-
-
Notifications
You must be signed in to change notification settings - Fork 220
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Consider increasing # of default workers? #126
Comments
Somehow I tab+enter'd that. Didn't mean to submit yet. What do you think about:
|
We should also update the documentation to recommend triggering the garbage collector of PHP after every request in worker mode. This should improve the performance. |
Please, add the patch to the benchmark. |
When you want we talk!! |
The default number of workers is highly specific to the physical architecture it is running on but there's no actual scheduling of each worker on a different core. The default is wholly arbitrary and something that kinda makes sense. Right now, it appears the default is too low. I have no desire to update the benchmarks, and that isn't all I tested with to arrive at 2x cores. I did share the fact that it improves the benchmark... because I'm trying to improve performance... and that would improve the benchmarks. The nice thing about these benchmarks is that they're nicely reproducible and thus, useful. PS: I have a big day tomorrow and if it works out, I'll be around throughout the holidays to dig into some issues here. I've been seriously itching to write some code but life has been crazy these last few weeks. |
Currently, there is one worker spawned per core (if I'm understanding correctly).
This severely limits concurrent connections as it appears only one request per worker can be handled at a time.
Changing this to 2x dramatically affects the techempower benchmarks:
from:
to
Anything beyond 2x has diminishing returns in my testing. FWIW, 2x on my hardware is 32 workers.
The text was updated successfully, but these errors were encountered: