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

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

doi:10.1007/978-3-642-31137-6_21

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

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

Department of Computer Science and Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference 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 language	English
Title of host publication	Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings
Pages	272-282
Number of pages	11
Edition	PART 3
DOIs	https://doi.org/10.1007/978-3-642-31137-6_21
Publication status	Published - 2012
Event	12th International Conference on Computational Science and Its Applications, ICCSA 2012 - Salvador de Bahia, Brazil Duration: 2012 Jun 18 → 2012 Jun 21

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 3
Volume	7335 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	12th International Conference on Computational Science and Its Applications, ICCSA 2012
Country	Brazil
City	Salvador de Bahia
Period	12/6/18 → 12/6/21

Keywords

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

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

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

Fingerprint Dive into the research topics of 'Reducing last level cache pollution in NUMA multicore systems for improving cache performance'. Together they form a unique fingerprint.

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 Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings (PART 3 ed., 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.

Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings. 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 proceeding › Conference contribution

An, D, Kim, J, Han, J & Eom, YI 2012, Reducing last level cache pollution in NUMA multicore systems for improving cache performance. in Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings. PART 3 edn, 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 Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings. PART 3 ed. 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. Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings. 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",

doi = "10.1007/978-3-642-31137-6_21",

language = "English",

isbn = "9783642311369",

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 = "Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings",

edition = "PART 3",

note = "12th International Conference on Computational Science and Its Applications, ICCSA 2012 ; Conference date: 18-06-2012 Through 21-06-2012",

}

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

Y1 - 2012

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

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

SP - 272

EP - 282

BT - Computational Science and Its Applications - 12th International Conference, ICCSA 2012, Proceedings

T2 - 12th International Conference on Computational Science and Its Applications, ICCSA 2012

Y2 - 18 June 2012 through 21 June 2012

ER -