More information about the Underscore mailing list

[_] AWS users

Andy Gale andy at
Tue Feb 5 16:58:25 GMT 2013

On Tue, Feb 5, 2013 at 4:37 PM, Jon Free < at> wrote:

> What because Amazon went down a couple of times and then so did they?
> As with anything it's a balance. But if using it meant they could scale a
> tiny company into a billion dollar sale without having to buy millions of
> dollars worth of hardware then surely it was worth it?

No it's not shit obviously and there's no excuse for all that "EC2
went down so my site went down" nonsense either. Check out the Netflix
Technical Blog for examples of how they detect downtime and route
stuff away from Amazon failures.

A really good one is the Chaos Monkey that randomly kills their
instances to ensure they can always deal with failure.

The problem is that at small scale the costs of having a redundant EC2
setup are more than having a standard VM with someone like Bytemark.
Amazon EC2 instances do seem to be the more likely to disappear and
most other cloud providers guarantee your instance won't get killed
off unless something terrible happens.

Using Amazon EC2 does encourage you to design for failure which is a
good thing but that shouldn't really be considered a feature in their



Andy Gale