NetBackup client Buffer_Size KB value and Media Server NET_BUFFER_SZ byte value for optimal performance

[yobole]

We have a very large file server with 4 volumes and millions of small files we are struggling to backup adequately. The client is a windows server 2012 server virtual machine running Netbackup 7.6.0.3 the largest is 6 TB while all the other are about 1.5 TB.

The media server which has an LTO 5 tape drives attached also has the Dedupe pool running windows 2008R2 Netbackup 7.6.0.3

We have 1 policy to back up the 3 X 1.5TB to a Dedupe pool which takes about 8 – 15hrs. We now have another policy to backup just the largest 6TB drive directly to tape once a week.

For each of the volumes we have added an extra volume and configured VSS snapshots to use this for Windows Open File Backup

However the speed we are getting are quite shocking its taking about 7 days to back up the 6TB and this tend to fails as it’s difficult to get 7 days uninterrupted service on the server which is a very busy file server

Looking into more details at the performance and buffer seize discovered this article

http://www.symantec.com/business/support/index?page=content&id=TECH19206

NetBackup client Buffer_Size KB value should match the Media Server NET_BUFFER_SZ byte value for optimal performance

However at the Moment the client Buffer_Size KB is 128KB and the Media Server NET_BUFFER_SZ  is 256KB

However then I find this article which contradicts the above saying they should just be 0

http://www.symantec.com/business/support/index?page=content&id=TECH28339

Any ideas on what the buffer setting should be?? Or any advice of backing up the server in an appropriate time frame??

Kind Regards

P/S we have multiple stream, accelerator enabled. Cannot use change journal as the sever is also an Enterprise vault file server so the reparse point prevents the change journal from working 

[Nicolai]

0 should be used - the last tech note (Updated: 2014-05-08) is the most current and explain why NET_BUFFER_SZ does not apply anymore for Netbackup.

Hope this clarify

Best Regards

Nicolai

[Marianne]

You need to look elsewhere to increase performance.

Slow read speed from disk due to fragmentation and NTFS battling to navigate filesystem tree with millions of files and large directory depth are normally the culprits.

No much you can do about this for filesystem backup.

One options you may want to look at, is FlashBackup (where raw volume is backed up with ability to restore individual files/folders) or else simply break the Backup Selection up into multiple streams.
I have seen how a customer's file server that took more than 48 hours to run full backup of one large volume completed in 8 hours by breaking up the Backup Selection.
(Only do this before next full backup is due.)
Ensure Max Jobs per Client is set to 4 or more and 'Allow multiple data streams' is selected in policy attributes along with MPX enabled in schedule and Storage Unit properties.

A good understanding of filesystem/folder layout is needed. Example:

NEW_STREAM
Z:\folder\a*
Z:\folder\b*
Z:\folder\c*
.....
Z:\folder\g*
NEW_STREAM
Z:\folder\h*
Z:\folder\i*
...
...
NEW_STREAM
Z:\folder\x*