In the stress test, it may require the use of graduated compression jmeter. And in the use of graduated compression, most of this tester would doubt: the total set of threads is 100, but the number of threads polymerization report far more than 100, the pressure gradient Why is there such a phenomenon? The facts speak, we use an example to explain this.
surroundings:
First set pressure gradient scenes, visit a website, as follows:
From this pressure gradient parameter set, we can see the total number of threads is 50, an increase of 5 per 2s, increase completed within lS; then holding thread 30s; And then, 5 threads per 2s stopped.
To properly understand the final number of requests, we need to understand that every second thread to release the number of requests
analysis:
Pressure gradient has three stages: the pressure gradient, continuous load, thread releases.
Graduated compression:
If the average response time for the request is 100ms, then the request may 1s clock 10 iterations;
So if you start in the 1s 5 threads, the number is 5 * 10 = 50 requests made within this 1s
2s then run the next wave after the start of loading thread 2s in which the number of requests it issues is 2 * 5 * 10 = 100
After 2s, thread groups and release requests within 1s 5, and runs 2s, 2s in which the number of requests which are issued 2 * 10 * 10 = 200 times (in this case, running threads 10)
And so on, until the thread 50 has been loaded, the number of requests the release of the thread is such that: (2 * 5 * 10) + (2 * 10 * 10) + (2 * 15 * 10) + (2 * 20 * 10) + (25 * 10 * 2) + .... + (2 * 45 * 10) = 4500
Continuous load:
(Note:? Why is not the last 2 * 50 * 10 continuous load because it is from the 30s after 50 thread finished loading, conducted)
This is the 30s, the total number of requests is 30 * 50 * 10 = 15,000
Thread release:
(After the end of the 30s load, gradient release thread starts)
At this time, even if the thread release, still remaining thread initiation request, requests: (2 * 45 * 10) + (2 * 40 * 10) + (2 * 35 * 10) + (2 * 30 * 10) + (2 * 25 * 10) + .... + (2 * 5 * 10) = 4500
Therefore, the total number of requests = 4500 + 15000 + 4500 = 24,000
(Actual test is not the case, because as the load increases, the response time is increased, reducing the number of iterations per second, the final number of requests is reduced. This article describes the idea only.)
This article learn someone else's article.