Back-up seems to be one of those topics which can cause huge polarisation in the storage community and especially when we start talking how to back-up.
Can snaps and replication replace the traditional back-up server based architectures; the answer is generally ‘kind of’ but you probably still want to get your data into another system just in case. And in a lot of cases, tape is pretty good enough for what you want to do. But I’m not going to enter that debate in any detail today in this little snippet of an article.
Why do we back-up? We back-up so that we can restore in the event that we need to do; without the necessity to restore, there would be no need to back-up. What kinds of things do we back-up and how often do we restore? Well, apart from the huge amount of data that our data-centre applications seem to generate; we also back-up more and more user-files.
And users are a lot less reliable than most of our applications and randomly delete the wrong file; the more that they create, the more mistakes are going to be made. It is probably this fact more than anything which will drive the move to more on-line back-up solutions and improvements to the interfaces to restore files.
We’ve already put the ability to restore their own files to a small subset of our users; no more calling the help-desk feeling stupid about having deleted the wrong file, they just call up the interface and recall the file.
And that’s the beauty of a zero-tolerance policy to back-up failures; we have enough faith to let our users see what has been backed-up and let them restore. It also means that we can get on and do more interesting things instead!
(BTW, these back-ups go to tape 😉 But that’s a different story.)
My disclosure statement is here; http://infosmackpodcasts.com/about/
Martin, you state:
“Can snaps and replication replace the traditional back-up server based architectures; the answer is generally ‘kind of’ but you probably still want to get your data into another system just in case…”
What do you mean by “another system”? A level of redundancy is backup?
Well, I tend to think of replica pairs as single logical system; I think it is always best to get a regular back-up copy into a different system. This mitigates against something horrible which corrupts both your replicas and snapshots; I like to think of this as a position of last resort.
I think this is a great topic and one I’ve been poking around with people on it for awhile. I’d really like to see a scenario where tape backup is better than snap/CDP.
In your example with a corruption, you would simply roll back to the last good snap before the corruption occurred just as you would roll back to the last tape backup before the corruption occurred. The advantage to the snap is that it’s disk based and recovery is much, much faster.
(Cheap Plug) On our last podcast, I asked FalconStor CEO Jim McNiel if you could do complete backups with just snaps/CDP and he said yes… assuming you were doing replication and had an archive solution in place.
It’s a great topic and like I stated earlier, I’d like to understand the backup scenario that would deem tape a better option than a snap.
Position of last resort; firmware update has just destroyed your snaps. Multiple disk failure has just wiped your snaps. There’s lots of cases where you might want a position of last resort just in case. Your position of last resort might be your archive copy but you do need one.
This is good – I understand your point of redundant backups. For arguments sake:
1) Each of those scenarios is a corruption of the local backup disk targets. Assuming you are replicating daily, your last replicated snap gives you an RPO <24 hours. If you snap/replicate a corrupt snap, only that last snap is corrupt so you should be good restoring from the last good replicated snap. With some solutions, you could have a nearly instantaneous rebuild from the remote site for faster recovery than tape.
2) For a complete data loss like you are describing, an incredible series of events have to take place. It would involve primary data loss combined with loss of local backup disk target and loss of remote DR data. It's always possible, but it is somewhat remote.
3) I think it would be difficult to argue that archive tape is a good position of last resort… but it all depends on how frequently copies are sent to tape. A 1 month RPO vs 1 week RPO.
I'm not saying you are wrong, I'm just trying to challenge the conventional wisdom of tape backup in 2011. Good discussion.
And why do people keep assuming that I’m talking about tape?? It could be bulk disk, it could be etched into granite!
Just get it out of the logical primary disk construct on a regular basis. Incredible events happen…we’ve seen them in our life-times.