Poor performance on NAS1200s

[Clinton Thomson]

I have recently installed an HP NAS1200s into my 2000 domain environment to be used as a disk backup solution. We are trying to use Tivoli to backup the Oracle databases to the NAS device. The problem we are encountering is that the backups fail with write errors. These appear to be linked to a timeout error received by Tivoli. What has been noticed is the Avg Disk Queue counter peaks when the backup is started, in particular when backing up large files. The volume we are backing up onto is configured as a RAID-5 set.

Has anyone encountered this problem and if so what are the possible solutions / config options which we may try.

Regards

[Chris Watts_1]

I have noticed an issue when using Veritas to do this.

The CPU runs ar 100%, the image name according to Task Manager is system and the owner is system. It doesn't do this until approx 300Mb - 400Mb has been copied.

I suspect it's a RAID sync issue, but have no fix as yet

[Chris Watts_1]

I have found a fix that works for my NAS system.

http://support.microsoft.com/?id=835152

Stopping the NFS service fixed the issue, I now see good throughput of 422Mb + on all backups.

[Clinton Thomson]

Hi Chris,

Thanks for response. I have disabled the Server for NFS as I am not currently using it. This does not seem to have helped in my instance as my throughput has not improved. A work around at the moment has been to split the backup into multiple files..very messy but it alolwed the backup to complete.

[CA1131670]

Try format the volume with larger clustersize than 4kB (default).
4kB is "normal" for clients and filesharing.
A backup-engine typically allocates a BIG file to create a virtual tape in. So we're talking single stream and usually larger I/O's. A volume used for backup purposes, formated with larger clustersize performs much better. Had a customer that was very happy after using 16kB.
Bear in mind though that this unit uses SW-raid 5 and only four diskdrives. You can't get 100's of MB/s !!!!
try formating first, then you could also try using 0+1 instead (will loose some space though).

Reg. Joel, HP Sweden

[CA1125448]

Hi Chris,

Customer need using NFS and during backups CPU runs at 100% even after hotfix Q835152 was applyed. He is trying to use TSM but it hangs ! Taskmanager screens is attached. Any more issue will be welcome !

Ronaldo - HPService - Brazil

[CA1118839]

I agree with Joel. Essentially the 1200s system does not perform that well when it comes to large files (in excess of system memory, which is anything larger that ~500MB). Additionally, Joel points out that backup apps create one BIG file that is the 'virtual' tape. So if you are backing up 10GB of data that started out as 1000s of files, the backup app takes that and makes it one big 10GB file (one, continuous write session). This kills the NAS low-end platform like the 1200s.

Stopping services for NFS is a good idea, but it is not the end all fix. Unfortunately, even if you upped system RAM to 1GB the problem will likely continue. Hope this helps.