Is it worth converting from Netbackup from Windows to Linux?

[John_Deackman]

I read or heard somewhere that larger Netbackup systems run much more efficiently on a Linux system than a Windows system.  Our shop is primarily a Windows shop with just a few Linux machines. 

I am currently up to 170 policies  with close to 200 servers writing a DataDomain system and our data is growing.  Is it worth the headache to convert over to a linux based NetBackup environment or do I just keep pushing hardware at Windows for growth?

Just looking for opinions and shop stories or those who have done this.

[mph999]

Personally I would go for Linux over windows, but then again I would go for Solaris over Linux.

Part of that is based on the fact I have a Unix background in a previous life, which in no way can be used to justify a decision for someone elses environment.

Designed properly and left to run without being fiddled with, NBU should run well on any platform.  I 'personally' only run NBU on test systems and I therefore swap and change, as you will image.

Looking at customers who run large environments (big enough to make your eyes water)  the vast majority if not all, are either Linux or Unix (usually Solaris).  I don't see many large windows systems, but that does not mean they are not out there, simply the majority are unixy ...

Then again, it depends what you call 'big'.

My persoanl view, based on real life experience with differnt enterprise backup products, is that it is very sensible to split backups across multiple masters, and have x hours of idle time on each system, there is a very very good reason for this.

Backup servers go wrong - like it or not, it happens.  If this is unacceptable to someone, they are in the wrong job.

With idle time, you have the ability to rerun jobs, so sure, you may get a failure, but will have time to re-run and therefore meet your SLA.

When things break, you have time to fix, or at least a little flexability.

We do see single master site, where the server is busy pretty much 24/7 - and these do work, but when it breaks things get bad real quick ...  and it's especially bad if the system is big, why, log size - this can present a combination of issues.

It's slightly harder to shift through a big logs than a little ones,  but it also takes longer to get the logs to us.  These are probably only fairly minor points, but that's how it is and it all adds up.  The biggest issue, however, is that on bigger systems it can be preety much impossible to turn the logs verbosity up, either for disk space reasons (and therefore incorrect design...) or the fact that a restart could be required.  Quite how we are expected to investigate/ fix an issue when the customer refuses to send in the logs I am not too sure, but occassionally they do.

Another big advantage with separate backup environments is that you only lose part of the backups in the event of an outage.  If the design of the whole system is really clever, you may simply be able to redirect any really important backups to another server, which of course, as it has some idle time, will be able to cope with the extra load.

So in summary if you go for separate environments you will certainly be fine, and will gain what you may count of the advantages I've suggested above.  If you want to go for a single environment, based on the fact that the majority of large systems I see are non-windows you may wish to further investigate the Linux install.

Note, you can't just catalog restore from Win > Linux, it won't work.  It can be done, but only via a Symantec partner.

Hope that this helps,

Martin

[Stavros41 · Accepted solution]

I agree with a lot of what Martin has said but I would say that given the size of your environment and the current make up ie it being Windows mostly, it would make sense to stick with windows unless it has been giving major issues.

Performance and reliability was always held up in the old days as a reason to choose Unix (Normally Solaris) over any other platform for Master/Media servers. You could get significantly more performance out of this combination than Windows was able to deliver. This was of course true for far more than just backup.

The performance argument and to a lesser degree the reliability argument has become far less relevant over time as the new versions of Windows became more performant and reliable. The same could be said for Linux as it has matured it certainly wasn't cutting against Unix when it first started being looked at in comparison to establish unix ie Solaris, HP-UX, AIX etc

In a recent conversation with Symantec management I asked a similar question about the deployment breakdown over NetBackup Master servers specifically and any trends in deployments. The information I got back was that over 80% of all new deployments of NetBackup are on Windows Master servers.

The other things that spring to mind on this issue are

1) How is the current setup performing? Has any optimisation been done to ensure the best performance?

2) If you went through the aggravation of moving over, do you have the staff to optimise and get the best out of the flavour of Linux you deploy? How big an improvement would make this worthwhile?

3) How much of the envrionment is virtualised and do you use VMware?

A recent deployment worked on was much larger than your size in clients but used data domain and has been implemented on Windows Master/Media servers without issue and the performance levels have been very good utilizing DD Boost.

Hope that helps