eValid™ -- Automated Web Quality Solution Browser-Based, Client-Side, Functional Testing & Validation, Load & Performance Tuning, Page Timing, Website Analysis, and Rich Internet Application Monitoring.

Synopsis
A well known national high-end retail and web-based home furnishings distributor was undergoing a major refurbishment of their web store, and needed confirmation of web server capacity in terms of number of simultaneous users in preparation for a big on-the-web promotion. The eValid team prepared tests that mimiced the existing customer behavior based on usage analysis of the pre-upgrade website. We used this behavior model to demonstrate a capacity increase in excess of 150% in the upgraded servers.

Background
This is a very typical case of confirmation of capacity of a web server upgrade, where the current site already had a very good statistical base of experience. At issue was whether the change to the new versions of the server software would equal or exceed the capacity.

eValid Application Description
The server loading and capacity analysis process involved these stages:

1. Creation of a "standard scenario" that closely matched the activities of users based on detailed statistics from the existing website. This resulted in a set of 10 scripts with the property that when run in parallel the mix of work (navigation, login, item selection, pricing and shipping calculation, and final order entry, etc.) was as close as possible to existing measured norms.
2. Perfection of the scripts so that they could run in multiple-parallel and multiple-machine configurations.
3. Establishment of a feedback loop that revealed basic CPU utilization statistics on one of the multiple machines in the server complex.
4. Application of the standard scenario in increasing multiples while observing the achieved CPU utilization.
5. Interpolation and extrapolation of the numeric results to arrive at a predicted baseline load capacity.

This work required about 6 weeks of technical analysis, script development, and a series of eValid LoadTest runs.

Results Achieved
Once the basic 10-user scenario was established, we applied increasing numbers of simultaneous numbers of eValid users using multiple machines and multiple DSLs. As the number of simultaneous users increased we studied the amount of CPU utilization. From the collected data we could make a reliable prediction on the ultimate capacity

Customer Comments
"We have been very pleased with the results we've seen from the eValid team's work. The main thing we have learned is the value of ultra-realistic load testing -- it assures we have the right server complex, with enough capacity, and sufficient stability, to meet our upcoming needs. -- Y. N.