Hi,

we are currently looking for a bottleneck in our application stack. We use Nginx as a reverse proxy which dispatches requests to an asynchronous backend (Twisted Python).
When looking at the stub_stats output we notice a large deviation between reads and writes. Strangely, we notice substantially more reads than writes (about 4-6 times as much). Obviously read and write stats are only a snapshot of the counts in this very moment in which we poll stub_stats. Should we worry about such a huge discrepancy between reads and writes? How should we interpret these numbers?

More detailed plots can be found here:

Read/Writes
https://picasaweb.google.com/lh/photo/u1wm7uUBECTgSh_hrCtsutMTjNZETYmyPJy0liipFm0?feat=directlink

Waits
https://picasaweb.google.com/lh/photo/1A4I6GJblxoZMGWemx25GtMTjNZETYmyPJy0liipFm0?feat=directlink

thanks,
Peter

I just realized that I've posted this question on the wrong mailing list (Php-fpm) - I've re-opened it on the Nginx Mailing list:
http://forum.nginx.org/read.php?2,223788

Sorry for any inconveniences caused.