Reply to post: Re: Optional DR/Resiliency

Tuesday's AWS S3-izure exposes Amazon-sized internet bottleneck

Lee D Silver badge

Re: Optional DR/Resiliency

The number of times that I've had to explain this:

If you want a backup system, it will cost you what the real system cost, again, and a bit more for whatever tech to make it fall over.

And, yes, that functionality, hardware, processing power, storage, etc. will NOT be available to you to use. It will literally be idle (from a user point of view, but hopefully replicating etc.!) most of the time.

If you want something that tolerates a failure, you have to buy two of them and one of them does nothing all day long but wears, depreciates and costs just as much as the first. If not, it's not a suitable replacement.

And then you get into the depth you take this to - a redundant disk is just another disk. A redundant array is just another array. A redundant server is just another server. But a redundant site is another site. A redundant datacentre is another, fully-funded, fully-functional, datacentre. That sits and does nothing but can break in exactly the same kinds of ways over time.

And then you have to have a controller card, or another storage array, or a licensing for the server and software to make it failover, and site-failover logic and hardware, etc. on top of that cost.

I'm currently working at a place that can put a value on their data. They very nearly lost everything, and it would have cost an awful lot to get back running, let alone try and get their data back. Thus their DR is "proper" as they realised how much it would have cost in time and money, realised how much it would cost to avoid that (including my salary, for instance) and choose the "good" side of the coin.

As such, despite being a tiny employer by global standards, we have remote sites, remote servers, remote backups, full remote operation in an emergency, redundant leased-lines, redundant cabling around the site, redundant servers and all the logic to tie this together nicely.

But to secure System A against failure requires System A and System B of the same spec - MINIMUM - sensibly System C and maybe System D as well, plus the additional licensing and logic to fail them over and complete copies of EVERYTHING on them all. So you would have to pay 2-5x the total price your system cost originally, just to do a basic job of it.

When you do the maths, that STILL works out better than data loss, however. But nobody ever costs data loss properly until it happens and they realise how much it REALLY costs in terms of lost custom, legal requirements, hassle, time and money, the complete INABILITY to recover some data (no, you can't just post it off to 'a specialist' and expect anything to come back except a bill), etc.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon


Biting the hand that feeds IT © 1998–2020