Stream Processing on Hybrid CPU/Intel® Xeon Phi™ Systems

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Stream processing is currently central to handle large volumes of data generated at high rates. However, the efficient processing of such quantity of data demands massively parallel hardware. The usual approach is to rely on clusters of multi-processors, where network communication may become a bottleneck. Some work has also been done in the GPU computing field. Yet, the GPUs’ programming complexity and the existence of synchronization-related overheads, when the streaming graph scales, have hampered the integration of GPUs in the Big Data streaming frameworks. In this paper we explore the unique characteristics of the Intel Xeon Phi processor to develop a stream processing framework for hybrid CPU/Intel Xeon Phi systems. We built atop the Intel Threading Building Blocks library and the Marrow algorithmic skeleton framework to offer an easily programmable high performance system. Our experimental results show that offloading the computationally heavy nodes of a streaming graph to the Xeon Phi may earn considerable speed-ups. Furthermore, additional gains may be obtained by sharing the processing load between the CPU(s) and the Xeon Phi processor(s).

Original languageEnglish
Title of host publicationEuro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings
EditorsM. Torquati, M. Aldinucci, L. Padovani
PublisherSpringer Verlag
Pages796-810
Number of pages15
ISBN (Print)9783319969824
DOIs
Publication statusPublished - 2018
Event24th International European Conference on Parallel and Distributed Computing, Euro-Par 2018 - Turin, Italy
Duration: 27 Aug 201831 Aug 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Volume11014 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference24th International European Conference on Parallel and Distributed Computing, Euro-Par 2018
CountryItaly
CityTurin
Period27/08/1831/08/18

Fingerprint

Stream Processing
Program processors
Streaming
Processing
Streaming Data
Network Communication
Graph in graph theory
Multiprocessor
Skeleton
Building Blocks
Sharing
Synchronization
Programming
High Performance
Hardware
Telecommunication networks
Computing
Experimental Results
Vertex of a graph
Framework

Keywords

  • Algorithmic skeletons
  • Intel Xeon Phi
  • Parallel computing
  • Stream processing

Cite this

Ferrão, P., Marques, H., & Paulino, H. (2018). Stream Processing on Hybrid CPU/Intel® Xeon Phi™ Systems. In M. Torquati, M. Aldinucci, & L. Padovani (Eds.), Euro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings (pp. 796-810). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11014 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-96983-1_56
Ferrão, Paulo ; Marques, Hélder ; Paulino, Hervé. / Stream Processing on Hybrid CPU/Intel® Xeon Phi™ Systems. Euro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings. editor / M. Torquati ; M. Aldinucci ; L. Padovani. Springer Verlag, 2018. pp. 796-810 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{6f28a7dc9f054ae6841a025e2271e86f,
title = "Stream Processing on Hybrid CPU/Intel{\circledR} Xeon Phi™ Systems",
abstract = "Stream processing is currently central to handle large volumes of data generated at high rates. However, the efficient processing of such quantity of data demands massively parallel hardware. The usual approach is to rely on clusters of multi-processors, where network communication may become a bottleneck. Some work has also been done in the GPU computing field. Yet, the GPUs’ programming complexity and the existence of synchronization-related overheads, when the streaming graph scales, have hampered the integration of GPUs in the Big Data streaming frameworks. In this paper we explore the unique characteristics of the Intel Xeon Phi processor to develop a stream processing framework for hybrid CPU/Intel Xeon Phi systems. We built atop the Intel Threading Building Blocks library and the Marrow algorithmic skeleton framework to offer an easily programmable high performance system. Our experimental results show that offloading the computationally heavy nodes of a streaming graph to the Xeon Phi may earn considerable speed-ups. Furthermore, additional gains may be obtained by sharing the processing load between the CPU(s) and the Xeon Phi processor(s).",
keywords = "Algorithmic skeletons, Intel Xeon Phi, Parallel computing, Stream processing",
author = "Paulo Ferr{\~a}o and H{\'e}lder Marques and Herv{\'e} Paulino",
note = "info:eu-repo/grantAgreement/FCT/5876/147279/PT#",
year = "2018",
doi = "10.1007/978-3-319-96983-1_56",
language = "English",
isbn = "9783319969824",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "796--810",
editor = "M. Torquati and M. Aldinucci and L. Padovani",
booktitle = "Euro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings",

}

Ferrão, P, Marques, H & Paulino, H 2018, Stream Processing on Hybrid CPU/Intel® Xeon Phi™ Systems. in M Torquati, M Aldinucci & L Padovani (eds), Euro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11014 LNCS, Springer Verlag, pp. 796-810, 24th International European Conference on Parallel and Distributed Computing, Euro-Par 2018, Turin, Italy, 27/08/18. https://doi.org/10.1007/978-3-319-96983-1_56

Stream Processing on Hybrid CPU/Intel® Xeon Phi™ Systems. / Ferrão, Paulo; Marques, Hélder; Paulino, Hervé.

Euro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings. ed. / M. Torquati; M. Aldinucci; L. Padovani. Springer Verlag, 2018. p. 796-810 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11014 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Stream Processing on Hybrid CPU/Intel® Xeon Phi™ Systems

AU - Ferrão, Paulo

AU - Marques, Hélder

AU - Paulino, Hervé

N1 - info:eu-repo/grantAgreement/FCT/5876/147279/PT#

PY - 2018

Y1 - 2018

N2 - Stream processing is currently central to handle large volumes of data generated at high rates. However, the efficient processing of such quantity of data demands massively parallel hardware. The usual approach is to rely on clusters of multi-processors, where network communication may become a bottleneck. Some work has also been done in the GPU computing field. Yet, the GPUs’ programming complexity and the existence of synchronization-related overheads, when the streaming graph scales, have hampered the integration of GPUs in the Big Data streaming frameworks. In this paper we explore the unique characteristics of the Intel Xeon Phi processor to develop a stream processing framework for hybrid CPU/Intel Xeon Phi systems. We built atop the Intel Threading Building Blocks library and the Marrow algorithmic skeleton framework to offer an easily programmable high performance system. Our experimental results show that offloading the computationally heavy nodes of a streaming graph to the Xeon Phi may earn considerable speed-ups. Furthermore, additional gains may be obtained by sharing the processing load between the CPU(s) and the Xeon Phi processor(s).

AB - Stream processing is currently central to handle large volumes of data generated at high rates. However, the efficient processing of such quantity of data demands massively parallel hardware. The usual approach is to rely on clusters of multi-processors, where network communication may become a bottleneck. Some work has also been done in the GPU computing field. Yet, the GPUs’ programming complexity and the existence of synchronization-related overheads, when the streaming graph scales, have hampered the integration of GPUs in the Big Data streaming frameworks. In this paper we explore the unique characteristics of the Intel Xeon Phi processor to develop a stream processing framework for hybrid CPU/Intel Xeon Phi systems. We built atop the Intel Threading Building Blocks library and the Marrow algorithmic skeleton framework to offer an easily programmable high performance system. Our experimental results show that offloading the computationally heavy nodes of a streaming graph to the Xeon Phi may earn considerable speed-ups. Furthermore, additional gains may be obtained by sharing the processing load between the CPU(s) and the Xeon Phi processor(s).

KW - Algorithmic skeletons

KW - Intel Xeon Phi

KW - Parallel computing

KW - Stream processing

UR - http://www.scopus.com/inward/record.url?scp=85052930688&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-96983-1_56

DO - 10.1007/978-3-319-96983-1_56

M3 - Conference contribution

SN - 9783319969824

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 796

EP - 810

BT - Euro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings

A2 - Torquati, M.

A2 - Aldinucci, M.

A2 - Padovani, L.

PB - Springer Verlag

ER -

Ferrão P, Marques H, Paulino H. Stream Processing on Hybrid CPU/Intel® Xeon Phi™ Systems. In Torquati M, Aldinucci M, Padovani L, editors, Euro-Par 2018: Parallel Processing - 24th International Conference on Parallel and Distributed Computing, Proceedings. Springer Verlag. 2018. p. 796-810. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-96983-1_56