with an existing programming framework and running experiments with a synthetic and a real-world application. The results show that the proposed solution is able to dynamically self-adapt to the most suitable parallel pattern configuration and achieve performance competitive with the best static cases. The feasibility of the proposed solution encourages future optimizations and other applicabilities.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Stream processing applications compute streams of data and provide insightful results in a timely manner, where parallel computing is necessary for accelerating the application executions. Considering that these applications are becoming increasingly dynamic and long-running, a potential solution is to apply dynamic runtime changes. However, it is challenging for humans to continuously monitor and manually self-optimize the executions. In this paper, we propose self-adaptiveness of the parallel patterns used, enabling flexible on-the-fly adaptations. The proposed solution is evaluated with an existing programming framework and running experiments with a synthetic and a real-world application. The results show that the proposed solution is able to dynamically self-adapt to the most suitable parallel pattern configuration and achieve performance competitive with the best static cases. The feasibility of the proposed solution encourages future optimizations and other applicabilities. |
| Garcia, Adriano Marques; Griebler, Dalvan; Schepke, Claudio; Fernandes, Luiz Gustavo Introducing a Stream Processing Framework for Assessing Parallel Programming Interfaces Inproceedings doi In: 29th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP), pp. 84-88, IEEE, Valladolid, Spain, 2021. @inproceedings{GARCIA:PDP:21,
title = {Introducing a Stream Processing Framework for Assessing Parallel Programming Interfaces},
author = {Adriano Marques Garcia and Dalvan Griebler and Claudio Schepke and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/PDP52278.2021.00021},
doi = {10.1109/PDP52278.2021.00021},
year = {2021},
date = {2021-03-01},
booktitle = {29th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP)},
pages = {84-88},
publisher = {IEEE},
address = {Valladolid, Spain},
series = {PDP'21},
abstract = {Stream Processing applications are spread across different sectors of industry and people's daily lives. The increasing data we produce, such as audio, video, image, and text are demanding quickly and efficiently computation. It can be done through Stream Parallelism, which is still a challenging task and most reserved for experts. We introduce a Stream Processing framework for assessing Parallel Programming Interfaces (PPIs). Our framework targets multi-core architectures and C++ stream processing applications, providing an API that abstracts the details of the stream operators of these applications. Therefore, users can easily identify all the basic operators and implement parallelism through different PPIs. In this paper, we present the proposed framework, implement three applications using its API, and show how it works, by using it to parallelize and evaluate the applications with the PPIs Intel TBB, FastFlow, and SPar. The performance results were consistent with the literature.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Stream Processing applications are spread across different sectors of industry and people's daily lives. The increasing data we produce, such as audio, video, image, and text are demanding quickly and efficiently computation. It can be done through Stream Parallelism, which is still a challenging task and most reserved for experts. We introduce a Stream Processing framework for assessing Parallel Programming Interfaces (PPIs). Our framework targets multi-core architectures and C++ stream processing applications, providing an API that abstracts the details of the stream operators of these applications. Therefore, users can easily identify all the basic operators and implement parallelism through different PPIs. In this paper, we present the proposed framework, implement three applications using its API, and show how it works, by using it to parallelize and evaluate the applications with the PPIs Intel TBB, FastFlow, and SPar. The performance results were consistent with the literature. |
 | Vogel, Adriano; Griebler, Dalvan; Fernandes, Luiz G. Providing High‐Level Self‐Adaptive Abstractions for Stream Parallelism on Multicores Journal Article doi In: Software: Practice and Experience, vol. 51, no. 6, pp. 1194-1217, 2021. @article{VOGEL:SPE:21,
title = {Providing High‐Level Self‐Adaptive Abstractions for Stream Parallelism on Multicores},
author = {Adriano Vogel and Dalvan Griebler and Luiz G. Fernandes},
url = {https://doi.org/10.1002/spe.2948},
doi = {10.1002/spe.2948},
year = {2021},
date = {2021-01-01},
journal = {Software: Practice and Experience},
volume = {51},
number = {6},
pages = {1194-1217},
publisher = {Wiley},
abstract = {Stream processing applications are common computing workloads that demand parallelism to increase their performance. As in the past, parallel programming remains a difficult task for application programmers. The complexity increases when application programmers must set non-intuitive parallelism parameters, i.e. the degree of parallelism. The main problem is that state-of-the-art libraries use a static degree of parallelism and are not sufficiently abstracted for developing stream processing applications. In this paper, we propose a self-adaptive regulation of the degree of parallelism to provide higher-level abstractions. Flexibility is provided to programmers with two new self-adaptive strategies, one is for performance experts, and the other abstracts the need to set a performance goal. We evaluated our solution using compiler transformation rules to generate parallel code with the SPar domain-specific language. The experimental results with real-world applications highlighted higher abstraction levels without significant performance degradation in comparison to static executions. The strategy for performance experts achieved slightly higher performance than the one that works without user-defined performance goals.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Stream processing applications are common computing workloads that demand parallelism to increase their performance. As in the past, parallel programming remains a difficult task for application programmers. The complexity increases when application programmers must set non-intuitive parallelism parameters, i.e. the degree of parallelism. The main problem is that state-of-the-art libraries use a static degree of parallelism and are not sufficiently abstracted for developing stream processing applications. In this paper, we propose a self-adaptive regulation of the degree of parallelism to provide higher-level abstractions. Flexibility is provided to programmers with two new self-adaptive strategies, one is for performance experts, and the other abstracts the need to set a performance goal. We evaluated our solution using compiler transformation rules to generate parallel code with the SPar domain-specific language. The experimental results with real-world applications highlighted higher abstraction levels without significant performance degradation in comparison to static executions. The strategy for performance experts achieved slightly higher performance than the one that works without user-defined performance goals. |
| Scheer, Claudio; Hoffmann, Renato; Griebler, Dalvan; Manssour, Isabel; Fernandes, Luiz Performance Data Visualization of Linux Events on Multicores Inproceedings doi In: Anais do XXII Simpósio em Sistemas Computacionais de Alto Desempenho, pp. 108-119, SBC, Belo Horizonte, 2021. @inproceedings{SCHEER:WSCAD:21,
title = {Performance Data Visualization of Linux Events on Multicores},
author = {Claudio Scheer and Renato Hoffmann and Dalvan Griebler and Isabel Manssour and Luiz Fernandes},
url = {https://doi.org/10.5753/wscad.2021.18516},
doi = {10.5753/wscad.2021.18516},
year = {2021},
date = {2021-01-01},
booktitle = {Anais do XXII Simpósio em Sistemas Computacionais de Alto Desempenho},
pages = {108-119},
publisher = {SBC},
address = {Belo Horizonte},
abstract = {Profiling tools are essential to understand the behavior of parallel applications and assist in the optimization process. However, tools such as Perf generate a large amount of data. This way, they require significant storage space, which also complicates reasoning about this large volume of data. Therefore, we propose VisPerf: a tool-chain and an interactive visualization dashboard for Perf data. The VisPerf tool-chain profiles the application and pre-processes the data, reducing the storage space required by about 50 times. Moreover, we used the visualization dashboard to quickly understand the performance of different events and visualize specific threads and functions of a real-world application.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Profiling tools are essential to understand the behavior of parallel applications and assist in the optimization process. However, tools such as Perf generate a large amount of data. This way, they require significant storage space, which also complicates reasoning about this large volume of data. Therefore, we propose VisPerf: a tool-chain and an interactive visualization dashboard for Perf data. The VisPerf tool-chain profiles the application and pre-processes the data, reducing the storage space required by about 50 times. Moreover, we used the visualization dashboard to quickly understand the performance of different events and visualize specific threads and functions of a real-world application. |
2020
|
 | Bordin, Maycon Viana; Griebler, Dalvan; Mencagli, Gabriele; Geyer, Claudio F. R.; Fernandes, Luiz Gustavo DSPBench: a Suite of Benchmark Applications for Distributed Data Stream Processing Systems Journal Article doi In: IEEE Access, vol. 8, no. na, pp. 222900-222917, 2020. @article{BORDIN:IEEEAccess:20,
title = {DSPBench: a Suite of Benchmark Applications for Distributed Data Stream Processing Systems},
author = {Maycon Viana Bordin and Dalvan Griebler and Gabriele Mencagli and Claudio F. R. Geyer and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/ACCESS.2020.3043948},
doi = {10.1109/ACCESS.2020.3043948},
year = {2020},
date = {2020-12-01},
urldate = {2020-12-01},
journal = {IEEE Access},
volume = {8},
number = {na},
pages = {222900-222917},
publisher = {IEEE},
abstract = {Systems enabling the continuous processing of large data streams have recently attracted the attention of the scientific community and industrial stakeholders. Data Stream Processing Systems (DSPSs) are complex and powerful frameworks able to ease the development of streaming applications in distributed computing environments like clusters and clouds. Several systems of this kind have been released and currently maintained as open source projects, like Apache Storm and Spark Streaming. Some benchmark applications have often been used by the scientific community to test and evaluate new techniques to improve the performance and usability of DSPSs. However, the existing benchmark suites lack of representative workloads coming from the wide set of application domains that can leverage the benefits offered by the stream processing paradigm in terms of near real-time performance. The goal of this paper is to present a new benchmark suite composed of 15 applications coming from areas like Finance, Telecommunications, Sensor Networks, Social Networks and others. This paper describes in detail the nature of these applications, their full workload characterization in terms of selectivity, processing cost, input size and overall memory occupation. In addition, it exemplifies the usefulness of our benchmark suite to compare real DSPSs by selecting Apache Storm and Spark Streaming for this analysis.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Systems enabling the continuous processing of large data streams have recently attracted the attention of the scientific community and industrial stakeholders. Data Stream Processing Systems (DSPSs) are complex and powerful frameworks able to ease the development of streaming applications in distributed computing environments like clusters and clouds. Several systems of this kind have been released and currently maintained as open source projects, like Apache Storm and Spark Streaming. Some benchmark applications have often been used by the scientific community to test and evaluate new techniques to improve the performance and usability of DSPSs. However, the existing benchmark suites lack of representative workloads coming from the wide set of application domains that can leverage the benefits offered by the stream processing paradigm in terms of near real-time performance. The goal of this paper is to present a new benchmark suite composed of 15 applications coming from areas like Finance, Telecommunications, Sensor Networks, Social Networks and others. This paper describes in detail the nature of these applications, their full workload characterization in terms of selectivity, processing cost, input size and overall memory occupation. In addition, it exemplifies the usefulness of our benchmark suite to compare real DSPSs by selecting Apache Storm and Spark Streaming for this analysis. |
| Hoffmann, Renato B.; Griebler, Dalvan; Danelutto, Marco; Fernandes, Luiz G. Stream Parallelism Annotations for Multi-Core Frameworks Inproceedings doi In: XXIV Brazilian Symposium on Programming Languages (SBLP), pp. 48-55, ACM, Natal, Brazil, 2020. @inproceedings{HOFFMANN:SBLP:20,
title = {Stream Parallelism Annotations for Multi-Core Frameworks},
author = {Renato B. Hoffmann and Dalvan Griebler and Marco Danelutto and Luiz G. Fernandes},
url = {https://doi.org/10.1145/3427081.3427088},
doi = {10.1145/3427081.3427088},
year = {2020},
date = {2020-10-01},
booktitle = {XXIV Brazilian Symposium on Programming Languages (SBLP)},
pages = {48-55},
publisher = {ACM},
address = {Natal, Brazil},
series = {SBLP'20},
abstract = {Data generation, collection, and processing is an important workload of modern computer architectures. Stream or high-intensity data flow applications are commonly employed in extracting and interpreting the information contained in this data. Due to the computational complexity of these applications, high-performance ought to be achieved using parallel computing. Indeed, the efficient exploitation of available parallel resources from the architecture remains a challenging task for the programmers. Techniques and methodologies are required to help shift the efforts from the complexity of parallelism exploitation to specific algorithmic solutions. To tackle this problem, we propose a methodology that provides the developer with a suitable abstraction layer between a clean and effective parallel programming interface targeting different multi-core parallel programming frameworks. We used standard C++ code annotations that may be inserted in the source code by the programmer. Then, a compiler parses C++ code with the annotations and generates calls to the desired parallel runtime API. Our experiments demonstrate the feasibility of our methodology and the performance of the abstraction layer, where the difference is negligible in four applications with respect to the state-of-the-art C++ parallel programming frameworks. Additionally, our methodology allows improving the application performance since the developers can choose the runtime that best performs in their system.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Data generation, collection, and processing is an important workload of modern computer architectures. Stream or high-intensity data flow applications are commonly employed in extracting and interpreting the information contained in this data. Due to the computational complexity of these applications, high-performance ought to be achieved using parallel computing. Indeed, the efficient exploitation of available parallel resources from the architecture remains a challenging task for the programmers. Techniques and methodologies are required to help shift the efforts from the complexity of parallelism exploitation to specific algorithmic solutions. To tackle this problem, we propose a methodology that provides the developer with a suitable abstraction layer between a clean and effective parallel programming interface targeting different multi-core parallel programming frameworks. We used standard C++ code annotations that may be inserted in the source code by the programmer. Then, a compiler parses C++ code with the annotations and generates calls to the desired parallel runtime API. Our experiments demonstrate the feasibility of our methodology and the performance of the abstraction layer, where the difference is negligible in four applications with respect to the state-of-the-art C++ parallel programming frameworks. Additionally, our methodology allows improving the application performance since the developers can choose the runtime that best performs in their system. |
| Maliszewski, Anderson M.; Roloff, Eduardo; Carreño, Emmanuell D.; Griebler, Dalvan; Gaspary, Luciano P.; Navaux, Philippe O. A. Performance and Cost-Aware in Clouds: A Network Interconnection Assessment Inproceedings doi In: IEEE Symposium on Computers and Communications (ISCC), pp. 1-6, IEEE, Rennes, France, 2020. @inproceedings{larcc:network_azure_cost_perf:ISCC:20,
title = {Performance and Cost-Aware in Clouds: A Network Interconnection Assessment},
author = {Anderson M. Maliszewski and Eduardo Roloff and Emmanuell D. Carreño and Dalvan Griebler and Luciano P. Gaspary and Philippe O. A. Navaux},
url = {https://doi.org/10.1109/ISCC50000.2020.9219554},
doi = {10.1109/ISCC50000.2020.9219554},
year = {2020},
date = {2020-07-01},
booktitle = {IEEE Symposium on Computers and Communications (ISCC)},
pages = {1-6},
publisher = {IEEE},
address = {Rennes, France},
series = {ISCC'20},
abstract = {The availability of computing resources has significantly changed due to the growing adoption of the cloud computing paradigm. Aiming at potential advantages such as cost savings through the pay-per-use method and resource allocation in a scalable/elastic way, we witnessed consistent efforts to execute high-performance computing (HPC) applications in the cloud. Performance in this environment depends heavily upon two main system components: processing power and network interconnection. If, on the one hand, allocating more powerful hardware theoretically boosts performance, on the other hand, it increases the allocation cost. In this paper, we evaluated how the network interconnection impacts on performance and cost efficiency. Our experiments were carried out using NAS Parallel Benchmarks and Alya HPC application on Microsoft Azure public cloud provider, with three different cloud instances/network interconnections. The results revealed that through the use of the accelerated networking approach, which allows the instance to have a high-performance interconnect without additional charges, the performance of HPC applications can be significantly improved with a better cost efficiency.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The availability of computing resources has significantly changed due to the growing adoption of the cloud computing paradigm. Aiming at potential advantages such as cost savings through the pay-per-use method and resource allocation in a scalable/elastic way, we witnessed consistent efforts to execute high-performance computing (HPC) applications in the cloud. Performance in this environment depends heavily upon two main system components: processing power and network interconnection. If, on the one hand, allocating more powerful hardware theoretically boosts performance, on the other hand, it increases the allocation cost. In this paper, we evaluated how the network interconnection impacts on performance and cost efficiency. Our experiments were carried out using NAS Parallel Benchmarks and Alya HPC application on Microsoft Azure public cloud provider, with three different cloud instances/network interconnections. The results revealed that through the use of the accelerated networking approach, which allows the instance to have a high-performance interconnect without additional charges, the performance of HPC applications can be significantly improved with a better cost efficiency. |
| Maliszewski, Anderson M.; Roloff, Eduardo; Griebler, Dalvan; Gaspary, Luciano P.; Navaux, Philippe O. A. Performance Impact of IEEE 802.3ad in Container-based Clouds for HPC Applications Inproceedings doi In: International Conference on Computational Science and its Applications (ICCSA), pp. 158-167, Springer, Cagliari, Italy, 2020. @inproceedings{larcc:ieee802.3ad_containers:ICCSA:20,
title = {Performance Impact of IEEE 802.3ad in Container-based Clouds for HPC Applications},
author = {Anderson M. Maliszewski and Eduardo Roloff and Dalvan Griebler and Luciano P. Gaspary and Philippe O. A. Navaux},
url = {https://doi.org/10.1007/978-3-030-58817-5_13},
doi = {10.1007/978-3-030-58817-5_13},
year = {2020},
date = {2020-07-01},
booktitle = {International Conference on Computational Science and its Applications (ICCSA)},
pages = {158-167},
publisher = {Springer},
address = {Cagliari, Italy},
series = {ICCSA'20},
abstract = {Historically, large computational clusters have supported hardware requirements for executing High-Performance Computing (HPC) applications. This model has become out of date due to the high costs of maintaining and updating these infrastructures. Currently, computing resources are delivered as a service because of the cloud computing paradigm. In this way, we witnessed consistent efforts to migrate HPC applications to the cloud. However, if on the one hand cloud computing offers an attractive environment for HPC, benefiting from the pay-per-use model and on-demand resource allocation, on the other, there are still significant performance challenges to be addressed, such as the known network bottleneck. In this article, we evaluate the use of a Network Interface Cards (NIC) aggregation approach, using the IEEE 802.3ad standard to improve the performance of representative HPC applications executed in LXD container based-cloud. We assessed the aggregation impact using two and four NICs with three distinct transmission hash policies. Our results demonstrated that if the correct hash policy is selected, the NIC aggregation can significantly improve the performance of network-intensive HPC applications by up to 40%.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Historically, large computational clusters have supported hardware requirements for executing High-Performance Computing (HPC) applications. This model has become out of date due to the high costs of maintaining and updating these infrastructures. Currently, computing resources are delivered as a service because of the cloud computing paradigm. In this way, we witnessed consistent efforts to migrate HPC applications to the cloud. However, if on the one hand cloud computing offers an attractive environment for HPC, benefiting from the pay-per-use model and on-demand resource allocation, on the other, there are still significant performance challenges to be addressed, such as the known network bottleneck. In this article, we evaluate the use of a Network Interface Cards (NIC) aggregation approach, using the IEEE 802.3ad standard to improve the performance of representative HPC applications executed in LXD container based-cloud. We assessed the aggregation impact using two and four NICs with three distinct transmission hash policies. Our results demonstrated that if the correct hash policy is selected, the NIC aggregation can significantly improve the performance of network-intensive HPC applications by up to 40%. |
| Garcia, Adriano M.; Serpa, Matheus; Griebler, Dalvan; Schepke, Claudio; Fernandes, Luiz G. L.; Navaux, Philippe O. A. The Impact of CPU Frequency Scaling on Power Consumption of Computing Infrastructures Inproceedings doi In: International Conference on Computational Science and its Applications (ICCSA), pp. 142-157, Springer, Cagliari, Italy, 2020. @inproceedings{GARCIA:ICCSA:20,
title = {The Impact of CPU Frequency Scaling on Power Consumption of Computing Infrastructures},
author = {Adriano M. Garcia and Matheus Serpa and Dalvan Griebler and Claudio Schepke and Luiz G. L. Fernandes and Philippe O. A. Navaux},
url = {https://doi.org/10.1007/978-3-030-58817-5_12},
doi = {10.1007/978-3-030-58817-5_12},
year = {2020},
date = {2020-07-01},
booktitle = {International Conference on Computational Science and its Applications (ICCSA)},
volume = {12254},
pages = {142-157},
publisher = {Springer},
address = {Cagliari, Italy},
series = {ICCSA'20},
abstract = {Since the demand for computing power increases, new architectures emerged to obtain better performance. Reducing the power and energy consumption of these architectures is one of the main challenges to achieving high-performance computing. Current research trends aim at developing new software and hardware techniques to achieve the best performance and energy trade-offs. In this work, we investigate the impact of different CPU frequency scaling techniques such as ondemand, performance, and powersave on the power and energy consumption of multi-core based computer infrastructure. We apply these techniques in PAMPAR, a parallel benchmark suite implemented in PThreads, OpenMP, MPI-1, and MPI-2 (spawn). We measure the energy and execution time of 10 benchmarks, varying the number of threads. Our results show that although powersave consumes up to 43.1% less power than performance and ondemand governors, it consumes the triple of energy due to the high execution time. Our experiments also show that the performance governor consumes up to 9.8% more energy than ondemand for CPU-bound benchmarks. Finally, our results show that PThreads has the lowest power consumption, consuming less than the sequential version for memory-bound benchmarks. Regarding performance, the performance governor achieved 3% of performance over the ondemand.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Since the demand for computing power increases, new architectures emerged to obtain better performance. Reducing the power and energy consumption of these architectures is one of the main challenges to achieving high-performance computing. Current research trends aim at developing new software and hardware techniques to achieve the best performance and energy trade-offs. In this work, we investigate the impact of different CPU frequency scaling techniques such as ondemand, performance, and powersave on the power and energy consumption of multi-core based computer infrastructure. We apply these techniques in PAMPAR, a parallel benchmark suite implemented in PThreads, OpenMP, MPI-1, and MPI-2 (spawn). We measure the energy and execution time of 10 benchmarks, varying the number of threads. Our results show that although powersave consumes up to 43.1% less power than performance and ondemand governors, it consumes the triple of energy due to the high execution time. Our experiments also show that the performance governor consumes up to 9.8% more energy than ondemand for CPU-bound benchmarks. Finally, our results show that PThreads has the lowest power consumption, consuming less than the sequential version for memory-bound benchmarks. Regarding performance, the performance governor achieved 3% of performance over the ondemand. |
 | Stein, Charles M.; Rockenbach, Dinei A.; Griebler, Dalvan; Torquati, Massimo; Mencagli, Gabriele; Danelutto, Marco; Fernandes, Luiz G. Latency‐aware adaptive micro‐batching techniques for streamed data compression on graphics processing units Journal Article doi In: Concurrency and Computation: Practice and Experience, vol. na, no. na, pp. e5786, 2020. @article{STEIN:CCPE:20,
title = {Latency‐aware adaptive micro‐batching techniques for streamed data compression on graphics processing units},
author = {Charles M. Stein and Dinei A. Rockenbach and Dalvan Griebler and Massimo Torquati and Gabriele Mencagli and Marco Danelutto and Luiz G. Fernandes},
url = {https://doi.org/10.1002/cpe.5786},
doi = {10.1002/cpe.5786},
year = {2020},
date = {2020-05-01},
journal = {Concurrency and Computation: Practice and Experience},
volume = {na},
number = {na},
pages = {e5786},
publisher = {Wiley Online Library},
abstract = {Stream processing is a parallel paradigm used in many application domains. With the advance of graphics processing units (GPUs), their usage in stream processing applications has increased as well. The efficient utilization of GPU accelerators in streaming scenarios requires to batch input elements in microbatches, whose computation is offloaded on the GPU leveraging data parallelism within the same batch of data. Since data elements are continuously received based on the input speed, the bigger the microbatch size the higher the latency to completely buffer it and to start the processing on the device. Unfortunately, stream processing applications often have strict latency requirements that need to find the best size of the microbatches and to adapt it dynamically based on the workload conditions as well as according to the characteristics of the underlying device and network. In this work, we aim at implementing latency‐aware adaptive microbatching techniques and algorithms for streaming compression applications targeting GPUs. The evaluation is conducted using the Lempel‐Ziv‐Storer‐Szymanski compression application considering different input workloads. As a general result of our work, we noticed that algorithms with elastic adaptation factors respond better for stable workloads, while algorithms with narrower targets respond better for highly unbalanced workloads.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Stream processing is a parallel paradigm used in many application domains. With the advance of graphics processing units (GPUs), their usage in stream processing applications has increased as well. The efficient utilization of GPU accelerators in streaming scenarios requires to batch input elements in microbatches, whose computation is offloaded on the GPU leveraging data parallelism within the same batch of data. Since data elements are continuously received based on the input speed, the bigger the microbatch size the higher the latency to completely buffer it and to start the processing on the device. Unfortunately, stream processing applications often have strict latency requirements that need to find the best size of the microbatches and to adapt it dynamically based on the workload conditions as well as according to the characteristics of the underlying device and network. In this work, we aim at implementing latency‐aware adaptive microbatching techniques and algorithms for streaming compression applications targeting GPUs. The evaluation is conducted using the Lempel‐Ziv‐Storer‐Szymanski compression application considering different input workloads. As a general result of our work, we noticed that algorithms with elastic adaptation factors respond better for stable workloads, while algorithms with narrower targets respond better for highly unbalanced workloads. |
| Araujo, Gabriell Alves; Griebler, Dalvan; Danelutto, Marco; Fernandes, Luiz Gustavo Efficient NAS Parallel Benchmark Kernels with CUDA Inproceedings doi In: 28th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP), pp. 9-16, IEEE, Västerås, Sweden, Sweden, 2020. @inproceedings{ARAUJO:PDP:20,
title = {Efficient NAS Parallel Benchmark Kernels with CUDA},
author = {Gabriell Alves Araujo and Dalvan Griebler and Marco Danelutto and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1109/PDP50117.2020.00009},
doi = {10.1109/PDP50117.2020.00009},
year = {2020},
date = {2020-03-01},
booktitle = {28th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP)},
pages = {9-16},
publisher = {IEEE},
address = {Västerås, Sweden, Sweden},
series = {PDP'20},
abstract = {NAS Parallel Benchmarks (NPB) are one of the standard benchmark suites used to evaluate parallel hardware and software. There are many research efforts trying to provide different parallel versions apart from the original OpenMP and MPI. Concerning GPU accelerators, there are only the OpenCL and OpenACC available as consolidated versions. Our goal is to provide an efficient parallel implementation of the five NPB kernels with CUDA. Our contribution covers different aspects. First, best parallel programming practices were followed to implement NPB kernels using CUDA. Second, the support of larger workloads (class B and C) allow to stress and investigate the memory of robust GPUs. Third, we show that it is possible to make NPB efficient and suitable for GPUs although the benchmarks were designed for CPUs in the past. We succeed in achieving double performance with respect to the state-of-the-art in some cases as well as implementing efficient memory usage. Fourth, we discuss new experiments comparing performance and memory usage against OpenACC and OpenCL state-of-the-art versions using a relative new GPU architecture. The experimental results also revealed that our version is the best one for all the NPB kernels compared to OpenACC and OpenCL. The greatest differences were observed for the FT and EP kernels.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
NAS Parallel Benchmarks (NPB) are one of the standard benchmark suites used to evaluate parallel hardware and software. There are many research efforts trying to provide different parallel versions apart from the original OpenMP and MPI. Concerning GPU accelerators, there are only the OpenCL and OpenACC available as consolidated versions. Our goal is to provide an efficient parallel implementation of the five NPB kernels with CUDA. Our contribution covers different aspects. First, best parallel programming practices were followed to implement NPB kernels using CUDA. Second, the support of larger workloads (class B and C) allow to stress and investigate the memory of robust GPUs. Third, we show that it is possible to make NPB efficient and suitable for GPUs although the benchmarks were designed for CPUs in the past. We succeed in achieving double performance with respect to the state-of-the-art in some cases as well as implementing efficient memory usage. Fourth, we discuss new experiments comparing performance and memory usage against OpenACC and OpenCL state-of-the-art versions using a relative new GPU architecture. The experimental results also revealed that our version is the best one for all the NPB kernels compared to OpenACC and OpenCL. The greatest differences were observed for the FT and EP kernels. |
| Vogel, Adriano; Rista, Cassiano; Justo, Gabriel; Ewald, Endrius; Griebler, Dalvan; Mencagli, Gabriele; Fernandes, Luiz Gustavo Parallel Stream Processing with MPI for Video Analytics and Data Visualization Inproceedings doi In: High Performance Computing Systems, pp. 102-116, Springer, Cham, 2020. @inproceedings{VOGEL:CCIS:20,
title = {Parallel Stream Processing with MPI for Video Analytics and Data Visualization},
author = {Adriano Vogel and Cassiano Rista and Gabriel Justo and Endrius Ewald and Dalvan Griebler and Gabriele Mencagli and Luiz Gustavo Fernandes},
url = {https://doi.org/10.1007/978-3-030-41050-6_7},
doi = {10.1007/978-3-030-41050-6_7},
year = {2020},
date = {2020-02-01},
booktitle = {High Performance Computing Systems},
volume = {1171},
pages = {102-116},
publisher = {Springer},
address = {Cham},
series = {Communications in Computer and Information Science (CCIS)},
abstract = {The amount of data generated is increasing exponentially. However, processing data and producing fast results is a technological challenge. Parallel stream processing can be implemented for handling high frequency and big data flows. The MPI parallel programming model offers low-level and flexible mechanisms for dealing with distributed architectures such as clusters. This paper aims to use it to accelerate video analytics and data visualization applications so that insight can be obtained as soon as the data arrives. Experiments were conducted with a Domain-Specific Language for Geospatial Data Visualization and a Person Recognizer video application. We applied the same stream parallelism strategy and two task distribution strategies. The dynamic task distribution achieved better performance than the static distribution in the HPC cluster. The data visualization achieved lower throughput with respect to the video analytics due to the I/O intensive operations. Also, the MPI programming model shows promising performance outcomes for stream processing applications.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The amount of data generated is increasing exponentially. However, processing data and producing fast results is a technological challenge. Parallel stream processing can be implemented for handling high frequency and big data flows. The MPI parallel programming model offers low-level and flexible mechanisms for dealing with distributed architectures such as clusters. This paper aims to use it to accelerate video analytics and data visualization applications so that insight can be obtained as soon as the data arrives. Experiments were conducted with a Domain-Specific Language for Geospatial Data Visualization and a Person Recognizer video application. We applied the same stream parallelism strategy and two task distribution strategies. The dynamic task distribution achieved better performance than the static distribution in the HPC cluster. The data visualization achieved lower throughput with respect to the video analytics due to the I/O intensive operations. Also, the MPI programming model shows promising performance outcomes for stream processing applications. |