IBM 5500 M20 Performance trouble

[klunde]

Hi!

Among our IBM servers I have two 5500 M20 servers, both with ServeRaid 3H controllers, dual CPU.

I recently discovered that both servers are using about 40% of its CPU power to handle hardware interrupts. Checking my other servers none of them seems to ever use more than one 1%. Also a perfom check showed me that the two servers in question have about 65000 Interrupts/sec (from perfmon processor count) - other servers have about 100-400.

I've tried calling IBM, but their only suggestion was to strip down the server and insert the hardware one-by-one and see if I could find the troubling part, but as this is a working production server that is not my first option.

Anyone with any suggestion on this issue?`

</Morten>

 

[Mammoth]

Well that really is the only way to see what piece of hardware is hitting the proc so hard. The reason why IBM suggested that method is because the hardware group does not support software. So suggesting to run a piece of software that monitors the bus is outside the scope of hardware support. You could try to google for some kind of hardware sniffer. It could also be software that is causing the hardware to send so many interrupts.

 

[klunde]

Well, problem is, as I said earlier, this is a working production server and I do not have the option to turn it off and strip it down. I either have to buy and configure a new on e or find the error. So, any suggestion on what kind of software that can be the cause of sending so many interrupts?


</Morten>

 

[catorze]

40% is within the operating specs of the processor...what's the problem? Are you sure you're comparing apples to apples?

 

[klunde]

No, I'm not sure, but when I found that I have two servers who are more or less identical in hardware, but totally different in usage / installed software (on Com+ application / File server and one Oracle DB Server) I found it suspicious that both these two servers has so high interrupt counts while none of my other 15 server have any numbers even in that range (70' plus compared to 100-400 as stated above)

And the problem? I want more juice out of my server - its running to slow and I suspect that this is a problem and therefor I would like to fix it rather than spending a lot of money on buying a new one.


</Morten>

 

[klunde]

Phuh!

I finally managed to figure out why the two servers was using up 40% of their CPU capacity to hardware interrupts. The Parallell port setup in bios was configured as "auto" and there was no printers connected to the port. I reset the bios and - viola! - now my servers are running with full CPU capacity again.

</Morten>

 

[rhayre]

had the same problem, it was the parellel port set to the wrong mode. try set it to the default mode.