Backup slow on one of two virtual drives located on the same SAN device
Updated: 30 Sep 2010 | 4 comments
We have a Backup Exec 12.5 tape backup server with 2 GB of ram. Around 86% of the ram is being used when the server is setting idle.
We are backing up a server with 2 virtual drives that are located on an AX100 storage array (SAN). Virtual drive E: is 1TB in size and virtual drive F: is 200GB.
When backing up the E: drive we get the following statistics:
Backed up 6,706,972 files in 2,857,129 directories.
Processed 220,444,794,362 bytes in 46 hours, 8 minutes, and 11 seconds.
Throughput rate: 75.9 MB/min
When backing up the F: drive we get the following statistics:
Backed up 170162 files in 1088 directories.
Processed 153,798,058,689 bytes in 3 hours, 21 minutes, and 55 seconds.
Throughput rate: 726 MB/min
Backed up 6,706,972 files in 2,857,129 directories.
Processed 220,444,794,362 bytes in 46 hours, 8 minutes, and 11 seconds.
Throughput rate: 75.9 MB/min
When backing up the F: drive we get the following statistics:
Backed up 170162 files in 1088 directories.
Processed 153,798,058,689 bytes in 3 hours, 21 minutes, and 55 seconds.
Throughput rate: 726 MB/min
Both virtual drives E: and F: are on the same AX100, so it seems like the backup speed should be the same. However, as you can see, the F: drive backs up nearly 10 times as fast as the E: drive. One big difference is that the E: drive has a much larger number of files and directories than the F: drive. I know more directories will slow it down some, but I wouldn't think it would be 10 times slower.
The buffer size for the tape backup device is set at 64KB. And the number of buffers is 10. I presume we should increase the buffer size, but I am not sure if we should raise it to the limit of 1MB.
The buffer size for the tape backup device is set at 64KB. And the number of buffers is 10. I presume we should increase the buffer size, but I am not sure if we should raise it to the limit of 1MB.
Any suggestions on what we can do to increase the backup speed of the E: drive would be appreciated.
discussion Filed Under:
Comments
A larger number of small
A larger number of small files that total the same data size as fewer large files will always backup slower due to the file system overhead to access the files. There is nothing you can do about this.
I do agree that a 10x differential may be on the high side Is anything else accessing the E: drive during the backup? what is the relative fragmentation of the two drives?
If this response answers your concern, please mark it as a "solution"
The E: drive contains images
The E: drive contains images that are not accessed during the weekend when the backup takes place. There was very little fragmentation on either drive.
Hi Aaron, Do you run backups
Hi Aaron,
Do you run backups off the AX100 at different times, or together?
My suggestion would be to check in Navisphere and see whether or not both LUNS are being accessed through 1 controller. If they are, try manage 1 of the LUNs through the other...ie. Drive E through SP B, Drive F through SP A. See if some sort of load-balancing there would help matters.
Other than that, Ken is quite right...taking a look at the number of files and folders shows you have a couple million each. Only other way around this here would be to look into either leaving off data that isn't needed, or cleaning up the drive if possible.
Laters!
If you find this is a solution, please mark it as such.
Hi Aaron, Did you come right
Hi Aaron,
Did you come right here...?
If you find this is a solution, please mark it as such.
Would you like to reply?
Login or Register to post your comment.