Exploiting the locality of virtual-machine images to boost the performance of a cloud platform

Shuo Han Chen, Tseng Yi Chen, Chi Heng Lee, Hsin Wen Wei, Tsan Sheng Hsu, Wei Kuan Shih

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Virtualization technology enables multiple virtual machines(VMs) to run on a physical mechanism. Due to the excessive data-intensive workloads of VMs, the system I/O performance of cloud platforms deteriorates. To improve the system I/O performance, one key technique is to exploit the spatial locality of data stored on the back-end disks. However, identifying data spatial locality on a cloud platform becomes challenging due to the transparency feature of virtualization. Therefore, to resolve the problem of poor disk I/O, this paper proposes an inter-VMs locality packing design to increase the number of sequential I/O accesses on back-end disks so as to improve disk I/O efficiency through exploiting the spatial locality of virtual machine images. Moreover, the proposed deign does not compromise the transparency of virtualization technique. A simulator architecture is also proposed to assess the performance of the proposed inter-VMs locality packing design.

Original languageEnglish
Title of host publication37th IEEE Sarnoff Symposium, Sarnoff 2016
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781509015405
DOIs
StatePublished - 7 Feb 2017
Event37th IEEE Sarnoff Symposium, Sarnoff 2016 - Newark, United States
Duration: 19 Sep 201621 Sep 2016

Publication series

Name37th IEEE Sarnoff Symposium, Sarnoff 2016

Conference

Conference37th IEEE Sarnoff Symposium, Sarnoff 2016
Country/TerritoryUnited States
CityNewark
Period19/09/1621/09/16

Fingerprint

Dive into the research topics of 'Exploiting the locality of virtual-machine images to boost the performance of a cloud platform'. Together they form a unique fingerprint.

Cite this