Veritas-bu

Re: [Veritas-bu] Sizing for NBU Master (a big one)

2013-04-12 12:05:11
Subject: Re: [Veritas-bu] Sizing for NBU Master (a big one)
From: Mikhail Nikitin <mikhail.nikitin AT gmail DOT com>
To: Rusty Major <rusty.major AT sungard DOT com>
Date: Sat, 13 Apr 2013 00:05:05 +0800

Hi Rusty,

Unfortunately it is not that easy :)
We already have multiple environments and consolidating ( all based on Linux x64 ). However, as you rightly pointed out to the NBU planning and performance tuning guide, sizing beyond 10k jobs is not documented :)

It's perfectly fine to add more resources to the master server, we just need to know what are those resources and how much to add, that is the question.

The environment is indeed based on two sites in HA cluster (VCS), but I doubt that it is possible to split them atm (there are a few data domains that are not supporting AIR quite yet. I heard it is coming in DD OS 5.3 and plugin 2.6 but they are not GA yet)

I have got a handful of responses from SYMC off-list but nobody is willing to commit on sizing figures :)

Thanks anyway, will share what I'll end up with.

M.

On Apr 12, 2013 10:05 PM, "Rusty Major" <rusty.major AT sungard DOT com> wrote:
Mikhail,

Theoretically, you can only have 86,400 jobs per day due to the limit of having the ctime at the end of a backupid unique for each backup image. As you probably know, the ctime is based on the seconds in a day, and there are only 86,400 of those.

This article discusses that, as well as mentioning a limit of 10,000 jobs per 12 hour period, so 20,000 jobs/day. www.symantec.com/business/support/index?page=content&id=HOWTO56073

Do not forget to include any duplications, replications, imports (AIR), image cleanups, or other jobs in your calculations as these can be significant numbers as well. I think you should consider splitting this up into 4-6 environments running 10K to 15K jobs each. If these are across different sites, putting a master in each site would be a logical way to separate those out.

I would recommend you at least double what Symantec's resources state, just because servers are so cheap. I would also recommend running it on a Solaris SPARC or Linux x64bit server and dedicate that server to being the Master only. You also should consider how to get that catalog backed up as quickly as possible as I'm sure a catalog for that many jobs is going to be over a TB in size.

If this environment is seriously going to be that size, you need to get with a Symantec Account Rep (who will probably love you) and have them engage some internal resources to help you plan this out. 60K jobs/day will be among the largest customers and you should get the attention that deserves.

Good luck with your project!
-Rusty

On Fri, Apr 12, 2013 at 4:31 AM, Mikhail Nikitin <mikhail.nikitin AT gmail DOT com> wrote:
Hi guys,

I'm curious if anyone here had a chance to get hold of knowledgeable chap in Symantec who knows what kind of hardware I need to run around 60k jobs per day.
The NBU Planning & Performance Tuning Guide does not go beyond 10k jobs per day, that is 8CPU and 16GB of RAM. 

Many thanks in advance.

Best regards, Mikhail

_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu


_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
<Prev in Thread] Current Thread [Next in Thread>