Enhancement request for NetBackup: better use of bandwith
Updated: 17 May 2010 | 5 comments
Status:
Reviewed
We use NetBackup 6.5.4 on Solaris master- and mediaservers. Every sunday we start 12 fullbackups (LAN), using the same mediaserver. These backups use our NIC for nearly 100percent. On Monday most of these backups are finished, 2-3 are still running but they only have a transfer speed of 10 Mb/sec (all Server and clients have Gigabit). Each of this clients are able to push 40/50 Meg. over LAN (tested). It seems, NetBackup can't speed up the running backups even if there is enough bandwith! As a workaround we suspend the 'running' backups and resume them, they continue much faster then before (25/30 Mb/sec).
It would be great to have NetBackup use automatically the full bandwith.
Idea Filed Under:
Comments
As a matter of interest
how are you measuring the throughput? Just through the Activity Monitor?
In my experience, the quoted values in the NetBackup Activity Monitor are averages based on the length of time the job has been running. So, if the jobs have been slowed down for any length of time and any reason (e.g. other backups, resource availability, processes or network traffic etc) then their average throughput will be effectively slower overall compared to if they were running 'solo'.
Or are you monitoring throughput directly & real-time & can confirm they are actually running slower?
Just a thought.
Regards Andy
"It's not too late to panic ..."
Re: As a matter of interest
We use a tool 'nicstat" to check the stats:
Time Int rKB/s wKB/s rPk/s wPk/s rAvs wAvs %Util Sat
09:59:08 nxge0 0.03 276.1 0.49 3720.0 60.00 76.00 0.02 0.00
09:59:08 nxge1 10922.0 0.00 7450.3 0.00 1501.2 0.00 0.89 0.0
For system monitoring we use Openview from HP (on netbackup-server and clients)
The point is: on monday morning only a few backups are running and they don't use the full bandwith. With 'suspend and resume' they run much faster.
NetBackup 7.7 on Solaris10
writing to TS1120/30 drives in IBM3494
That's fine.
Not what you are seeing, but how you are checking.
Only other thought, when you suspend & resume do the jobs start utilising a different tape drive? I know this isn't why you posted this (it is an 'idea' afterall) but was considering whether there maybe another reason why this is occurring.
Regards Andy
"It's not too late to panic ..."
Performance and Tuning Guide
There is a lot of information available on this topic. Start with the Backup Planning, Performance and Tuning Guide - http://seer.entsupport.symantec.com/docs/281842.htm
In general tuning the buffers used is often used as a step in making sure you are using resources well. It's not the only step, but once you understand the bottleneck it is used to ensure NetBackup is using resources well.
I have evidence from customers of NetBackup moving data at near the speed of the available data path, so this isn't a pervasive product problem.
As an Idea I can say with some confidence that the product already does what you asked for. In your case we need to do some tuning or engage support to take a deeper look at your environment. I've marked the Idea as reviewed to reflect this status.
Technical Product Manger Symantec
Client-Side Deduplication
Although I'm not offering this as a solution to your bandwidth concerns, one thing you may want to consider in long-term planning is deduplication at the client level. It's available now in PureDisk and available in NetBackup 7.0 as the client-side deduplication option. It greatly reduces the amount of data transferred and stored by orders of magnitude.
Todd D. Woodward
Sr. Technical Support Engineer
Symantec Corporation
www.symantec.com
Springfield, Oregon
Would you like to reply?
Login or Register to post your comment.