InformationWeek.com

Welcome Guest. | Log In| Register | Membership Benefits

January 17, 2000

Web-Site Testing Tools:
Is Your Web Site Scalable Enough?
Scalability of Web applications has never been more important. We looked at three products--RadView's WebLoad 3.51, RSW's e-Load 4.0, and Mercury Interactive's Astra LoadTest 3.0--that make it easy to stress-test your Web applications and keep your site up and running.

By Jason Levitt

Related links:

sidebar: Be Realistic When Testing Web Sites

New Software Tools Help Monitor And Test Web Apps

And from our sister publications:

EETimes Credence rolls built-off self-test

Send Us Your Feedback

n a perfect world, thousands of users, sitting at their computers, would be at your beck and call, ready to slam your Web application with their mouse clicks and keyboard entries, testing its ability to scale. In the frantic world of dot-com development, however, there are no such users--and if your Web application can't scale to handle hundreds, thousands, or even millions of real customers, you stand to lose not only revenue, but reputation. In the online realm, Web-application response is synonymous with customer service.

It's no wonder that Web-application stress-testing tools are hot and that there are several from which to choose. InformationWeek Labs tested three tools, RadView's WebLoad 3.51, RSW's e-Load 4.0, and Mercury Interactive's Astra LoadTest 3.0. All three products run under Windows NT 4.0 (though they can test the scalability of Web applications running under any operating system), are easy to set up and use, and can simulate hundreds of users pounding away at a Web site.

Though these three products are essentially similar, and any one of them will have you stress-testing your site within an hour of installing the software, they have very different overall capabilities, implementation, and ease-of-use. RSW's e-Load 4.0 was our favorite.

It offers the widest choice of client types, has excellent real-time charting capabilities, and its graphical user interface is well-designed for repetitive testing situations.

Mercury Interactive's Astra LoadTest 3.0 is a competent competitor that excelled at data validation, catching situations in which links and checkboxes changed during playback, and its playback client can handle complex user-interaction scenarios, such as those that require execution of ActiveX controls. But LoadTest's playback client is also a resource hog that limited the number of virtual users we could play back from a single machine (a service pack update we received near the end of testing adds an HTTP playback mode, but no medium functionality mode). Mercury's solution, for customers who want to generate more than 250 virtual users, is to buy its flagship LoadRunner product.

RadView's WebLoad 3.51 doesn't have the advanced data validation, functional testing, and robust client capabilities of the other two products, but its price is only about one quarter as much. WebLoad's pure JavaScript approach requires more manual script setup, but works fine for sites that generate fairly simple pages and don't need the missing features.

The tested products are focused on the busy dot-com development market, where stress-testing Web applications using HTTP is the primary concern. Quality-assurance testers, developers, performance analysts, and any IT staff responsible for maintaining operations of heavily used systems would find these tools useful. Several similar but more-general and powerful products for stress and functional testing were not considered for this review. Compuware's QALoad, Mercury Interactive's LoadRunner, Rational Software's Rational Suite Performance Studio, and Segue Software's SilkPerformer are high-end client-server testing products that can test a wide variety of applications (not just Web applications) with protocols other than HTTP. These products are more expensive than the tested products, as well as more difficult to set up and use.

Although evaluation versions of the tested products are freely downloadable, customers interested in getting their feet wet with a less-powerful, no-obligation, Web-application stress-testing tool should consider downloading Microsoft's Web Application Stress Tool (webtool.rte.microsoft.com), or the less-powerful Microsoft Web Capacity Analysis Tool (msdn.microsoft.com/workshop/server/toolbox/wcat.asp).

All three products are designed to be used offline. That means you use them against your development site, not your live site. Although they could be used against your live site, the possibility of denying service to customers, as well as the lack of a clean-room test-bed environment, makes it impractical.

For the most part, the products work similarly. The tools let you create an artificial workload script by recording your mouse clicks and keyboard input as you traverse through your Web application with a Web browser. The products then instantiate any number of virtual users, which can be distributed across multiple machines, that play back the scripts, creating a load on your Web application.

continued...page 2, 3, 4