On Apr 21, 2014, at 12:46 , Igor Sysoev wrote: > On Apr 20, 2014, at 21:17 , Lukas Tribus wrote: > >>> Hi >>> >>> I was watching this video by fastly ceo http://youtu.be/zrSvoQz1GOs?t=24m44s >>> he talks about the nginx ssl handshake versus apache and comes to the >>> conclusion that apache was more efficient at mass handshakes due to >>> nginx blocking while it calls back to openssl >>> >>> I was hoping to get other people's opinion on this and find out if what >>> he says is accurate or not >> >> I would be interested in opinions/statements about this as well. > > nginx worker blocks on SSL handshake as well as on disk I/O. But in contrast > to the disk I/O during the handshake you can not do anything else on this CPU > at this time until CPU time share slice will end for the current > process/thread. > If a typical time share slice is 1ms and the 1024-bit key handshake time is > 0.5ms > then the chances that another process/thread is able to run on the same CPU > are 25%.
Sorry, 50%. > The lesser handshake time the lesser chances. -- Igor Sysoev http://nginx.com _______________________________________________ nginx mailing list [email protected] http://mailman.nginx.org/mailman/listinfo/nginx
