From Zero to Thirty Servers

Mid last year, we made the decision to start evaluating cloud computing. I’ll admit, we approached the topic with some hesitation. Despite all the glowing press, we’d had some bad experiences with virtualization, including poor IO performance, instability and unpredictable slow-downs. To kick off our evaluation, we decided to host a prototype application on a handful of cloud servers to see how it went.

I’m happy to report the results have been mostly outstanding, in favor of the cloud. We can add and remove servers from our hosting stack at any time, and that flexibility opens up many possibilities that we just couldn’t consider otherwise. It’s incredibly liberating to have a new server instance up and running within 10 minutes of deciding you need one – and at an astonishing price.

That said, there are downsides. Servers can disappear without warning when the physical machine it’s hosted on goes down. It takes a different style of system architecture to allow for these failures. Failures don’t happen every day, but far more often than with high end, dedicated hardware. That said, we actually consider this a benefit…. our software practices have become more naturally fault tolerant, improving our processes overall. We’ve built an amazingly fault tolerant hosting infrastructure for our clients and their websites – any of our servers can suddenly go offline with no impact on our websites or clients, and we have the ability to quickly add (or remove) server capacity as needed to meet scaling challenges.

Based on our long trial period, we’ve made the decision to migrate most of our hosting infrastructure, minus a few strategic data servers, to cloud-based services. We now have around 30 server instances running in the cloud, with more on the way, and expect to enjoy greater flexibility and lower costs with very little (if any) negative impact.