Anyone have experience with these kernel parameters? I can't find a whole
lot of info on them.
I have a situation where I'm trying to stress test a server, dual proc quad
core with a tomcat application that is really fast. We can get it to about
60% CPU usage, and processing about 1,350 requests a second but we can't get
much past that. It seems we are maxing out the sockets on the system, with
more than 60,000 connections in TIME_WAIT. Our requests are really small and
light and short lived. So I was thinking of using these parameters to get
more out of the tcp stack. Our load balancer is configured to terminate
to the world after 25 seconds(a typical request takes 0.1 seconds to
complete), but I'm not sure if that applies to the systems behind the load
balancer as well(I suspect not because there's no apparent setting for
Currently I have the system configured to take a lower amount of traffic,
hovering at around 47% CPU, at around 833 req/s and it has 43k
connections in these states:
On a side note, the 'active tcp sockets' reported by sar seems wildly
inaccurate, it reports only ~10 active tcp sockets, it barely varies
from the system being idle to the system being maxxed out.
The docs I can find says seek expert assistance..any experts reading this ?
Running CentOS 4.6
CentOS mailing list