25 TB Windows Drive Backup

[erose]

Hello All

We have a physical Windows 2008R2 server with a 25TB hard drive and it is just about full.  I was reading that best practices would be to carve that drive up into separate policies based on directories.  So drive S: has 12 directories and if I understood a forum post that I can not find now, I would create 12 policies for each directory and have multiplex turned on for the tape drives.  

Would that be a fair assestment?  Or is there a better way to back up this data. Our environment is as follow:

Master server = VM running SLES 11SP4 2cpu and 4G of ram

Prod Media server = R710 24 cores and 196G of ram connected to 2 tape libraries via fiber(Scalar i80 and Dell LT4000 for a total of 9 LTO-5 Drives)

Non-Prod Media server = R710 24 cores and 196G of ram connected to 2 libraries via fiber (2 Scalar i80s for a total of 8 LTO-5 drives)

Exchange Media server = R710 24 cores and 196G of ram connected to 1 library via fiber (Scalar i500 total of 2 LTO-5 drives)

The non-prod media server has one of the tape libraries dedicated to this one server that has 25TB+ of data.  We have run into issues where the policy was assigned to the production storage unit group, but would fail once all the tapes in the first library were all used up.  The storage unit group is setup as failover, so as the first library fills the tapes while the policy in question is running, it will fail as there are no more tapes available in the first library.  Instead of it grabbing another tape from the second library and continuing, the policy fails and starts over for drive S:

So I guess I have two questions:

1. How do I setup the policy for this large number of files/data to run more efficiently?

2. How do I prevent the job from starting over from scratch if all the tapes in the library are full when I goes to continue the backup on a fresh scratch tape?

thanks

ed

[Marianne]

1. Create 1 policy with multiple streams enabled. If the option is enabled in policy attributes, then a new stream will be generated for each folder on X-drive in this example: X:\* (12 folders will generate 12 streams) 2. Enable checkpoints in policy attributes. Failed stream will restart from last checkpoint. Best to ensure sufficient scratch tapes are available before the backup starts. i80 tape library has more than enough slots to backup this data and some more a number of times.

[Marianne]

A bit more from my side... Extract from a reply in a similar post:

Without additional licenses to enable FlashBackup or dedupe with Backup Accelerator, you can enable Multistreaming and break down Backup Selection into multiple streams - between 4 and 8 should give good performance without affecting restore speed too much.

It all depends on layout of filesystem - something like:

NEW_STREAM
X:\folder1\a*
X:\folder1\b*
X:\folder1\c*
.....
..
NEW_STREAM
X:\folder1\g*
X:\folder1\h*
X:\folder1\i*
.....
NEW_STREAM
X:\folder1\m*
X:\folder1\n*
X:\folder1\o*
.....
NEW_STREAM
X:\folder1\r*
X:\folder1\s*
X:\folder1\t*
......
etc..

Plus another post: https://www.veritas.com/community/forums/netbackup-multi-steam-backup#comment-6221771

[nbutech]

1. How do I setup the policy for this large number of files/data to run more efficiently?

As stated earlier by Marianne I would first consider for flashbackup if I already have the license for it

2. How do I prevent the job from starting over from scratch if all the tapes in the library are full when I goes to continue the backup on a fresh scratch tape?

Checkpoint needs to be enabled on policy

Quick question

Is this data only expected to be backed up to tape or you also have a dedupe solution in your environment ?