This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Technical Report TR-08-07: An Evaluation of Amazon’s Grid Computing Services: EC2, S3 and SQS Simson L. Garfinkel Center for Research on Computation and Society School for Engineering and Applied Sciences Harvard University Cambridge, MA [email protected] Abstract Amazon.com’s Elastic Compute Cloud (EC2), Simple Storage Service (S3) and Simple Queue Service (SQS) offer enterprise-class computing, storage and coordina- tion facilities to any organization or individual in the world with a valid credit card. This paper details our experience working with these commodity grid comput- ing services between November 2006 and May 2007, including an analysis of the overall system’s API and ease-of-use; an analysis of EC2’s management and secu- rity facilities; an end-to-end performance analysis of S3’s throughput and latency as observed from Amazon’s EC2 cluster and other locations on the Internet; and an analy- sis of the SQS operation and performance. We conclude with a report of our experience moving a large-scale re- search application from dedicated hardware to the Ama- zon offering. We find that this collection of Amazon Web Services (AWS) has great promise but are hobbled by service consistency problems, the lack of a Service Level Agreement (SLA), and a problematic Web Services Li- censing Agreement (WSLA). 1 Introduction In 2006, Amazon Web Services (AWS), a subsidiary of Amazon.com, began offering three web services that allow organizations and individuals to use Ama- zon’s enterprise-class computing infrastructure on an as- needed basis and at commodity prices. Amazon’s Elas- tic Compute Cloud (EC2) rents Linux virtual machines at 10 cents per CPU hour; users may rent dozens, hun- dreds, or even thousands of CPUs simultaneously. Ama- zon’s Simple Storage Service (S3) allows users to store data at a cost of 15 cents per gigabyte per month, with bandwidth priced at between 10 and 18 cents per giga- byte transferred outside Amazon’s network. Amazon’s Simple Queue Service (SQS) is a reliable messaging ser- vice, making it relatively straightforward to coordinate a cluster of computers on a parallelizable large-scale prob- lem, at a cost of 10 cents per thousand messages. All of these services are sold without startup fees and with no minimum pricing, potentially making them attractive for individuals, universities and corporations alike. 1.1 Utility Computing with AWS Businesses, universities and government users may con- sider cluster computing as a potential solution to their computing needs. But building and operating even a relatively small cluster can be a formidable undertak- ing, requiring not just money but physical space, cooling, power, and management resources. Planning and build- ing a cluster requires considerable investment. Once built, it can be a challenge to keep the cluster both suf- ficiently utilized to justify the expense and sufficiently under-utilized so that there is room for existing projects to grow their resource requirements as necessary.to grow their resource requirements as necessary....
View Full Document
This note was uploaded on 12/08/2011 for the course CS 525 taught by Professor Gupta during the Spring '08 term at University of Illinois, Urbana Champaign.
- Spring '08
- Grid Computing