Scalable load balancing in cluster storage systems

Gae Won You, Seung Won Hwang, Navendu Jain

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

12 Citations (Scopus)

Abstract

Enterprise and cloud data centers are comprised of tens of thousands of servers providing petabytes of storage to a large number of users and applications. At such a scale, these storage systems face two key challenges: (a) hot-spots due to the dynamic popularity of stored objects and (b) high reconfiguration costs of data migration due to bandwidth oversubscription in the data center network. Existing storage solutions, however, are unsuitable to address these challenges because of the large number of servers and data objects. This paper describes the design, implementation, and evaluation of Ursa, which scales to a large number of storage nodes and objects and aims to minimize latency and bandwidth costs during system reconfiguration. Toward this goal, Ursa formulates an optimization problem that selects a subset of objects from hot-spot servers and performs topology-aware migration to minimize reconfiguration costs. As exact optimization is computationally expensive, we devise scalable approximation techniques for node selection and efficient divide-and-conquer computation. Our evaluation shows Ursa achieves cost-effective load balancing while scaling to large systems and is time-responsive in computing placement decisions, e.g., about two minutes for 10K nodes and 10M objects.

Original languageEnglish
Title of host publicationMiddleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings
Pages101-122
Number of pages22
DOIs
Publication statusPublished - 2011 Dec 23
Event12th ACM/IFIP/USENIX International Middleware Conference, Middleware 2011 - Lisbon, Portugal
Duration: 2011 Dec 122011 Dec 16

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7049 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other12th ACM/IFIP/USENIX International Middleware Conference, Middleware 2011
CountryPortugal
CityLisbon
Period11/12/1211/12/16

Fingerprint

Storage System
Load Balancing
Resource allocation
Reconfiguration
Servers
Server
Data Center
Costs
Hot Spot
Migration
Bandwidth
Vertex of a graph
Minimise
Set theory
Divide and conquer
Evaluation
Topology
Placement
Latency
Object

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

You, G. W., Hwang, S. W., & Jain, N. (2011). Scalable load balancing in cluster storage systems. In Middleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings (pp. 101-122). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7049 LNCS). https://doi.org/10.1007/978-3-642-25821-3_6
You, Gae Won ; Hwang, Seung Won ; Jain, Navendu. / Scalable load balancing in cluster storage systems. Middleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings. 2011. pp. 101-122 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{99f762d929a24e4e909e65beb06a4a30,
title = "Scalable load balancing in cluster storage systems",
abstract = "Enterprise and cloud data centers are comprised of tens of thousands of servers providing petabytes of storage to a large number of users and applications. At such a scale, these storage systems face two key challenges: (a) hot-spots due to the dynamic popularity of stored objects and (b) high reconfiguration costs of data migration due to bandwidth oversubscription in the data center network. Existing storage solutions, however, are unsuitable to address these challenges because of the large number of servers and data objects. This paper describes the design, implementation, and evaluation of Ursa, which scales to a large number of storage nodes and objects and aims to minimize latency and bandwidth costs during system reconfiguration. Toward this goal, Ursa formulates an optimization problem that selects a subset of objects from hot-spot servers and performs topology-aware migration to minimize reconfiguration costs. As exact optimization is computationally expensive, we devise scalable approximation techniques for node selection and efficient divide-and-conquer computation. Our evaluation shows Ursa achieves cost-effective load balancing while scaling to large systems and is time-responsive in computing placement decisions, e.g., about two minutes for 10K nodes and 10M objects.",
author = "You, {Gae Won} and Hwang, {Seung Won} and Navendu Jain",
year = "2011",
month = "12",
day = "23",
doi = "10.1007/978-3-642-25821-3_6",
language = "English",
isbn = "9783642258206",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "101--122",
booktitle = "Middleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings",

}

You, GW, Hwang, SW & Jain, N 2011, Scalable load balancing in cluster storage systems. in Middleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7049 LNCS, pp. 101-122, 12th ACM/IFIP/USENIX International Middleware Conference, Middleware 2011, Lisbon, Portugal, 11/12/12. https://doi.org/10.1007/978-3-642-25821-3_6

Scalable load balancing in cluster storage systems. / You, Gae Won; Hwang, Seung Won; Jain, Navendu.

Middleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings. 2011. p. 101-122 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7049 LNCS).

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

TY - GEN

T1 - Scalable load balancing in cluster storage systems

AU - You, Gae Won

AU - Hwang, Seung Won

AU - Jain, Navendu

PY - 2011/12/23

Y1 - 2011/12/23

N2 - Enterprise and cloud data centers are comprised of tens of thousands of servers providing petabytes of storage to a large number of users and applications. At such a scale, these storage systems face two key challenges: (a) hot-spots due to the dynamic popularity of stored objects and (b) high reconfiguration costs of data migration due to bandwidth oversubscription in the data center network. Existing storage solutions, however, are unsuitable to address these challenges because of the large number of servers and data objects. This paper describes the design, implementation, and evaluation of Ursa, which scales to a large number of storage nodes and objects and aims to minimize latency and bandwidth costs during system reconfiguration. Toward this goal, Ursa formulates an optimization problem that selects a subset of objects from hot-spot servers and performs topology-aware migration to minimize reconfiguration costs. As exact optimization is computationally expensive, we devise scalable approximation techniques for node selection and efficient divide-and-conquer computation. Our evaluation shows Ursa achieves cost-effective load balancing while scaling to large systems and is time-responsive in computing placement decisions, e.g., about two minutes for 10K nodes and 10M objects.

AB - Enterprise and cloud data centers are comprised of tens of thousands of servers providing petabytes of storage to a large number of users and applications. At such a scale, these storage systems face two key challenges: (a) hot-spots due to the dynamic popularity of stored objects and (b) high reconfiguration costs of data migration due to bandwidth oversubscription in the data center network. Existing storage solutions, however, are unsuitable to address these challenges because of the large number of servers and data objects. This paper describes the design, implementation, and evaluation of Ursa, which scales to a large number of storage nodes and objects and aims to minimize latency and bandwidth costs during system reconfiguration. Toward this goal, Ursa formulates an optimization problem that selects a subset of objects from hot-spot servers and performs topology-aware migration to minimize reconfiguration costs. As exact optimization is computationally expensive, we devise scalable approximation techniques for node selection and efficient divide-and-conquer computation. Our evaluation shows Ursa achieves cost-effective load balancing while scaling to large systems and is time-responsive in computing placement decisions, e.g., about two minutes for 10K nodes and 10M objects.

UR - http://www.scopus.com/inward/record.url?scp=83755228953&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=83755228953&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-25821-3_6

DO - 10.1007/978-3-642-25821-3_6

M3 - Conference contribution

AN - SCOPUS:83755228953

SN - 9783642258206

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

SP - 101

EP - 122

BT - Middleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings

ER -

You GW, Hwang SW, Jain N. Scalable load balancing in cluster storage systems. In Middleware 2011 - ACM/IFIP/USENIX 12th International Middleware Conference, Proceedings. 2011. p. 101-122. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-25821-3_6