Backup Stratergy: Suggestions Wanted!

[adamf]

I am after some suggestions on how we can improve our backup stratergy (mainly around the use of BackupExec), we currently do the following:

 

 

MS DPM 2010 - backs up to disk: various SQL databases, DC system states, Exchange mailboxes

 

Symantec BE 2010 R3 - backs up to tape: all staff/student files, shared drives, SQL databases, Exchange Mailboxes, DC system states

 

 

We used to use DPM for quick restores but no longer have available storage to reliably backup student/satff documents, BE is used for daily differentials and end of week/month backups. End of Week backups are kept for 4 weeks before being overwritten, End of month backups are kept indefinetly.

 

 

Our end of week/month backup totals around 3.5TB and is now taking upwards of 60 hours to run (starting at 7pm on a Friday). I suspect a lot of this is due to moving our main fileserver (which stores student/staff documents) over to a Hyper-V VM - the avg job rate when backing up that server is now 900Mb/min as against 2GB+/min when it was a physical server.

 

We cannot use the Hyper-V plugin for BE as when it comes to restoring data you need the size of the VHD in free space to stage the VHD when restoring from tape. The server has 2 VHDs both around the 900GB mark. We tried to restore a small folder and the job took 12hrs, most of which was staging!

 

I was thinking about disk-to-disk but then I have to make sure I have enough storage to get the sort of retention I am after. I could duplicate that off to tape but wouldn't it just take as long?

 

 

Is there anybody baking up similar amounts of data? How do you do it in a reasonable timeframe and make restores as pain free as possible?

[andydis]

any budget?

 

whats the hyper-v HOST hard drive capacity?

[adamf]

Budget: a possibility

 

The hyper-v setup we are using is clustered and we are using CSVs - the particular CSV the fileserver sits on is 3.64TB in size, 47% free

[andydis]

hi adam, sorry for all the questions . could you give more details on your CSV cluster?

 

Your hyper-V Vm's are all connected via 10GB virtual ethernet? ( and speed has been tested?)

[dhicks]

How do you do it in a reasonable timeframe and make restores as pain free as possible?

 

Backing up user/shared files is easy - just use rsync to copy any changed files nightly to a decent-sized backup server and have a script on that server deduplicate the files (or be really swish and get a backup server that supports block-level deduplication). Have the script place each day's backups in a new folder, labled with the date, and mount that folder as a read-only share when each user logs in. The user simply sees another Windows mapped drive that contains versions of their own and shared file areas going back however long you can keep backups for - probably a couple of weeks, at least.

[adamf]

hi adam, sorry for all the questions . could you give more details on your CSV cluster?

 

Your hyper-V Vm's are all connected via 10GB virtual ethernet? ( and speed has been tested?)

 

What sort of details do you want on the CSV?

 

Yes, all VMs are connected via 10GB virtual ethernet.

[glennda]

I use BackupPC for file server data. Backs then up effortlessly and rarely fairls unlike commercial counterparts!