NUMA memory arrangement by using BW-SA

Jian Zhang, Mei Wang, Huanmei Wu

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

Abstract

RAMCloud is a storage system that provides low-latency to access large-scale datasets. This paper focuses on the RAMcloud storage technology under NUMA (Non-Uniform Memory Access / Non-Uniform Memory Architecture) architecture - memories' arrangement. For the better performance of blocked memories and minimum cost, we present a prototype model and use blocked windows-simulated annealing (BW-SA) algorithm to sort-merge the memories and arrange them efficiently. The results converge to a best solution to prove the algorithm. After the arrangement we can assign the global address for the memories.

Original languageEnglish (US)
Title of host publication29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016
PublisherThe International Society for Computers and Their Applications (ISCA)
Pages239-244
Number of pages6
ISBN (Electronic)9781943436040
StatePublished - 2016
Event29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016 - Denver, United States
Duration: Sep 26 2016Sep 28 2016

Other

Other29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016
CountryUnited States
CityDenver
Period9/26/169/28/16

Fingerprint

Simulated annealing
Data storage equipment
Memory architecture
Costs

Keywords

  • Blocked windows
  • NUMA
  • RAMCloud
  • Simulated annealing

ASJC Scopus subject areas

  • Computer Science Applications
  • Industrial and Manufacturing Engineering

Cite this

Zhang, J., Wang, M., & Wu, H. (2016). NUMA memory arrangement by using BW-SA. In 29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016 (pp. 239-244). The International Society for Computers and Their Applications (ISCA).

NUMA memory arrangement by using BW-SA. / Zhang, Jian; Wang, Mei; Wu, Huanmei.

29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016. The International Society for Computers and Their Applications (ISCA), 2016. p. 239-244.

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

Zhang, J, Wang, M & Wu, H 2016, NUMA memory arrangement by using BW-SA. in 29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016. The International Society for Computers and Their Applications (ISCA), pp. 239-244, 29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016, Denver, United States, 9/26/16.
Zhang J, Wang M, Wu H. NUMA memory arrangement by using BW-SA. In 29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016. The International Society for Computers and Their Applications (ISCA). 2016. p. 239-244
Zhang, Jian ; Wang, Mei ; Wu, Huanmei. / NUMA memory arrangement by using BW-SA. 29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016. The International Society for Computers and Their Applications (ISCA), 2016. pp. 239-244
@inproceedings{da64e2b3ea9a4190ad69b489dd8f6670,
title = "NUMA memory arrangement by using BW-SA",
abstract = "RAMCloud is a storage system that provides low-latency to access large-scale datasets. This paper focuses on the RAMcloud storage technology under NUMA (Non-Uniform Memory Access / Non-Uniform Memory Architecture) architecture - memories' arrangement. For the better performance of blocked memories and minimum cost, we present a prototype model and use blocked windows-simulated annealing (BW-SA) algorithm to sort-merge the memories and arrange them efficiently. The results converge to a best solution to prove the algorithm. After the arrangement we can assign the global address for the memories.",
keywords = "Blocked windows, NUMA, RAMCloud, Simulated annealing",
author = "Jian Zhang and Mei Wang and Huanmei Wu",
year = "2016",
language = "English (US)",
pages = "239--244",
booktitle = "29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016",
publisher = "The International Society for Computers and Their Applications (ISCA)",

}

TY - GEN

T1 - NUMA memory arrangement by using BW-SA

AU - Zhang, Jian

AU - Wang, Mei

AU - Wu, Huanmei

PY - 2016

Y1 - 2016

N2 - RAMCloud is a storage system that provides low-latency to access large-scale datasets. This paper focuses on the RAMcloud storage technology under NUMA (Non-Uniform Memory Access / Non-Uniform Memory Architecture) architecture - memories' arrangement. For the better performance of blocked memories and minimum cost, we present a prototype model and use blocked windows-simulated annealing (BW-SA) algorithm to sort-merge the memories and arrange them efficiently. The results converge to a best solution to prove the algorithm. After the arrangement we can assign the global address for the memories.

AB - RAMCloud is a storage system that provides low-latency to access large-scale datasets. This paper focuses on the RAMcloud storage technology under NUMA (Non-Uniform Memory Access / Non-Uniform Memory Architecture) architecture - memories' arrangement. For the better performance of blocked memories and minimum cost, we present a prototype model and use blocked windows-simulated annealing (BW-SA) algorithm to sort-merge the memories and arrange them efficiently. The results converge to a best solution to prove the algorithm. After the arrangement we can assign the global address for the memories.

KW - Blocked windows

KW - NUMA

KW - RAMCloud

KW - Simulated annealing

UR - http://www.scopus.com/inward/record.url?scp=84991508746&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84991508746&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84991508746

SP - 239

EP - 244

BT - 29th International Conference on Computer Applications in Industry and Engineering, CAINE 2016

PB - The International Society for Computers and Their Applications (ISCA)

ER -