-3

I have been thinking about how large companies backup their data. I imagine that in some critical enterprises they must consider every possibility and can't afford to lose even one bit of data. Of course they use backups.

But what if the backup fails? The whole thing, not just one or two disk (because they're using RAID of course). No problem, just hurry up and copy the original data to a shiny new disk. But what happens if both systems failed. I don't know maybe zombies bit through the power lines and trashed both data centers, could happen. How do you prevent data loss?

Well, you just have a backup of a backup of a backup of a backup...

Does this problem occur in real life (sans zombies) and how do the pros deal with this conundrum?

noel
  • 103
  • 4
  • 2
    `how do the pros deal with this conundrum` - Just add more redundancy, multiple backup systems, stored with multiple methods. As much as the powers-that-be choose to pay for. – Zoredache Nov 04 '14 at 22:14
  • Yeah i see that, but that's kind of the paradox isn't it? There's only so many bits in the universe. – noel Nov 04 '14 at 22:16
  • 1
    How is that a paradox? You buy as much insurance (backup) as you can afford, or as much as the data is worth, or to the point where you get diminishing returns. There is no way to be 100% certain in backups, or anything else in our reality. – Zoredache Nov 04 '14 at 22:18
  • 5
    You keep using that word. I don't think it means what you think it means – mfinni Nov 04 '14 at 22:20
  • Well infinite redundancy seems kind of self-condradictory in a finite world. – noel Nov 04 '14 at 22:20
  • 1
    There's no such thing as infinite or guaranteed or perfect redundancy. All anyone can do is manage risk. You can spend more money and be smarter and reduce the risk to an acceptable likelihood. – mfinni Nov 04 '14 at 22:21
  • 1
    Nobody gets 100% uptime, absolutely nobody. 5 digits is about the best that can be planned for, and that costs astronomical amounts. I think your premise is flawed, everyone can tolerate downtime and dataloss, it's just a matter of what it's worth and the likelihood of such an event. – Chris S Nov 04 '14 at 22:22
  • Enterprise Backup Solutions (e.g. Commvault Simpana) have data integrity validation (e.g. CRC checks on source data blocks). Regardless this is why the maxim exists to always check your backups. It's about reducing your risk to a price point that your organization can afford. – HTTP500 Nov 04 '14 at 22:26
  • 1
    Its easy to get 5 9s uptime...as long as you define it as 9.9999%. – Grant Nov 04 '14 at 22:40

1 Answers1

2

This isn't a paradox.

There's no such thing as infinite or guaranteed or perfect redundancy. All anyone can do is manage risk. You can spend more money and be smarter and reduce the risk to an acceptable likelihood.

mfinni
  • 35,711
  • 3
  • 50
  • 86