问题
I need to achieve 100K or 1Million Request per Second using JMeter. Is it possible to achieve this target. If so, what is the procedure to achieve.
I am using custom plugin Throughput Shaping Timer.
回答1:
Practically it is very near to impossible to achieve such amount of load with a cluster of normal machines( or less powerful server class machines). There are 2 options possible,
- Distributed JMeter environment (with maximum tuned instances) with 100's or 1000's of machines in cluster
Drawback of this is requirement of very skilled personnel and infrastructure issues.
- Cloud based load testing.
Blazemeter, neoload with cloud support can be used to achieve the above target. These providers will have their own hardware and can generate that load also. positive side is you dont have to maintain hw issues, sw issues, log collection, result analysis etc. Negative is they are expensive.
回答2:
You can to use Yandex-tank with Phantom load generator (100k req per second): https://github.com/yandex/yandex-tank
Here the short presentation about their features: http://de.slideshare.net/AlexeyLavrenuke/yandextank-44162563
回答3:
If you thinking about jMeter to prepare test like this then there is one, but hard to bypass, issue - typically one jmeter shouldn't go with more than 100 threads (users) per instance. Of course you can configure remote testing clients, but you will need at least 1000 of them, so - like others mentioned - it can be quite pricey to set the proper machines to create that load. There is one not so expensive service that I know and it allows 100k clients / test - loader.io. It is less than 100$ per month if you need that kind of load so it's reasonably cheap comparing to other services or buying machines that can do it from your side, especially if you want to prepare just some tests now and then. I'm working with it and it is really good and easy service in my opinion.
来源：https://stackoverflow.com/questions/29773286/jmeter-load-testing-with-heavy-load