Windows 2008 Enterprise Server only user 1 CPU to handle network load

we have windows 2008 enterprise cluster sql server 2005 installed.

during peak time, when ping server, got pretty bad response time:

reply 10.1.1.137: bytes=32 time=12ms ttl=128
reply 10.1.1.137: bytes=32 time<1ms ttl=128
reply 10.1.1.137: bytes=32 time<1ms ttl=128
reply 10.1.1.137: bytes=32 time<1ms ttl=128
reply 10.1.1.137: bytes=32 time=104ms ttl=128
reply 10.1.1.137: bytes=32 time=47ms ttl=128
reply 10.1.1.137: bytes=32 time=10ms ttl=128
reply 10.1.1.137: bytes=32 time<1ms ttl=128
reply 10.1.1.137: bytes=32 time<1ms ttl=128
reply 10.1.1.137: bytes=32 time<1ms ttl=128

then notice cpu 0 seems cpu handling network load due following observations:

1. when cpu0 usage up, ping poor, when down, good.

2. check performance counter processor:%dpc time  , processor: %processor time , find matched, means cpu0 dedicated dpc. check other 7 cpus no dpc time found.

we have rss (receive-side scaling enhancements ) enabled on os , nic level should spread network load cpus, right?

so missing here?

btw:

cpu intel(r) xeon(r) cpu e7420@2.13ghz (2 processors)

nic intel(r) pro/1000pt dual port server adapter (teamed)

thanks

 

kelly

 

for reason working now.

the obvious difference can see dpc lantency somehow totally changed.

for example, dpc latency check, when copy big file, used red columns, green.

since problem scenario gone, cannot find root cause now.

btw, try turn off anti virus software.

thanks anyway.

 

 

 

Windows Server  >  High Availability (Clustering)