Video Screencast Help
Search Video Help Close Back
to help
Not able to make it to Vision this year? Get a sampling in the Best of Vision on Demand group.

Backup Performance Observations & Questions

Updated: 22 May 2010 | 1 comment
JasAld's picture
0 0 Votes
Login to vote
Our environment:
--Backup Exec 10.1 on Windows 2003 R2 Standard IBM server.
--Backup device: Certance (trademark now owned by Quantum) LTO 2 3200 Autoloader (8 slots, single drive).  Tapes hold 200GB native, and we are getting about 300GB hardwarecompressed.
--Library connected to BE server via SCSI LVD cable.  BE server SCSI card is ADAPTED 29160.
--Approx. 9 servers  plus backup server most with RAID 5 getting backed up.  Total 10 servers.  Most servers are Windows defragged nightly.
--Total nightly getting backed up about 505GB using full backup (but not backing up every single thing). 
Hours to completion including verify: about 20.
--Total weekend getting backed up about 702GB using full backup (backing up EVERY single thing).
Hours to completion including verify: about 27.
 
The problem:
Our data has increased to a point where backing up on a nightly basis is taking way too long.  So we have to make some decisions.
 
For some reason we are not getting near full performance on our backups.  For example, the drive is rated "Native Sustained Transfer Rate" :  34MB/s.
 
We really want to try to avoid differentials and certainly incrementals for various reasons.
 
Here are my tests & observations:
--Backing up 11.5GB from Backup server to tape without verify took 10m12s = 20MB/s
(not horrible, but still far from 34MB/s)
--Backing up same 11.5GB from another server across 1Gb Ethernet through 1Gb switch without verify took 21m 24s = 9.7MB/s
(HORRIBLE).
 
1Gb/s is 125MB/s so I don't see how the bottle neck can be with the network.  And no we don't have a ton of bandwidth usage across our network in general (other than backups), to my knowledge.
 
 
Questions:
1. What could the problem be here?
 
2. Although not my preferred solution:
Can a single instance of BE 10.1 run 2 drives backing up data concurrently?
 
3. Again, not my preferred solution, but:
Can BE 10.1 backup data via a SCSI connection to another server? In other words, keep the SCSI cable intact between the current BU server and the tape library, but then daisy-chain another SCSI cable to another server and backup through that interface?
 
THANK YOU.

Comments

Ken Putnam's picture
18
Apr
2007
0 Votes 0
Login to vote

1)  Verify that AutoNegotiate is NOT enabled on all NICs switchports, if possible.  Hardset to whatever is appropriate for your site
 
Is it only one remote server that is slow, or all of them?  How badly fragmented are the drives on the target servers?
 
2)  Certainly, but if the problem is network performance, this probably won't make much difference.  This will increase the RAM requirements, but modern servers almost certainly have enough for this (I ran 4 DLT7000s on two dedicated SCSI channels on a 640MB PII/400 at one point :smileywink:
 
3)  No, BackupExec needs the tape drives/libraries to be either directly connected or on a SAN, which would also require SSO for each media server (Shared Storage Option)

If this response answers your concern, please mark it as a "solution"