So fast forward to this week and it was time to try again. This time I went with an Extra Large instance running a stock 64-bit Ubuntu 8.04 and Lighttpd instead of Apache. And this time there were no problems. For the full 2 hours the site remained up and responsive, and because I could time the instance to come up just before the additional cost was negligible.

To give you an idea of the kind of traffic it had to deal with, the site took in just over 25,000 unique visitors in a 2 hour time frame. Glad to get the problem solved and now I know what to do next time I need some heavy traffic handled.

So I set up a small instance running on EC2, running Ubuntu 8.04 and tweaked Apache2 a little to increase the MaxClients setting.

Unfortunately not everything went to plan.  While the server itself was rock solid the response times from Apache grew very quickly to 20+ seconds.  So I’m trying to figure out what the next plan of attack is.  Do I set up a load balancer on EC2 and share the load with a number of instances, or do I go with a larger instance for the duration of the traffic spike, upgrading to a large or extra large instance.  Are there some settings I should be tweaking on Apache?  Should I be trying out an alternate webserver like Lighttpd?

Or maybe I give up and get back to the day job  Any advice appreciated.

For 2 years we’d been using a Fedora 6 box which had been running fine, but it was becoming difficult to keep the OS up to date given Fedora’s habit of ignoring previous releases fairly quickly.  Other issues, like the fact that there was a single hd in the machine which could potentially die at any time and cause much pain, prompted the desire to move everything to something more robust, flexible and easy to backup.  Enter Amazon EC2.

Fedora was ditched in favour of Ubuntu 8.04 – the latest LTS release – and within an hour we had set up a small instance in the Euro availability zone.  The doc’s for EC2 are excellent and take you through the process.  An Elastic IP was attached to the instance to ensure we could maintain the ip should the server ever go down.  

We also attached a 120gb Elastic Block Store instance, mounted at /var/www to make it easy to take all data easily to another instance should the need arise.  Backups are handled with the  AMI and EC2 tools.  Nightly the instance is bundled and stored on S3 and nightly a snapshot is take of the EBS block as well.

So far the instance has been rock solid reliable and much faster than the old Fasthosts dedicated box we had.  EC2 is exactly what anyone hosting websites need at the moment – no tie-in to any physical hardware and the flexibility to scale your hardware as required.

I’ve seen a lot of criticism about hosting web sites on EC2 but almost all of the issues I’ve seen are no longer relevant as Amazon have taken steps to improve EC2 and add critical features (like persistent external IP addresses).  

Really glad I got the opportunity to implement this as my last act at oceanseventy.  I’m definitely and EC2 advocate now and hopeful the skills i’ve learned will help my in my new position at STV.  Hopefully i’ll be able to use EC2 for my startup project as well.

I’ll write more soon about our setup.