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 connections 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 that).
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:
#### state ------------------ 65 CLOSING 94 ESTABLISHED 172 FIN_WAIT1 50 FIN_WAIT2 10 LAST_ACK 497 SYN_RECV 43480 TIME_WAIT
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
thanks
nate