forum.e-valid.com

I've heard I have to test my web application with 100,000 users...is that really true? What do you guys think?

yahirv wrote:I've heard I have to test my web application with 100,000 users...is that really true? What do you guys think?

Very good question.

It seems to us you have to ask yourself, first, if you can handle 100 simultaneous users without appreciable degradation in response times?

Then, can you do 1,000 with acceptable degradations?

Then, what about 10,000?

The 100,000 simultaneous user level is something achieved by very, very few websites -- in fact even on the biggest applications in the world (e.g. Google, or Amazon, or Microsoft) it is not clear that 100,000 users are all actually doing something simultaneously. [But there could be 10X that many "signed on" and not doing anything, but simply being signed on does not generate any dynamic response from the servers.]

{And, the total capacity in terms of maxiumum number of users signed on is a different kind of problem than measuring performance, by the way.}

So, we think you have to start small, build up in an orderly fashion, and along the way pay attention to the KEY metric: degradation of user response time as a function of simultaneous active user count.

-- eValid LoadTest Support