Version 10 slowed down

[teqmod]

I am running version 10.0 rev5484 and around a month ago the back-up speed has slowed to half of what it once was. Nothing has changed on the network and it is just file servers. The primary server is a NetApp with 1.5TB of images. We were getting approx 450mb/s and it has dropped to approx 225mb/s. All other network machines have slowed as well. There have been no changes on the network or to the server. I searched the site and saw lots of info but nothing that can help me yet. The server has 2 functions, it is a DC for the remote site and it does back-ups so the load is not high. I have checked the SCSI drivers and it has the latest. We are using an Overland powerloader with 2 HP tape drives. Originally we only had one drive, the problem started occurring while we had the second on order. The second drive has been added since we do have alot of data and I have added the license. Since the problem existed before the drive was added that should not be the issue. I attempted to go to Veritas' site to locate the latest drivers but came up with nothing. The machine is fairly old and it has an adaptec AHA-2940U SCSI adapter in it. Any ideas?

 

[penguinroundup]

what kind of files are you backing up? I have seen this issue happen (with version 9, it was a couple of years ago). My backups would slow to a crawl when backing up a directory full of *.zip files... Veritas tech support said it was because BE had to open each zip file to back it up (which doesn't make any sense to me)... we would up going into the SCSI bios and tweaking the settings in there to get an acceptable rate. It wasn't great, the backup still bogged down on those files, but it wasn't as bad as it had been.

 

[teqmod]

Mostly .tif image files which aren't even compressed. This problem is present on all back-ups but since the file server is over 1.5TB it is now taking in excess of 70hrs to back it up. I think if I cannot resolve the problem quickly I may put GB NIC in each nad run a direct connection between the 2 to rule out any network problem.

 

[steveot]

Try using Backup Exec and create a Backup to disk folder localling on the media server. Backup about 1 gig of remote data to disk.

Then Run NTbackup and backup the same data to disk.

Compare the rates..this will rule out the tape device and point you to a network or Veritas problem.

NTBACKUp you will have to do some calculations to get the mb/min rate but there are free online calculators out there.

 

[c0r0]

Just a thought, I have a similar problem. A 4 hour backup-to-disk of a remote server was taking 18 hours, Job is set to Overwrite.

Check the network activity in task manager ( or with perfmon ) Mine kept dropping down to zero for several minutes then picking up to 100% ( where I would expect it to be due to the concurrent backups I was running ). I found that if I deleted the backup files in the backup to disk folders and ran a fresh backup it would be back to normal. I think its a disk bottle neck maybe caused by too much disk fragmentation, which is where I'm looking now.

 

[Zelandakh]

The Adaptec 2940U is rated at 20MB/s. So getting 450mb/s is nearly 3 times the maximum speed of the card - not sure how you are managing that.

I'd say the card is the bottleneck. If its the Overland I think it is, try an Adaptec 29160 and see if the server backplane is also stifling you - PCI-X would be a result if you have one.

 

[teqmod]

Thanks for the info Zelandakh, the job speed has seemed to return to normal for all machines except the NetApp. I went through Veritas support for a while and ended up speaking with NetApp on the issue. The have made several recommendations since we are getting some irregularities in the diagnostic from the NetApp. I am going to be pulling the NetApp down this weekend and making some changes. I will post how it goes.