Monitoring and Tuning Your Server
Monitoring and Tuning Your Server |
|
In some situations, the Web server processor is the bottleneck. Increase the stress to the point where the number of requests per second starts to decrease, then back the stress off slightly. This is the maximum performance that the Web site can realistically achieve. (You increase the stress by increasing the Stress level setting—the number of threads— or by increasing the number of Web Application Stress Tool client machines.)
On the computer that is monitoring the Web server, use PerfMon to watch the Computername\ System\ % Total Processor Time, Computername\ Web Service\ Connection Attempts/sec, and Computername\ Active Server Pages\ Requests Queued counters of the Web server. If the processors are running at 80 to 85 percent of capacity, then they are most likely to create a bottleneck. On the other hand, if the number of Requests Queued fluctuates considerably during the stress test and the processor utilization remains relatively low, this is an indication that the script is calling a server COM component that is receiving more calls than it can handle. In this case, the server COM component is the source of the bottleneck.
It is a good idea to have an expected peak load that meets your business needs and then create a test that uses enough threads to reach that capacity. This will help you to determine the maximum request rate for the Web application and to confirm that it is in line with the expected peak load.
The amount of personalization in a Web application also plays a significant role in how it performs. The Web Application Stress Tool contains several features that make testing a personalized Web site easy. For example you can create users, and allow the Web Application Stress Tool to save a cookie with each of them. You can also use the Querystring editor to help create and store several sets of name-value pairs that are passed with each request.