Scheduling wise

[gplus]

Hi Gurus,

We are on a planning stage and wanted to ask help or any recommendation as to how we can set the proper timing for the schedules. I know that setting time in "start window" means its the start of backup window and not the actual start of the backup. We tried to create 5 policies with the same "start window" and it actually all starts at the same time we set. So since there will be a hundreds of policy, we think that scheduling them in this way (sample below) will give us a better performance at load-balance.

Policy1-10 - start of backup will be 7:00PM

Policy 11-20 - start of backup will be 7:15PM

Policy 21-30 - start of backup will be 7:30PM and so on.. 

What we wanted is to set intervals like the example.

Can someone tell us how we can achieve this? Or

Setting the "start window" to this timing is already fine?

Setup:

Master - 1x

Media - 2x

Direction of data will initially be to dedup pool with max of 15 concurrent jobs each media.

Regards,

Gino

[sdo · Accepted solution]

The fact is some poeple worry about this type of thing, and some do not.  With thousands of backup clients across hundreds of backup policies, then your actual active job count will be dependent upon the max concurrent write streams of the storage units that are being written to.  My advice would be to monitor your system and then decide.  i.e. if the CPU of your master goes crazy at 19:00 when thousands of backup jobs get queued up, then you know it's time to split the scheduling.

[gplus]

what were looking at is actually minimizing the rework that is why we're torn between scheduling the backups with the same 'start window' or split the schedules but i guess its inevitable. But you have a good point. We'll never know its (master) capacity until we see the outcome of what we've done.

But i'm pretty sure there will be spike in the CPU of our Master if we set all schedule at once. 

Thanks for the advice @sdo.. we'll try that approach. Cheers!! :)

[Nicolai]

With a dedupe pool  of max of 15 concurrent , i doubt the load will be high since the dedupe pool set the limit.

I would activate all policies at the same time and let Netbackup run the job by its own intern algorithm. It is actual very good at it.

Consider if yore design is right with "100 of policies". Can you really manage that many, and when Veritas publish a tech note describing data loss under certain circumstance. Can you figure our impact with with e.g 300 policies ? Just a heads-up ..

[gplus]

For managing alone i think we can, as they will have a dedicated support who will monitor it moving forward. 

For policies, they have its own general attributes and standards which is documented; per application, per server wise - so incase if veritas release a new technote atleast we have a baseline to start with. 

Thank you nicolai for the heads up. I will include this as part of the discussion.