Memory resource allocation for file system prefetching from a supply chain management perspective

Zhe Zhang, Amit Kulkarni, Xiaosong Ma, Yuanyuan Zhou

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

19 Citations (Scopus)

Abstract

As an important technique to hide disk I/O latency, prefetching has been widely studied, and dynamic adaptive prefetching techniques have been deployed in diverse storage environments. However, two issues are not well addressed by previous research: (1) how to handle the prefetching resource allocation between concurrent sequential access streams with different request rates, and (2) how to coordinate prefetching at multiple levels in the data access path. Interestingly, we found that these problems bear a strong resemblance to situations long studied in the field of supply chain management (SCM), used by retailers such as Wal-Mart. In this paper, we demonstrate how to perform the problem mapping and then apply SCM principles in practice, particularly from the branch of inventory theory, to improve data prefetching performance in storage systems. More specifically, we applied (1) two SCM policies to dynamically configure the sequential prefetching parameters, and (2) an SCM solution to correct the access pattern information distortion in multi-level prefetching. We implemented these SCM-based strategies in the Linux kernel prefetching algorithm and a multi-level storage simulator, and evaluated the performance with three types of work- loads. The results indicate that the SCM approaches are able to generate up to a 55.0% of performance improvement for a real-world server workload benchmark, and up to 33.3% for a combination of Linux I/O-intensive applications.

Original languageEnglish
Title of host publicationProceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09
Pages75-88
Number of pages14
DOIs
Publication statusPublished - 21 Sep 2009
Externally publishedYes
Event4th ACM European Conference on Computer Systems, EuroSys'09 - Nuremberg, Germany
Duration: 1 Apr 20093 Apr 2009

Other

Other4th ACM European Conference on Computer Systems, EuroSys'09
CountryGermany
CityNuremberg
Period1/4/093/4/09

Fingerprint

Supply chain management
Resource allocation
Data storage equipment
Servers
Simulators

Keywords

  • Memory cache
  • Prefetching
  • Supply chain management

ASJC Scopus subject areas

  • Hardware and Architecture
  • Information Systems

Cite this

Zhang, Z., Kulkarni, A., Ma, X., & Zhou, Y. (2009). Memory resource allocation for file system prefetching from a supply chain management perspective. In Proceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09 (pp. 75-88) https://doi.org/10.1145/1519065.1519075

Memory resource allocation for file system prefetching from a supply chain management perspective. / Zhang, Zhe; Kulkarni, Amit; Ma, Xiaosong; Zhou, Yuanyuan.

Proceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09. 2009. p. 75-88.

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

Zhang, Z, Kulkarni, A, Ma, X & Zhou, Y 2009, Memory resource allocation for file system prefetching from a supply chain management perspective. in Proceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09. pp. 75-88, 4th ACM European Conference on Computer Systems, EuroSys'09, Nuremberg, Germany, 1/4/09. https://doi.org/10.1145/1519065.1519075
Zhang Z, Kulkarni A, Ma X, Zhou Y. Memory resource allocation for file system prefetching from a supply chain management perspective. In Proceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09. 2009. p. 75-88 https://doi.org/10.1145/1519065.1519075
Zhang, Zhe ; Kulkarni, Amit ; Ma, Xiaosong ; Zhou, Yuanyuan. / Memory resource allocation for file system prefetching from a supply chain management perspective. Proceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09. 2009. pp. 75-88
@inproceedings{fb406fddf0a54bf6a6b78bc6eac71a20,
title = "Memory resource allocation for file system prefetching from a supply chain management perspective",
abstract = "As an important technique to hide disk I/O latency, prefetching has been widely studied, and dynamic adaptive prefetching techniques have been deployed in diverse storage environments. However, two issues are not well addressed by previous research: (1) how to handle the prefetching resource allocation between concurrent sequential access streams with different request rates, and (2) how to coordinate prefetching at multiple levels in the data access path. Interestingly, we found that these problems bear a strong resemblance to situations long studied in the field of supply chain management (SCM), used by retailers such as Wal-Mart. In this paper, we demonstrate how to perform the problem mapping and then apply SCM principles in practice, particularly from the branch of inventory theory, to improve data prefetching performance in storage systems. More specifically, we applied (1) two SCM policies to dynamically configure the sequential prefetching parameters, and (2) an SCM solution to correct the access pattern information distortion in multi-level prefetching. We implemented these SCM-based strategies in the Linux kernel prefetching algorithm and a multi-level storage simulator, and evaluated the performance with three types of work- loads. The results indicate that the SCM approaches are able to generate up to a 55.0{\%} of performance improvement for a real-world server workload benchmark, and up to 33.3{\%} for a combination of Linux I/O-intensive applications.",
keywords = "Memory cache, Prefetching, Supply chain management",
author = "Zhe Zhang and Amit Kulkarni and Xiaosong Ma and Yuanyuan Zhou",
year = "2009",
month = "9",
day = "21",
doi = "10.1145/1519065.1519075",
language = "English",
isbn = "9781605584829",
pages = "75--88",
booktitle = "Proceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09",

}

TY - GEN

T1 - Memory resource allocation for file system prefetching from a supply chain management perspective

AU - Zhang, Zhe

AU - Kulkarni, Amit

AU - Ma, Xiaosong

AU - Zhou, Yuanyuan

PY - 2009/9/21

Y1 - 2009/9/21

N2 - As an important technique to hide disk I/O latency, prefetching has been widely studied, and dynamic adaptive prefetching techniques have been deployed in diverse storage environments. However, two issues are not well addressed by previous research: (1) how to handle the prefetching resource allocation between concurrent sequential access streams with different request rates, and (2) how to coordinate prefetching at multiple levels in the data access path. Interestingly, we found that these problems bear a strong resemblance to situations long studied in the field of supply chain management (SCM), used by retailers such as Wal-Mart. In this paper, we demonstrate how to perform the problem mapping and then apply SCM principles in practice, particularly from the branch of inventory theory, to improve data prefetching performance in storage systems. More specifically, we applied (1) two SCM policies to dynamically configure the sequential prefetching parameters, and (2) an SCM solution to correct the access pattern information distortion in multi-level prefetching. We implemented these SCM-based strategies in the Linux kernel prefetching algorithm and a multi-level storage simulator, and evaluated the performance with three types of work- loads. The results indicate that the SCM approaches are able to generate up to a 55.0% of performance improvement for a real-world server workload benchmark, and up to 33.3% for a combination of Linux I/O-intensive applications.

AB - As an important technique to hide disk I/O latency, prefetching has been widely studied, and dynamic adaptive prefetching techniques have been deployed in diverse storage environments. However, two issues are not well addressed by previous research: (1) how to handle the prefetching resource allocation between concurrent sequential access streams with different request rates, and (2) how to coordinate prefetching at multiple levels in the data access path. Interestingly, we found that these problems bear a strong resemblance to situations long studied in the field of supply chain management (SCM), used by retailers such as Wal-Mart. In this paper, we demonstrate how to perform the problem mapping and then apply SCM principles in practice, particularly from the branch of inventory theory, to improve data prefetching performance in storage systems. More specifically, we applied (1) two SCM policies to dynamically configure the sequential prefetching parameters, and (2) an SCM solution to correct the access pattern information distortion in multi-level prefetching. We implemented these SCM-based strategies in the Linux kernel prefetching algorithm and a multi-level storage simulator, and evaluated the performance with three types of work- loads. The results indicate that the SCM approaches are able to generate up to a 55.0% of performance improvement for a real-world server workload benchmark, and up to 33.3% for a combination of Linux I/O-intensive applications.

KW - Memory cache

KW - Prefetching

KW - Supply chain management

UR - http://www.scopus.com/inward/record.url?scp=70349085675&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70349085675&partnerID=8YFLogxK

U2 - 10.1145/1519065.1519075

DO - 10.1145/1519065.1519075

M3 - Conference contribution

AN - SCOPUS:70349085675

SN - 9781605584829

SP - 75

EP - 88

BT - Proceedings of the 4th ACM European Conference on Computer Systems, EuroSys'09

ER -