------=_NextPart_000_0BE8_01BFF237.915C4690
Content-Type: text/plain;
charset="koi8-r"
Content-Transfer-Encoding: quoted-printable
I'm running 2.2.x kernel in specific network load (web counter serving =
15millions of buttons per day). Currently I increased=20
/proc/sys/net/ipv4/tcp_max_syn_backlog to 2048=20
/proc/sys/ner/ipv4/route/max_size to 16384
The number of connections open is:
[root@mars dem]# netstat -nt | wc -l
15395
[root@mars dem]# netstat -nt | grep SYN | wc -l
358
So it does not seems to hit the limit of max_syn backlog.
The problem is - it seems in 1-3% of the requests connect takes several =
seconds which, which is the problem in my case.
Looking at VMSTAT shows 40-60% of CPU is idle anytime.
Any ideas how to solve such problem ?
I have tried several stable kernels from 2.2.14 to 2.2.16 with =
different patch sets (RedHat,aa-series) all seems to have the same =
problem....
------=_NextPart_000_0BE8_01BFF237.915C4690
Content-Type: text/html;
charset="koi8-r"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">