The growth of e-commerce and the Internet in general has pushed Web-based application to their limits and beyond according to a study by Newport Group, Inc., which found that more than half of Web-based applications failed to scale as expected.
According to the study, the average number of concurrent business users that Web applications are designed and developed to support is 3,700. However, once the applications were set for deployment, they handled only 2,650 users on average (72 percent of the traffic volume originally planned).
“Meeting user performance expectations is a formidable challenge given the rapidly evolving technologies, unpredictable user volumes, and fierce competitive pressures of the Web,” said Billie Shea of The Newport Group. “Many companies depending on transaction-based Web applications have suffered significant losses in terms of application redesign costs, unnecessary upgrades, lost business, and weakened competitive position.”
Fifty-two percent of those surveyed said their applications failed to scale as expected. Sixty percent of those said no load or performance tests were conducted prior to deployment of the application. More than three-quarters of those whose application failed to scale did not use load testing until late in the development stage, with 15 percent performing tests after the application’s deployment.
Respondents whose application failed to scale experienced higher time and cost overruns (averaging 23.5 days and $67,083) in comparison to those whose applications did scale. The fail to scale group had a tendency to set their scalability expectations higher than the study’s population as a whole.
Of the 48 percent of respondents whose application scaled as expected, 69 percent reported using a load-testing tool. More than half (56 percent) of the respondents whose application scaled as expected adopted load testing tools early in the process; 21 percent in the design/requirement stage, and 35 percent early in the development process.
Almost two-thirds (66 percent) of those using load-testing equipment said the tools helped them find scalability problems before the application went live. Nearly three-quarters (75 percent) of all study participants concluded the return on investment for load-testing applications was high.
The Newport Group survey was based on responses from 172 Web developers and IT managers overseeing Web applications. The most popular type of Web application deployed by the study participants is business-to-consumer commerce.