Reducing last level cache pollution in NUMA multicore systems for improving cache performance

Deukhyeon An, Jeehong Kim, Junghyun Han, Young Ik Eom

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

Abstract

Non-uniform memory architecture (NUMA) system has numerous nodes with shared last level cache (LLC). Their shared LLC has brought many benefits in the cache utilization. However, LLC can be seriously polluted by tasks that cause huge I/O traffic for a long time since inclusive cache architecture of LLC replaces valid cache line by back-invalidate. Many research on the page coloring, partitioning, and pollute buffer mechanism handled this cache pollution. But, there are no scheduling approaches considering I/O-intensive tasks in NUMA systems. To address the above problem, OS scheduling that reduces cache pollution is highly needed in NUMA systems. In this paper, we propose a software-based mechanism that reduces shared LLC miss in NUMA systems. Our mechanism includes I/O traffic measurement and devil conscious scheduling. The experimental results show that LLC miss rate can be reduced up to 37.6%, and our approach improves execution time to 1.48%.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages272-282
Number of pages11
Volume7335 LNCS
EditionPART 3
DOIs
Publication statusPublished - 2012 Jul 23
Event12th International Conference on Computational Science and Its Applications, ICCSA 2012 - Salvador de Bahia, Brazil
Duration: 2012 Jun 182012 Jun 21

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 3
Volume7335 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other12th International Conference on Computational Science and Its Applications, ICCSA 2012
CountryBrazil
CitySalvador de Bahia
Period12/6/1812/6/21

Fingerprint

Memory architecture
Pollution
Cache
Computer systems
Scheduling
System Architecture
Coloring
Architecture
Traffic Measurement
Execution Time
Colouring
Buffer
Partitioning
Traffic
Valid

Keywords

  • Cache Performance
  • Cache Pollution
  • I/O Intensive Task
  • Last Level Cache
  • NUMA Scheduling
  • Task Characteristics

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

An, D., Kim, J., Han, J., & Eom, Y. I. (2012). Reducing last level cache pollution in NUMA multicore systems for improving cache performance. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 3 ed., Vol. 7335 LNCS, pp. 272-282). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7335 LNCS, No. PART 3). https://doi.org/10.1007/978-3-642-31137-6_21

Reducing last level cache pollution in NUMA multicore systems for improving cache performance. / An, Deukhyeon; Kim, Jeehong; Han, Junghyun; Eom, Young Ik.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7335 LNCS PART 3. ed. 2012. p. 272-282 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7335 LNCS, No. PART 3).

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

An, D, Kim, J, Han, J & Eom, YI 2012, Reducing last level cache pollution in NUMA multicore systems for improving cache performance. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 3 edn, vol. 7335 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 3, vol. 7335 LNCS, pp. 272-282, 12th International Conference on Computational Science and Its Applications, ICCSA 2012, Salvador de Bahia, Brazil, 12/6/18. https://doi.org/10.1007/978-3-642-31137-6_21
An D, Kim J, Han J, Eom YI. Reducing last level cache pollution in NUMA multicore systems for improving cache performance. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 3 ed. Vol. 7335 LNCS. 2012. p. 272-282. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3). https://doi.org/10.1007/978-3-642-31137-6_21
An, Deukhyeon ; Kim, Jeehong ; Han, Junghyun ; Eom, Young Ik. / Reducing last level cache pollution in NUMA multicore systems for improving cache performance. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7335 LNCS PART 3. ed. 2012. pp. 272-282 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3).
@inproceedings{c1416d6cd9ef448f99dbd858458d198a,
title = "Reducing last level cache pollution in NUMA multicore systems for improving cache performance",
abstract = "Non-uniform memory architecture (NUMA) system has numerous nodes with shared last level cache (LLC). Their shared LLC has brought many benefits in the cache utilization. However, LLC can be seriously polluted by tasks that cause huge I/O traffic for a long time since inclusive cache architecture of LLC replaces valid cache line by back-invalidate. Many research on the page coloring, partitioning, and pollute buffer mechanism handled this cache pollution. But, there are no scheduling approaches considering I/O-intensive tasks in NUMA systems. To address the above problem, OS scheduling that reduces cache pollution is highly needed in NUMA systems. In this paper, we propose a software-based mechanism that reduces shared LLC miss in NUMA systems. Our mechanism includes I/O traffic measurement and devil conscious scheduling. The experimental results show that LLC miss rate can be reduced up to 37.6{\%}, and our approach improves execution time to 1.48{\%}.",
keywords = "Cache Performance, Cache Pollution, I/O Intensive Task, Last Level Cache, NUMA Scheduling, Task Characteristics",
author = "Deukhyeon An and Jeehong Kim and Junghyun Han and Eom, {Young Ik}",
year = "2012",
month = "7",
day = "23",
doi = "10.1007/978-3-642-31137-6_21",
language = "English",
isbn = "9783642311369",
volume = "7335 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 3",
pages = "272--282",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 3",

}

TY - GEN

T1 - Reducing last level cache pollution in NUMA multicore systems for improving cache performance

AU - An, Deukhyeon

AU - Kim, Jeehong

AU - Han, Junghyun

AU - Eom, Young Ik

PY - 2012/7/23

Y1 - 2012/7/23

N2 - Non-uniform memory architecture (NUMA) system has numerous nodes with shared last level cache (LLC). Their shared LLC has brought many benefits in the cache utilization. However, LLC can be seriously polluted by tasks that cause huge I/O traffic for a long time since inclusive cache architecture of LLC replaces valid cache line by back-invalidate. Many research on the page coloring, partitioning, and pollute buffer mechanism handled this cache pollution. But, there are no scheduling approaches considering I/O-intensive tasks in NUMA systems. To address the above problem, OS scheduling that reduces cache pollution is highly needed in NUMA systems. In this paper, we propose a software-based mechanism that reduces shared LLC miss in NUMA systems. Our mechanism includes I/O traffic measurement and devil conscious scheduling. The experimental results show that LLC miss rate can be reduced up to 37.6%, and our approach improves execution time to 1.48%.

AB - Non-uniform memory architecture (NUMA) system has numerous nodes with shared last level cache (LLC). Their shared LLC has brought many benefits in the cache utilization. However, LLC can be seriously polluted by tasks that cause huge I/O traffic for a long time since inclusive cache architecture of LLC replaces valid cache line by back-invalidate. Many research on the page coloring, partitioning, and pollute buffer mechanism handled this cache pollution. But, there are no scheduling approaches considering I/O-intensive tasks in NUMA systems. To address the above problem, OS scheduling that reduces cache pollution is highly needed in NUMA systems. In this paper, we propose a software-based mechanism that reduces shared LLC miss in NUMA systems. Our mechanism includes I/O traffic measurement and devil conscious scheduling. The experimental results show that LLC miss rate can be reduced up to 37.6%, and our approach improves execution time to 1.48%.

KW - Cache Performance

KW - Cache Pollution

KW - I/O Intensive Task

KW - Last Level Cache

KW - NUMA Scheduling

KW - Task Characteristics

UR - http://www.scopus.com/inward/record.url?scp=84863952156&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84863952156&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-31137-6_21

DO - 10.1007/978-3-642-31137-6_21

M3 - Conference contribution

AN - SCOPUS:84863952156

SN - 9783642311369

VL - 7335 LNCS

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

SP - 272

EP - 282

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

ER -