Use Application Center Test (ACT) to do stress testing "go"

After we have completed the development based on SPS2003 and implemented our specific application, can we directly invite users to use it? If I do this, then experienced developers will certainly scoff at this: even stress test do not do! I really don't want to live ...

Oh, yes. Development environment is often only consider the function, to the specific environment, you need to consider a large number of users to access the time, many features will not be wrong? What about performance? ...... Here's a quick look at how to do stress tests.

I believe there are a lot of tools for stress testing, and many of them are now application Center Test (ACT). This east is a vs.net in the formation, very simple, easy to get started, and support scripts, can also achieve complex functions ...

This omits the test steps, assuming we are just testing a website directly, such as Test.sendrequest ("http://server/default.aspx"). Now how to analyze the results?

Here are some of the information I just learned, and share with you, I hope for experienced friends, play a starting function.

1. First, check that there are no errors, such as 401 users do not verify the error. If there is an error, then the result must be wrong, and do not have to read.

2. Analysis of average requests per second should be "average request per second".

We can test several times, using 1, 2, 5, 10, 50, 100, 200 ... Number of concurrent browser connections. Then, we can put a few results in a chart to analyze.

In general, as the number of concurrent browser connections increases, the number of Average requests per second increases, but when a certain value is added, it causes Average requests per second to drop. The value, then, is almost the maximum number of concurrent browser connections that the server can support.

3. Average time to last byte

Is the time after the request is sent to the end of the received server response.

Obviously, in general, as the number of concurrent browser connections increases, this value will become larger. In general, the analysis of this value is not reasonable, you can refer to the following criteria:

0.1 seconds: very fast.

1 seconds: The speed is still very fast, basically do not consider the performance problem

3–4 seconds: An acceptable result for the internal network

5–8 seconds: An acceptable result for an external network

More than 10 seconds: too slow some

4. Average time to first byte

is just the time it takes to send a request to the beginning of receiving the server response.

You are welcome to discuss and publish your experience on the act.