Maximizing the job rate

[mufka]

I'm running Backup Exec 12.5 on Storage Server 2008. Using a HP 1760 LTO 4 SCSI tape drive. The stated rate for the tape drive is 576GB/hr. I'd like to get up to half that if possible. Currently I'm getting about 78GB/hr for a local backup. I've tried changing to the HP driver and changing the buffer settings on the device within Backup Exec. Are there any other tricks to getting the rate to increase?

The server is in a test environment and has no load whatsoever (other than the backup). Should I be looking at Backup Exec or hardware? Unfortunately, you can't backup to tape with the Windows backup software so I can't test that.

 

[SteveRoadWarrior]

You could first attempt to identify your bottleneck. If your disk volume is fragmented, busy or not laid out with the right block size (versus the underlying block size and position of the array), you can experience I/O bottlenecks. If this is the case you could try turning on software compression.

Symantec tech support has told me that disabling the on-media catalogs is a way to get better performance, but they don't really recommend that either, for obvious reasons.

Have you run the HP Library & Tape Tools utility? It has a throughput test and also the ability to check firmware updates. Shut down all BUE services first.

It may also be your terminator (needs to be active). Also check the drivers of your SCSI card. You'd better not have that attached to an array controller, even on a spare channel that's not supported.

 

[SteveRoadWarrior]

Any luck with the LT&T benchmarking?

 

[mufka]

I just ran the utility. The device performance tests came up with about 600GB/hr whereas the system performance tests came up with about 110GB/hr. This is a bit confusing because Backup Exec is reporting about 300GB/hr for backups.

It appears that the problem isn't with the tape drive, but with the disk subsystem. It's a HP x1400 with 4 1TB SATA drives in RAID5 on a P212. I've reached half the stated rate, but I'd like to see if I can improve it. I don't know if there's anything I can do to remedy the situation. Suggestions are welcome.

 

[SteveRoadWarrior]

It may be a little late, but if you Gen the server using smart start, it will lay the disk out more correctly than Windows setup. So, this means that 64k sectors on the Array subsystem will match up with 64k sectors in Windows. This makes for more efficient file reads.

You can also switch from Raid5, but that may not be an option. 1+0 is noticably faster.

Enabling Disk Write cache may help, as it might allow spurious write functions to be delayed until the tape system is the bottleneck.

Defragging the server is always a good idea, verifying your MFT is defragmented and less than 80% full. The standard Defrag "Analyze" will tell you some of this in the report.

 

[mufka]

It's a brand new install, still in a test environment. I can reconfigure. I do want some redundancy in the disks (with about 3TB usable), but the rest I can change. It's primary function is NAS so I don't want to do anything that will affect network access to the shares.