Characterizing Infrastructure Cloud Performance for Scientific Computing

Project Information

Discipline
Computer Science (401) 
Orientation
Research 
Abstract

There is much variation in the world of cloud computing that may impact application performance. The use of system virtualization technologies in Infrastructure-as-a-Service clouds (IaaS) allows users to deploy custom software stacks, down the the operating system level. These same virtualization technologies also allow multiple virtual machines (VMs) to be co-located on a single node. Cloud providers also have different hardware and software configurations. These issues can lead to performance variability. This project will focus on identifying metrics and benchmarks to evaluate the performance of IaaS clouds. Furthermore, it will develop necessary tools and frameworks to automate the process of benchmarking and analyzing the performance of IaaS cloud offerings. Automated benchmarking toolkits will enable users to repeatedly evaluate the performance of IaaS clouds as configurations and offerings continually change.

Intellectual Merit

This project will provide a platform for understanding performance variations between different IaaS clouds, giving users the ability to thoroughly evaluate and compare different offerings.

Broader Impacts

Understanding the performance characteristics of IaaS clouds will provide useful information to users of all disciplines looking to leverage IaaS clouds for their work.

Project Contact

Project Lead
Theron Voran (voran) 
Project Manager
Theron Voran (voran) 
Project Members
Domenic Murtari  

Resource Requirements

Hardware Systems
  • alamo (Dell optiplex at TACC)
  • foxtrot (IBM iDataPlex at UF)
  • hotel (IBM iDataPlex at U Chicago)
  • india (IBM iDataPlex at IU)
  • sierra (IBM iDataPlex at SDSC)
 
Use of FutureGrid

FutureGrid will aid in the development of the performance characterization tools by providing access to IaaS clouds, such as Nimbus and OpenStack clouds.

Scale of Use

Initially this project will consist of 2-3 members using IaaS clouds for development and minimal benchmarking. Most work should typically involve tens of VMs.

Project Timeline

Submitted
09/03/2013 - 17:06 
»