Reputation: 38083
I've created a web load test for my site, ramping up from 1 to 300 users. The performance recorded on the test is outrageous, totally off the scale from the worst performance I could imagine - imagine response times for a simple login script taking up to 10 minutes!
My immediate suspicion falls on the fact that I'm simulating 300 clients on a single computer. I would guess that my own computer's processing power and internet bandwidth are much more of a bottleneck than the server.
Is there any better way of testing my web site's performance that will yield more realistic results?
Upvotes: 2
Views: 2799
Reputation: 24556
At first, you can run load tests from a developer machine but it's generally not recommanded for large load tests : you machine will host a web site, run a load test, simulate virtual users, store tests results in a local db ... It's easy to understand that your machine will become the bottleneck in your load test. It's only good for small smoke tests.
In order to simulate a lot of users, you'll set up what is known as a Load Test Rig.If possible try to Host your web site on a separate machine, and use TestController and Agents. The controller manages and coordinates the agent machines and the agents generate load against the application (hosted somewhere else).
I have been a user of Visual Studio Load Testing for 2 years, and I have found it very powerfull and easy to use. There are also many hidden of not documented features/properties that are very important to understand
Some of them are listed here.
I also generally do not run load tests using a step pattern but a goal based pattern targeting 50% CPU on the web server.
Upvotes: 2