back
7 comments
I've been saying this about EC2 for some time (and we have a product that manages EC2 instances, so I have a vested interest in seeing EC2 become successful). It's just not very reliable, when compared to the best hosts in the business--while EC2 downtime per year can be measured in hours, the best hosts are down minutes or seconds each year (of course, your apps and your system administration abilities can easily make up the difference--we've been offline a couple of times due to Joomla problems, and there's nothing a more reliable network or server could do about it). But, every bit of downtime adds up...and maybe your next big customer or investor was looking at your site at that very moment.
Our hosting spend is around $1m AUD a year (although a lot of that is bandwidth which is expensive in AU) and the datacentre we use took all its customers offline for 2 hours last week. It happens. Even the to the big guys.
You should check with Netcraft's reliability rankings next time. 2 hours is an awfully long time for a "big guy".

Sure, downtime happens. Routers are upgraded and power bounced (30 seconds of downtime if the admins are good), UPS infrastructure is kept up-to-date and in working order (minutes of downtime), and server components break. The last one is possibly a lot of downtime...but hopefully MTBF works for you and the server that you keep for the average lifecycle of hosting servers (about 30 months) never fails catastrophically.

From our experience last night, 1 or more Amazon datacenters went off the net. Our instance kept chugging along while it was unavailable (but we didn't know this at the time). We were firing up another instance in a different data center (using the new trusted zones feature) when the outage ceased and our site on the original instance resurfaced.
Nice - there's a vote of confidence.
EC2 is still in beta. The fact that a beta was down for an hour at 2am shouldn't be newsworthy.
Even as a beta that beats the uptime for a lot of other services I've used.