Optimizing Resource Utilization of a Data Center

Xiang Sun, Nirwan Ansari, Ruopeng Wang

Research output: Contribution to journalReview articlepeer-review

47 Scopus citations

Abstract

To provision IT solutions with reduced operating expenses, many businesses are moving their IT infrastructures into public data centers or are starting to build their own private data centers. Data centers can provide flexible resource provisioning in order to accommodate the workload demand. In this paper, we present a comprehensive survey of most relevant research activities on resource management of data centers that aim to optimize the resource utilization. We first describe the resource overprovisioning problem in current data centers. Then, we summarize two important components in the resource management platform and present the benefit of accurately predicting the workload in resource management. Afterwards, we classify existing resource management in a data center into three categories: 1) virtual machine-based, 2) physical machine-based, and 3) application-based resource management mechanisms. We discuss the performance degradation for implementing these three kinds of resource management in a heterogeneous data center. Finally, we present three important issues arose in the data center resource management and some potential approaches to address the issues. This paper presents a timely survey on resource management in a data center, and provides a comprehensive reference for further research in this field.

Original languageEnglish (US)
Article number7458812
Pages (from-to)2822-2846
Number of pages25
JournalIEEE Communications Surveys and Tutorials
Volume18
Issue number4
DOIs
StatePublished - Oct 1 2016

All Science Journal Classification (ASJC) codes

  • Electrical and Electronic Engineering

Keywords

  • Resource management
  • data center
  • resource allocation
  • resource utilization
  • workload prediction

Fingerprint

Dive into the research topics of 'Optimizing Resource Utilization of a Data Center'. Together they form a unique fingerprint.

Cite this