Backup Plan

NiCkNaMe

NiCkNaMe

NiCkNaMe

Soldato
Joined
17 Jul 2005
Posts
3,192
Hello All,

I'm looking at re-jigging our backups somewhat as with various upgrades they've become outdated and do nothing for us IMO. It's mainly the exchange backups I want to look at.

We've got;

  • 2 x Ultrium LTO3 tape drives
  • 1 x Multi tape loader (Quantum - 16 bays) with an LTO 4 drive
  • Backup Exec 12.5

We have various other systems backing up that I don't want to adjust for the moment. With the above i'm specifically looking at backing up our file server and exchange 2007 enviroment. Currently the multi loader is not used and we backup a SG each evening (to keep to one tape), with one SG backed up on friday on a four week rotation. Crap!

The current total Exchange IS hits approx. 850gb (200ish per Storage Group). The file server hits about 500. I'll continue backing the file server up to one of the LTO3 ultrium drives... it sits on one tape at the moment anyway.

What would you reccomend to get the most out of the exchange backups? Some issues i'm having at the moment (away from the obvious lack of being unable to restore any recent data due to backing up a SG each evening) are that currently snapshot backups are being used (apparantly should not be used with databases, e.g. exchange?) and that restores from tape take an absolute age.

People have mentioned duplicating the backup to disk... how would you do this? I have approximately 1tb available via a SAN that could be used for this.. but would I just duplicate each evening? What's the point? Should I be backing the catalogues up seperately to disk? How can I speed them up (without buying new super fast tape drives ;))

My overall plan is four tapes on a two week rotation with one tape on a 12 week rotation over the 5 working days.

Any hints and tips would be much appreciated.#

*This will leave us with a spare LTO3 drive that I was going to use for restores...
 
In backup exec, you can add Backup-to-Disk folders, and then you create a job to duplicate the backup to disk job to tape.

We backup all of our servers to a SAN then stream it off to tape, meaning we can backup all of our boxes in an evening, and we have 4 LTO3s a night just for our NT servers

In answer to your other question about restores, we do two kinds of Exchange, we do a full server backup to tape. but we also back up the information store to disk only, this allows us to recover emails, folders etc, without having to get a tape back from our tape store, and makes things a bit quicker.

Kimbie
 
Nikumba said:
In backup exec, you can add Backup-to-Disk folders, and then you create a job to duplicate the backup to disk job to tape.

We backup all of our servers to a SAN then stream it off to tape, meaning we can backup all of our boxes in an evening, and we have 4 LTO3s a night just for our NT servers

In answer to your other question about restores, we do two kinds of Exchange, we do a full server backup to tape. but we also back up the information store to disk only, this allows us to recover emails, folders etc, without having to get a tape back from our tape store, and makes things a bit quicker.

Kimbie
Click to expand...

I've been the B2D folders... I haven't ran a full backup to one of them as opposed to tape yet but I know they've been used previously here and were canned due to speed problems? What sort of speeds do you see backing up to disk?

And when you back up the IS to disk... do you do this each evening? (e.g. writing over it each night?)... 800gb IS uses a fair whack of disk space :o

One more question... when you stream the job from B2D to tape... is this faster than running a direct backup job to tape?

Thanks
 
NiCkNaMe said:
I've been the B2D folders... I haven't ran a full backup to one of them as opposed to tape yet but I know they've been used previously here and were canned due to speed problems? What sort of speeds do you see backing up to disk?

And when you back up the IS to disk... do you do this each evening? (e.g. writing over it each night?)... 800gb IS uses a fair whack of disk space :o

One more question... when you stream the job from B2D to tape... is this faster than running a direct backup job to tape?

Thanks
Click to expand...

The speed of our B2D folders ranges from 200mb/min for our older servers up to about 1000mb a min for the newer ones.

We only back up one of our IS and we keep that for two days as that is the retention on the B2D folder

Streaming to tape we often get 1500mb/min easy, since its a large file its streaming it can get up a decent speed apposed to streaming lots of files to tape

Kimbie
 
another thing to bear in mind is with exchange IS backups Backupexec (i'm basing this on 12 not 12.5) has to stage the whole thing before it can back it up. If you're just copying the B2D file then this process isn't necessary, it can just copy it to the tape which is a lot faster and more efficient than doing the tape backup from the IS directly. It also means you can do it whenever and however you want because backing up the backup won't affect performance of the live system at all.
 
Off hand, does anyone know if it's possible to share a backup tape device connected to one exec server, to another exec server?

Thanks,
 
NiCkNaMe said:
Off hand, does anyone know if it's possible to share a backup tape device connected to one exec server, to another exec server?

Thanks,
Click to expand...

Usually you would configure the jobs to run on the server with the tape drive attached and install remote agent on the other servers to allow it to back them up over the network.
Though I suspect there are many ways to skin a cat that's probably most cost effective as I don't think remote agent licenses cost as much as full monty Backup exec licenses.
 
Thanks guys.

I did not know you could run a B2D and a backup to tape drive simultanously on the same backup exec server...

I've now got the entire exchange IS backing up to disk (900gb) and at the same time another job running that backs up various servers to tape during the same time frame (evening/overnight). Once both are complete (morning) and the tape library becomes free I then duplicate the exchange IS B2D to tape.

Now this i'm happy with, however restores still seem to be a big issue.

I can obviously restore from the B2D no problem.. matter of seconds to retrieve mail from the backup. However i'm only keeping the B2D folder for 1 day before the next kicks off as I only have 1TB on the SAN attached to that backup server.

When I did a test restore from the duplicated tape it was still running at 200 MB/min (compared with 1,500 MB/min backup)... when staging the entire storage group (2-250GB) this takes about 30 hours... crap when you just want to restore a couple of emails :-(

I'm running through some tests, e.g. updating all drivers, installing all hotfixes etc etc... but the performance is still dire.

Any ideas, workarounds or suggestions fellas?
 
12.5 has GRT. See how it works for you?

I have had limited success but basically granular restores of Mailboxes/Individual mails from a 'regular' IS backup with GRT enabled on the job.

It was buggy, but it may of been worked on since I played with it.

Can you not configure Deleted Item retention to suit your restore needs?

Brick level restores are god awfully slow. I have a BackupExec 10d environment that can backup at 1.7gig a minute and I have had 36 MEGABYTES a minute throughput on the restore. :/
 
Last edited:
GRT is enabled on the original B2D job and hence enabled on the duplicated tape backup, i'd presume? - Restore still attempts to stage the data when restoring from tape, not direct from the B2D however.

From what i've read even if you use GRT with a tape backup you still have to stage the storage group when restoring?

I've managed to increase the restore speed by an additional 25 MB/Min... it's now steady at 230 MB/min after upgrading drivers on both the SAS card and device itself - however this is still shocking.

I know read speeds are often slower... but a massive 70%+ decrease is crazy in my opinon... I just cannot see how you could justify staging an entire storage group for a single email...

I'll be configuring deleted item retention - but generally we seem to be getting requested to restore items from 8-12 months previous (queries from other companies on what was said, etc)

Thanks for the reply mate.
 
Yeah, damn sales section of the business causing grief :p Nightmare.

Archiving has been on my mind - we had enterprise vaults I believe at my old firm however it caused no end of problems... although it was perhaps mis-managed.
 
If people asked me to restore something they deleted 8 months ago I'd tell them tough titties. But if your policy states you have to be able to then I think you need to be talking archiving. I wouldn't suggest archiving the whole thing each time, but certainly creating a master copy then taking snapshots and archiving those somewhere.

Backup exec is more Disaster recovery orientated than for the recovery of deleted items. Live snapshots are more appropriate for that. This is especially easy if using a SAN to store the data.
 
Look at a Grandfather policy set up on the backups.

We set off VrangerPro to backup all our VM's at 1900, then a bit later, Backup Exec starts a B2D job, then after that the library backs it all up to tape.

Monday starts with a Full Backup, then differentials each day.
Dailys are on a 2 week cycle,
Weekly's run on a 4 week cycle
Every 4 weeks, a Weekly set is removed and kept.

in theory we can restore a file from a point within the last two weeks at a day, then at a week, then at a month to 12 months ago.
 
You must log in or register to reply here.
Back
Top Bottom