Slow Disk Performance on 2012 Standard

[andytruelove]

Hi

 

I have a problem with slow disk access on a new Server 2012 Standard edition build. The spec. of the server is:

 

Supermicro 7047R with 2 x Xeon 2.3Ghz 6 Core

128Gb RAM

LSI 2208 SAS/3 RAID Controller

2 x 147Gb Seagate 10K (RAID-1) – O/S

4 x 600Gb Seagate 15K (RAID-10) – “Data”

 

All hardware has had the latest firmware\bios updates applied and Windows Server is fully patched. The server will be a HyperV Host for 3 – 4 guest 2008/R2 servers.

 

The performance problem we’re having occurs with sustained disc activity. For example if we copy multiple 17Gb files to new folder(s) on the same volume it’s fine for the first three files with an average transfer rate of around 300Mbps. Once the total data transferred reaches between 50-70Gb the transfer rate then drops down until it reaches around 2Mbps and the copy then seems to freeze up altogether. This kills the other services on the server and leaves it needing a hard reset.

 

Once the transfer rate starts dropping the memory usage on the server rapidly increases from around 10Gb to 50-70Gb. I’ve also noticed that all the files copied continue to show up under Disk Activity in Resource Manager with a PID of 4 (system). Any ideas why please? This doesn’t seem to happen on other servers I’ve checked. There’s literally nothing else running on the server whilst this is happening, bare basic install plus drivers.

 

I’m not 100% sure if the problem is just confined to large file copies or whether it also happens under other sustained disk load. The server isn’t yet in production (fortunately) and I’m unsure how to generate a similar load. If I don’t do a large file copy then it seems to be OK running the 4 Guests but obviously worried about making it live with this problem. I’m running an overnight burn in test to hopefully help prove this. Write caching is disabled by the way.

 

Hopefully someone reading has some ideas of how I can go about fixing this please? The hardware vendors don’t think it’s a hardware problem and all their diagnostics come out clear. I’ve also been through various articles without any joy which has been a wasted two days.

 

Thanks in advance and sorry for long post!

 

Andy Truelove

[xenonive]

How are you copying the files? Tried xcopy, robocopy or similar via command prompt to see if faster? Maybe try older raid drivers if you have latest one in case bug introduced. Reason to have write caching disabled? Would think that would give you big performance increase if enabled. I find using a different program to copy files better. I use teracopy on windows server 2008R2 with no issues so might work for 2012 but it uses its own file buffers rather than windows ones and reports more accurately the copy speed.

[xenonive]

Other things to look could be antivirus scanning or UAC when copying which might explain system process if not running as administrator

[andytruelove]

Thanks for your responses. For anyone else reading this with the same problem we were able to confirm the issue only occured when copying large files with a buffer so using XCOPY /J etc is fine. We had some dialogue with Microsoft and this is the same problem that's been around since Server 2008! Enabling Write Cache both on the RAID controller and with the Windows setting has helped (just need to get a BBU now) and other intensive read\write ops with smaller files have been fine so we're carrying on and hoping for the best.

 

Thanks again