Dedupe, Dedupe, Dedupe.... In the last few weeks, every time I look online, someone is talking about dedupe. New models, customer testimonials or new places to use dedupe. Everyone sees the value in dedupe when looking at backup to disk, but where else can we use it? Some talk about dedupe on primary disk storage, I would love to dedupe hills when riding my bike, I am sure Captain Kirk would like to dedupe the Tribbles and I think we would all love to find a way to dedupe our bills. At the same time, I don't want to dedupe everything - I already don't go on as many dates as I would like. What about deduping backups to tape?
There has been some fun and exciting conversation about dedupe to tape in the last few days. I started reading this latest flourish of conversation at Storage Soup. As someone that's evening plans got deduped, I had plenty of time to poke around online. It all seems to have started with W. Curtis Preston's comments that dedupe to tape has its place. Of course, others jumped in, disagreeing. I have been paying attention to the CommVault feature, and learning what I can about it for a while now. CommVault was my first gig out of the Air Force, so I have a soft spot for them. I also think the make good software, so while I never really thought the idea of dedupe to tape was a good one, I kept an open mind about their implementation.
Should an organization dedupe to tape? It seems that everyone writing about it this week agrees that the answer is no if they want to recover from that copy of the data. That certainly makes sense to me, and is in line with what I have told people for a couple of years now when they have asked what I thought. Curtis' comment makes a lot of sense:
"They recommend it for a very specific user case: the tapes you know you're making that you never plan to restore from"
After 10 years on the vendor side of backup and storage, I have been in a lot of data centers all over the country. My idle curiosity often gets the best of me, and I ask - what's on those tapes on the wall? It became a pretty typical refrain - "We have no idea, they have been here longer than anyone in the shop." We really do have backups we never think we will need to recover; we just store them because we have to. Overall, this speaks to process problems and using backup for something it isn't designed for. (Why don't more organizations actually start archiving?) That is a conversation for a different blog entry.
When I work with an account team and customer to architect a data protection scheme, I always think worst case. Today, when we backup most of data to disk first, recovery from tape most likely means you are having a bad day. 50 tape swaps for a server recovery would make that a very bad day. At that moment no one will feel better that they saved a couple hundred bucks on tapes six months ago.
It seems the ideal application for deduped tape would be:
o Large amount of data – LTO-4 is big; LTO-5 will be huge. If you are not filling lots of these tapes, there does not seem to be any benefit
o Archived for a long time - if your retention period is only a few months, the savings in media won't be that great, as you would get to recycle your tapes.
o Expect very, very few recoveries from tape copy
o Expect no large recoveries from tape copy
o Will never have to bring tapes in from off site location for recovery
So, would I dedupe to tape? I don’t think so, but then I would hope my retention policies would not include long term retention with almost no expectation of recovery. I think some organizations will see great benefits from dedupe to tape, and others will not. If I found myself in that situation would I? Yes. Ultimatly, I think the value in the dedupe to tape conversation with many organizations is it will force them to think about the process that creates all those old copies of data. That kind of review can always lead to better data protection.
Follow me at www.Twitter.com/3pedal.
There are no comments for this entry.
[Add Comment] [Subscribe to Comments]