Disasters
July 11th, 2005 by Peter Scott
We are currently putting the final stages of a DW migration for one of our customers to bed – they have a small DW (just over half a terabyte) on old and unsupportable equipment – it was good stuff when it is was new, but now…
Like all conscientious developers we insist on a full recovery test prior to go-live. This has two purposes: to prove our recovery process works and to benchmark how long it takes to get the data back. One of my objectives in designing the recovery process was to have something that was reliable, easy to manage and quick to recover. The old system’s backup failed on all three counts. Last year they had a serious problem, made worse by a hardware engineer trashing half of the disk controllers – the only way back was a full restore. The original system was designed to backup quickly, unfortunately the reverse was not true; data was written across multiple tapes, the tapes were often in the wrong drives at the wrong time. To restore from tape took 4 days (or 8 times longer than backing it up) On the new system it took 3.5 hours, a significant improvement!
Another recovery test last week also went well – 2.4 TB restored to a DR system, that took a lot longer, especially when you take it account the time to get the DR system delivered to our data centre and the systems team to load the whole lot from tape, but still it was less than two days – pretty good really, I was pleased and so was the customer
July 12th, 2005 at 5:18 pm
Hello again. I know that tape is pretty much the standard for what to store backups on. But, I’m curious, why not disk drives? They are so cheap now, and hold so much data. And they can be unplugged and carried offsite.
Are they an option you consider in general, but chose not to use in this specific case? Or is it tape all the way?
July 12th, 2005 at 6:12 pm
The larger system uses on-disk backup, this is copied to tape and stored off-site - result: fast backups and quick recoveries. But in this test we were going for the complete rebuild / loss of data centre type of recovery. I do not fancy moving 2.5 TB of disk around - it’s over 40 disks!
I would love to the same with small DW - but if the customer won’t pay for the disk and is happy with hit on recovery time then it’s their call.
But to disk is a great way to back-up, but it still needs to be copied off-site for full protection and tape is still a viable way to do this on a DW.
July 13th, 2005 at 3:39 am
Hi Pete,
How plan for backup recovery (and offsite) requires both options (disk and tape). We use a D2D (disk to disk configuration to support faster MTTR) But only use tape to support offsite requirements to meet D&R plans. (our tape backup is based on the disk backup). The cost of a disk backup solution can actually be lower than that of a comparable tape implementation. Still don’t get why tapes are so expensive.
Though we have not taken advantage of it, theD2D appliances support taking the disks offsite.
But we mainly use it to recover data faster.
I am impressed with your improved speeds for the backup.
July 13th, 2005 at 7:49 am
For the 2.4 TB system we backup the disk copy to tape - so that is much as you describe. For us though, all of the disk sits in the same SAN (20TB - it supports other systems too) so physically moving disk of site is not viable.
Price - maybe there’s more market for disk