[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

[Libevent-users] Libevent Schedule Mechanism /Performance



Hi ,
I have a 8 core machine , with 8 threads which have registered with
libevent for accept callbacks.
All it does it accept a connection and close the connection from the thread.
I could see that performance could scale linearly with cpu utilization
till 4 threads , if i spawn more threads the performance does not
increase and also the CPU utilization remains less than 500%.
For a single thread - i get 30k req/per sec - cpu utilization - 99%
2 threads- 40k -198%
4 threads - 60k - 390%
8 threads - 60k - 500-650 %


Can someone shed some light on this behavior ?
How is the scheduling done by libevent when the accepts happen ?
Is there any reference on the libevent code i could tweak ?

I m using this on Linux Platform using pthreads .
***********************************************************************
To unsubscribe, send an e-mail to majordomo@xxxxxxxxxxxxx with
unsubscribe libevent-users    in the body.