| There have been a few threads recently discussing backups and restores.
I thought I'd make a few comments.
Things go wrong. If you depend on any technology or service, you should be aware that it can fail.
You can decide you do not need to cope with failures. This makes sense if the service you're depending on is not very important, or if you know can obtain an alternative elsewhere. Planning to deal with failures takes time, effort, and money, and it may make more sense to spend those resources elsewhere.
But if you do plan not to cope with failures, then remember that decision. It's the worst of all worlds to plan not to cope with failures, then change your mind after a failure has occurred.
If instead you are going to plan to cope with failures, list out which failures may occur, and what you plan to do about them.
The kinds of failures you should think about include:
1) Storage failure; eg failure of a disk
2) Server failure; your server must be replaced
3) Software failure; you discover your software had been creating invalid data for the last month
4) Supplier failure; your data centre operator goes bankrupt.
5) Your own site failure. If you run a business from home, can you keep running the business if your home is broken into and your computers stolen?
And possibly others.
For each failure, you should estimate the likely frequency, decide what is the benefit of recovering, decide on the required recovery time, and work out how to meet that recovery time requirement and how much it will cost. (For example, if data is used by a team of 20 programmers whose work will stop if the data is unavailable for more than three hours, then recovering the data in three hours is worth quite a lot; and every extra hour you take prices $2000 or whatever in lost productivity)
(1) is the failure that most people seem to think about. But it's actually the easiest to deal with. All you have to do is replace the failed storage, and put the data back on it at a known point.
Even so, people still get it wrong. If you have data across several drives, and one fails necessitating a restore from the last backup, you may also need to restore files on other drives to keep them in step with the restored data.
One benefit of duplicating online data with RAID-1, RAID-5, or RAID-6 is that failure of a storage device does not result in data being set back in time.
(2) is harder to deal with. Ideally, you'd be able to restore data to the new server, and have all the software work the same way as it did on the failed server. But that may not work; for example, software licence keys may be tied to the physical hardware.
You need to be sure you can reinstall any software you installed, and choose the same installation options. And if you downloaded the software, the version you downloaded may no longer be available at the download site.
(3) is, I think, the hardest to cope with. Depending on the application, there may actually be no practical way to deal with it, and all you can do is stop it recurring and deal with the data corruption tactically (that is, see what has happened, and use that info to decide what to do.)
For (4), note that backups taken for you by the service supplier might not help. If the service supplier sends backups to a offsite storage facility, that facility may not release them if the service supplier goes out of business. And if the backups are not offsite, they may not be available anyway (eg if there were a fire.)
(5) not only means keep backups of your home data, but also make sure that the thieves cannot disrupt your business using information from the computers they have stolen.
Personally, I think that if you can cope with (4), you can use the same solution for several other failures So it makes sense to devise a solution for (4), see how practical it is, and plan to use it for (1) and (2) if necessary.
Sure, your service provider may be able to restore data to a failed drive for you; but it will take them whatever time it takes, and you may not be at the top of their priority list. (If you are a supplier with 100 customers each needing 1GB of data restored, and one customer needing 100GB, you might leave that customer till last to get the others off your back ... )
Curiously, if you are using a low-cost service such as shared hosting for $10 a month or less, the cheapest solution may be to run a duplicate service with another supplier. It will cost you $10 a month, plus the cost of keeping data in step.
It's worth remembering that. Most backup/restore plans in the past have been based on the premise that it's cheaper to implement backup/restore than to provide a duplicate service.
(Keeping the data in step is the trick of course. But you may prefer to spend effort solving that problem than spend effort on backups/restores.)
Well, that's my two cents worth.
Have at it!
|
Add to Favourites 
Print this Article