You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I went to this page [1] as suggested but could not figure out how to comment on the page to suggest a new framework. So I'm opening an issue here to use as the suggestion.
My suggestion is to make a variation of the existing swoole benchmark. This line [2] tells swoole to use all available cores for its workers. However, in my own tests I have found that if I have e.g. an n CPU box and start n swoole instances with each 1 worker listening on its own unique port, and then put an nginx load balancer infront of it, then with this configuration I have been able to achieve up to 24% more TPS. I think this is because swoole does some things really well and other things not so well, like managing requests between its workers.
If you added this variation of swoole behind nginx then for some tests swoole might be in the #2 spot, barking directly on the heels of vert-x :-)
If you think it works better, give it a shot. That is the nice thing about the TechEmpowered benchmark. It allows for testing frameworks under extreme loads that can differ from small personal tests.
You can copy the Swoole test code here and make a Swoole 2 test with your changes. The current Swoole test will not get any updates until matyhtf has officially solved the plain-text test issue. And he is too busy with Swoole 4.0 and the new Coroutine system.
If your solution works better, i will push any changes and updates to your version in the future and ask for the old version to be removed.
If you added this variation of swoole behind nginx then for some tests swoole might be in the #2 spot, barking directly on the heels of vert-x :-)
I am fairly sure that Swoole with proper PosggreSQL drivers will be able to hit Nr 1 spot in a lot of benchmarks ( its already number one for MySQL in several ). But the driver is unfortunately only half implemented ( missing stuff like prepare statements etc ).
Hi @Wulfklaue, Thanks for the encouragement. I'm willing to try it, but how do I get access to the test environment and boxes etc? Are there instructions for adding components like nginx to a test? Thanks, Simon
I went to this page [1] as suggested but could not figure out how to comment on the page to suggest a new framework. So I'm opening an issue here to use as the suggestion.
My suggestion is to make a variation of the existing swoole benchmark. This line [2] tells swoole to use all available cores for its workers. However, in my own tests I have found that if I have e.g. an n CPU box and start n swoole instances with each 1 worker listening on its own unique port, and then put an nginx load balancer infront of it, then with this configuration I have been able to achieve up to 24% more TPS. I think this is because swoole does some things really well and other things not so well, like managing requests between its workers.
If you added this variation of swoole behind nginx then for some tests swoole might be in the #2 spot, barking directly on the heels of vert-x :-)
HTH,
Simon
[1] https://github.com/TechEmpower/FrameworkBenchmarks/wiki/Suggested-Frameworks,-Languages-and-Features
[2] https://github.com/TechEmpower/FrameworkBenchmarks/blob/master/frameworks/PHP/swoole/swoole-server.php#L8
The text was updated successfully, but these errors were encountered: