Upcoming Events

Cloud Connect
Santa Clara
Feb 13-16, 2012

Cloud Connect brings together the entire cloud eco-system to better understand the transformation we're experiencing and promises to be the defining event of the cloud computing industry. Learn about the latest cloud technologies and platforms from thought leaders in Cloud Connect’s comprehensive conference.

Register Now!

F E A T U R E

AppCelera Burns Up the Last Mile

November 26, 2001
By Lori MacVittie

How We Tested Content Accelerators


	Print Full Article Print This Page Download the PDF E-Mail This URL

Once we got these content accelerators in our Green Bay, Wis., Real-World Labs®, we wanted to hammer on real-life performance and ease of installation and configuration. Installation and configuration testing was straightforward, but gauging actual performance was trickier. We needed to emulate Internet conditions inside the lab to determine which products provided the best increase in performance for end users dialing up via low-speed links (14.4 Kbps to 19.2 Kbps).

To do this we configured half our lab as a corporate site, hosting our Web site and the devices under test; we set up the other half to support remote customers. To emulate the Internet, we inserted Shunra Software's Storm between our routers and configured it to throttle the client side of the lab down to 14.4 Kbps with a fixed latency of 100 milliseconds. We also added 4 percent packet loss through the cloud to emulate the real Internet more accurately -- we arrived at the 4 percent number by watching the Internet Traffic Report (www.internettrafficreport.com) for several weeks and taking the average packet-loss rate.

Our corporate Web site was hosted on a Hewlett-Packard Co. NetServer LH6000 running an Apache Web server on Red Hat Linux 7.1. The site itself comprised two pages, one containing 4 KB of text, three static GIF images (22 KB), three GIF banner ads (30 KB) and a small portion of text (256 bytes) dynamically generated via PHP on every request, for an average total of 64 KB. The second page, a static "brochureware" page, comprised 15 KB of text, 105 KB in five JPEG images, 12 KB in GIF images and a 40-KB PNG image, for a total of 172 KB.

Each device was configured as a transparent or a reverse proxy for the Web server. After configuration we turned on the Storm's emulation and retrieved each page numerous times (clearing the cache in between each run) using Microsoft Internet Explorer 5.5 on a single client. We measured the results with Systems Software Technology's TraceWeb32 Plus, which provided accurate data sizes for each object in the Web page as well as download speeds (in bits per second), elapsed transfer times and a view of the HTTP response headers. Unfortunately, no testing tool is available to provide all the functionality of a browser necessary for this test, so we were unable to test the ability of the products to handle massive amounts of traffic. Instead, we concentrated on what we could measure: the decrease in file sizes and the impact on download times.

Storm, Shunra Software, www.shunra.com

TraceWeb32, Systems Software Technology, www.sstinc.com