Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I have two Windows 2008 R2 64-bit servers that are identical in pretty much every way. Each of them has an app that basically does a lot of calculations. The hard drive is not really accessed at all. The app is very much CPU bound.

Inexplicably, though, one of the servers is processing 1/3 less data than the other one.

I checked and rechecked but everything seems to be the same. I've gone tab by tab of CPUZ to make sure that absolutely every last piece of memory and timing is the same - all to no avail. I've checked the affinity of each app to make sure it's utilizing all the CPUs.

Each box has 2 physical CPUs (each of which sports 4 cores).

How can I troubleshoot why the one box is slower than the other?

share|improve this question
    
How are you measuring it's speed? –  Soviero Feb 6 '12 at 21:29
    
Start with perfmon's Monitor, Add Counters for Processor->% User Time for <All Instances>. Compare the graphs while the app is running and see if anything looks terribly askew (something should stand out). –  Chris S Feb 6 '12 at 21:31
    
@Kevin I send it 1000 requests to work on and and server 1 is done 33% faster. –  AngryHacker Feb 6 '12 at 21:33
    
Check CPU temperatures. What CPU is this? (If it's a CPU with an auto-overclock / turbo capability, it could just be natural variation among the CPUs.) –  David Schwartz Feb 6 '12 at 21:59
1  
@AngryHacker Which CPU? And what are the temps? (The most likely explanation, I think, is that the slower machine has poor CPU cooling due to a bad fan, improperly mounted heat sink, blocked vent, intake near some other machine's exhaust, or the like.) –  David Schwartz Feb 7 '12 at 0:03
show 2 more comments

3 Answers

up vote 4 down vote accepted

Much like many IT problems you're going to have to divide the problem, measure and then divide again...

If you can I'd move the hard disk/s of the 'slow' server to the hardware of the 'fast' server, does the problem follow the disks? If yes then you've probably got an OS/filesystem/config issue, if not then maybe you've got some broken/semi-broken hardware.

If the problem did move with the disks then I'd suggest you boot from a CD/ISO-based OS and carry out some form of test suite of CPU, memory and disk then repeat the tests but on the regular disk-based OS, this should hel identify if you have an OS or filesystem problem.

Then basically keep dividing up the problem into areas, kind of 50/50 each time (or as close as you can get anyway).

share|improve this answer
add comment

I agree with Chopper3, but I would like to add the following things to consider:

Check the Bios of both systems for differences. Pay special attention to settings related to Power managment, Speedstep/EIST (Intel) or Cool'n Quit (AMD).

If the bios has screen's showing the voltages and/or temperatures compare those as well. Maybe the power-supply's or case-temperatures are different causing one system to have more leeway for the automatic speedup/slowdowns modern CPU's can do.

This sort of thing is not always visible in CPU-Z so it can't hurt to check.

Another thing to check: Are all drivers for the hardware the same ? Pay special attention to motherboard drivers and network cards. A flaky driver here may cause significant slowdowns.

Are both machines on the same network and see the same traffic come by ? If one of them has to deal with much more broadcast traffic than the other this may cause additional load. (I once had a server with a network-card that was left in promiscuous mode after a colleague had experimented with network monitoring software. Took me a day or 2 to figure out what was going on.)

share|improve this answer
add comment

1 - Update all the drivers. 2 - Check in the system properties -> performance options -> tab Processor Scheduling select adjust best performance of: (Programs) I would think thats probably the better, :) At least check if they are equal in both servers.

If they're equal we'll have to try other thing. :)

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.