The Ohio State University Computer Science and Engineering
Search CSE
    Home  |  Events |  People
Around Campus
  Department
  Calendar
    Calendar Index
    Upcoming Speakers
   
  News
  Research
  Undergrad Programs
  Graduate Programs
  People
  Courses
  Computing Services
 

Faculty Candidate Presentation

Dynamic Resource Management in Internet Data Centers

Bhuvan Urgaonkar
University of Massachusetts

Tues., March 8th
3:30; 480 Dreese Labs
All interested parties are invited.
Refreshments will be served immediately preceding the talk.

Abstract:
Internet applications such as on-line news, retail, and financial sites have become commonplace in recent years. These applications are typically hosted on large clusters of servers known as data centers. Data centers provide performance guarantees to the hosted applications in return for revenue. Two key features of Internet applications make the design of data centers challenging. First, these applications exhibit highly dynamic workloads with multi-time-scale variations. Second, modern Internet applications employ a complex multi-tier architecture, an aspect not fully captured by existing abstractions. Designing a data center to realize the often opposing goals of meeting application performance targets and achieving high resource utilization is therefore a difficult endeavor. In this talk, I will present dynamic resource management mechanisms that a data center can employ to address these difficulties.

I will show how a data center can improve its revenue by careful statistical multiplexing and overbooking of its resources among hosted applications. I will then present a dynamic capacity provisioning technique that allows a data center to meet performance guarantees by dynamically varying the allocated capacity in the presence of varying workloads. A novel aspect of this work is the combination of predictive and reactive mechanisms to deal with the multi-time-scale variations seen in Internet workloads. I have implemented these resource management mechanisms in a prototype data center of forty Linux servers. Using an extensive experimental study with realistic Internet applications on this prototype data center, I will demonstrate the benefits and feasibility of these mechanisms.

Host: Gagan Agrawal

 

 

Home  |  Department  |  Calendar  |  News  |  Research  |  Undergrad Programs  |  Graduate Programs
People  |  Courses  |  Computing Services  |  Ohio State Home  |  Diversity Program  |  College of Engineering  |  E-mail Us

The Ohio State University, Copyright 2004.