How to increase bandwidth utilization in SLP

[liuyang]

Hi, my master server and media servers are all NBU 7.0.1 on Windows 2008 R2 Enterprise. I have one media dedupe server in HQ and another one at DR site. I use SLP to duplicate some critical backup data from media dedupe server in HQ to media dedupe server at DR site. The two sites are connected by 2MB WAN link.

Every day I have about 10 duplication jobs to run from morning to afternoon. I found each job will occupy some bandwidth of the WAN link. So in the morning when all the jobs are running, the bandwidth of the WAN link is nearly fully utilized. But in the afternoon, when some duplicaiton jobs complete, the bandwidth utilization drops. The remaning jobs seems still occupy the same amount of bandwidth as in the morning.

For example, each duplication job uses 200KB bandwidth in the morning. In the afternoon, maybe only 5 duplication jobs are still running. However, each of these jobs still uses 200KB. I thought it should increase to 400KB so the bandwidth will be fully occupied

My question is: are there any settings that I can ajust to increase the duplication performance? How to increase bandwidth uitlization in duplication jobs in SLP? Thanks in advance.

[RiaanBadenhorst]

Hi,

There is no setting, it uses what is available. One thing you might consider is that the source or target is the factor slowing down the speed i.e. its not the network that is a bottle neck but rather the source / target processing the backup image/information. The same concept really as having slow clients or slow media server when referring to a backup.

There are other parameters you can modify, take a look at this note. http://www.symantec.com/docs/HOWTO33715

Maybe you'll get better results if you modify the size of the duplication batches.

[thesanman]

I have been looking at something similar and a lot of the work is being at the source end; working out what blocks do and don't have to be sent across the wire.

I doubled my memory; that made a difference.  In my case my media server was showing 80% memory used and all went slow.  Also look to your IO subsystem; what's the queue length like?

[Rajesh_s1]

One thing i would like to add here is , once you send the full backup of an perticular client from the next time only the changes will be replicated to another site . Suppose if you are getting a dedupe of 90 % then only the 10 % changes needs to replicate to offsite .

If you have configured OpsCentre then you can check the actually how much data is getting transfered. If you check the 'Pre vs. Post Deduplication Size' report on the OpsCentre you will come to know exactly how much data you are transfering .

Comming to your point on the replication, what ever the network bandwidth you have it will completey occupy if your producation site can able to push that much data.