Backup to tape speed query

[Frankie]

Hi there

We have a Dell PE2950 with a RAID1 mirror for the OS (Windows Server 2003 x64 SP2) and a RAID5 array (6 x 750GB SATA drives). The server also has two standalone LTO4 drives connected to two SAS cards in the server. This is our backup server running Backup Exec 11d.

On this server we carry out backups to disk on the RAID5 array and then off to tape (the two standalone LTO4 drives). I have noticed recently that when I backup one job to tape with no other backups going on I get 4-5GB/min as the job rate. When I have backups to disk going on as well as backing up to both tape drives I get a lower throughput, maybe 1.5GB/min for each tape drive. When all jobs finish apart from a backup to one tape drive the throughput doesn't get back to 4-5GB/min. It creeps up but will often not get much past 2GB/min. This job is backing up about 580GB of data and is usually backing up the last three quarters of the job with nothing else going on. Can anyone explain why the job rate doesn't jump back up to an optimum throughput?

Cheers

Sean

[Ken_Putnam]

The speed reported is not a snapshot or incremental measurement.  It is total bytes divided by total time, so if it was really slow for a long time, total throughput at the end of the job will be reported as quite slow, even if the last third or so of the job flies

Are the tape drives on separate SCSI channels? 

When the concurrent jobs are running, are they all backing up network data?  perhaps the NIC on the media server is saturated?

[Frankie]

Hi Ken

Thanks for the info re speed reported. The tape drives are on separate SCSI channels, there are two SAS cards installed in the server and a tape drive connected to each one. When the concurrent jobs are running they are backing up network data to disk on the media server and then backing up this disk backup to tape. The backup to tape speed seems to be affected by the backup to disk jobs. I believe it's disk contention rather than network throughput that's slowing things down. At the same time that jobs are being backed up to disk, jobs just completed are reading from the same disk to backup to tape. I was contemplating putting more disks in and creating a second array to see if that would help.

Cheers

Franko

[Ken_Putnam]

Yep, sound like you are thrashing the disk(s)

Don't have any solution except what you have already come up with - More drives