Storage performance virtualization via throughput and latency control

Jianyong Zhang, Anand Sivasubramaniam, Qian Wang, Alma Riska, Erik Riedel

Research output: Contribution to journalArticle

54 Citations (Scopus)

Abstract

I/O consolidation is a growing trend in production environments due to increasing complexity in tuning and managing storage systems. A consequence of this trend is the need to serve multiple users and/or workloads simultaneously. It is imperative to ensure that these users are insulated from each other by virtualization in order to meet any service-level objective (SLO). Previous proposals for performance virtualization suffer from one or more of the following drawbacks: (1) They rely on a fairly detailed performance model of the underlying storage system; (2) couple rate and latency allocation in a single scheduler, making them less flexible; or (3) may not always exploit the full bandwidth offered by the storage system. This article presents a two-level scheduling framework that can be built on top of an existing storage utility. This framework uses a low-level feedback-driven request scheduler, called AVATAR, that is intended to meet the latency bounds determined by the SLO. The load imposed on AVATAR is regulated by a high-level rate controller, called SARC, to insulate the users from each other. In addition, SARC is work-conserving and tries to fairly distribute any spare bandwidth in the storage system to the different users. This framework naturally decouples rate and latency allocation. Using extensive I/O traces and a detailed storage simulator, we demonstrate that this two-level framework can simultaneously meet the latency and throughput requirements imposed by an SLO, without requiring extensive knowledge of the underlying storage system.

Original languageEnglish (US)
Pages (from-to)283-308
Number of pages26
JournalACM Transactions on Storage
Volume2
Issue number3
DOIs
StatePublished - Aug 1 2006

Fingerprint

Throughput
Bandwidth
Consolidation
Tuning
Simulators
Scheduling
Feedback
Controllers
Virtualization

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture

Cite this

Zhang, Jianyong ; Sivasubramaniam, Anand ; Wang, Qian ; Riska, Alma ; Riedel, Erik. / Storage performance virtualization via throughput and latency control. In: ACM Transactions on Storage. 2006 ; Vol. 2, No. 3. pp. 283-308.
@article{c7e131cc19194617b2e5671025ccd640,
title = "Storage performance virtualization via throughput and latency control",
abstract = "I/O consolidation is a growing trend in production environments due to increasing complexity in tuning and managing storage systems. A consequence of this trend is the need to serve multiple users and/or workloads simultaneously. It is imperative to ensure that these users are insulated from each other by virtualization in order to meet any service-level objective (SLO). Previous proposals for performance virtualization suffer from one or more of the following drawbacks: (1) They rely on a fairly detailed performance model of the underlying storage system; (2) couple rate and latency allocation in a single scheduler, making them less flexible; or (3) may not always exploit the full bandwidth offered by the storage system. This article presents a two-level scheduling framework that can be built on top of an existing storage utility. This framework uses a low-level feedback-driven request scheduler, called AVATAR, that is intended to meet the latency bounds determined by the SLO. The load imposed on AVATAR is regulated by a high-level rate controller, called SARC, to insulate the users from each other. In addition, SARC is work-conserving and tries to fairly distribute any spare bandwidth in the storage system to the different users. This framework naturally decouples rate and latency allocation. Using extensive I/O traces and a detailed storage simulator, we demonstrate that this two-level framework can simultaneously meet the latency and throughput requirements imposed by an SLO, without requiring extensive knowledge of the underlying storage system.",
author = "Jianyong Zhang and Anand Sivasubramaniam and Qian Wang and Alma Riska and Erik Riedel",
year = "2006",
month = "8",
day = "1",
doi = "10.1145/1168910.1168913",
language = "English (US)",
volume = "2",
pages = "283--308",
journal = "ACM Transactions on Storage",
issn = "1553-3077",
publisher = "Association for Computing Machinery (ACM)",
number = "3",

}

Storage performance virtualization via throughput and latency control. / Zhang, Jianyong; Sivasubramaniam, Anand; Wang, Qian; Riska, Alma; Riedel, Erik.

In: ACM Transactions on Storage, Vol. 2, No. 3, 01.08.2006, p. 283-308.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Storage performance virtualization via throughput and latency control

AU - Zhang, Jianyong

AU - Sivasubramaniam, Anand

AU - Wang, Qian

AU - Riska, Alma

AU - Riedel, Erik

PY - 2006/8/1

Y1 - 2006/8/1

N2 - I/O consolidation is a growing trend in production environments due to increasing complexity in tuning and managing storage systems. A consequence of this trend is the need to serve multiple users and/or workloads simultaneously. It is imperative to ensure that these users are insulated from each other by virtualization in order to meet any service-level objective (SLO). Previous proposals for performance virtualization suffer from one or more of the following drawbacks: (1) They rely on a fairly detailed performance model of the underlying storage system; (2) couple rate and latency allocation in a single scheduler, making them less flexible; or (3) may not always exploit the full bandwidth offered by the storage system. This article presents a two-level scheduling framework that can be built on top of an existing storage utility. This framework uses a low-level feedback-driven request scheduler, called AVATAR, that is intended to meet the latency bounds determined by the SLO. The load imposed on AVATAR is regulated by a high-level rate controller, called SARC, to insulate the users from each other. In addition, SARC is work-conserving and tries to fairly distribute any spare bandwidth in the storage system to the different users. This framework naturally decouples rate and latency allocation. Using extensive I/O traces and a detailed storage simulator, we demonstrate that this two-level framework can simultaneously meet the latency and throughput requirements imposed by an SLO, without requiring extensive knowledge of the underlying storage system.

AB - I/O consolidation is a growing trend in production environments due to increasing complexity in tuning and managing storage systems. A consequence of this trend is the need to serve multiple users and/or workloads simultaneously. It is imperative to ensure that these users are insulated from each other by virtualization in order to meet any service-level objective (SLO). Previous proposals for performance virtualization suffer from one or more of the following drawbacks: (1) They rely on a fairly detailed performance model of the underlying storage system; (2) couple rate and latency allocation in a single scheduler, making them less flexible; or (3) may not always exploit the full bandwidth offered by the storage system. This article presents a two-level scheduling framework that can be built on top of an existing storage utility. This framework uses a low-level feedback-driven request scheduler, called AVATAR, that is intended to meet the latency bounds determined by the SLO. The load imposed on AVATAR is regulated by a high-level rate controller, called SARC, to insulate the users from each other. In addition, SARC is work-conserving and tries to fairly distribute any spare bandwidth in the storage system to the different users. This framework naturally decouples rate and latency allocation. Using extensive I/O traces and a detailed storage simulator, we demonstrate that this two-level framework can simultaneously meet the latency and throughput requirements imposed by an SLO, without requiring extensive knowledge of the underlying storage system.

UR - http://www.scopus.com/inward/record.url?scp=33750456172&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33750456172&partnerID=8YFLogxK

U2 - 10.1145/1168910.1168913

DO - 10.1145/1168910.1168913

M3 - Article

AN - SCOPUS:33750456172

VL - 2

SP - 283

EP - 308

JO - ACM Transactions on Storage

JF - ACM Transactions on Storage

SN - 1553-3077

IS - 3

ER -