ToolBox -

Amazon makes great tools! Not cheap, but amazing to be sure.

Unfortunately, just having access to servers by the hour doesn’t automatically create a highly-available, scalable, fast WordPress cluster. Here’s why.

First, ELB won’t make a single page fast, it just warms up more Apache servers when you’re under heavy traffic load. Also it doesn’t warm them up quickly which means a traffic spike is usually a problem until more boxes are allocated, launched, and join the cluster. That can be mitigated by keeping more running even when not necessary; of course that means running servers you don’t normally need, which is expensive.

ELB doesn’t scale the database, which is often the first thing to choke under load. Amazon does have the RDS service making it easier to build things like MySQL slaves. That’s good, but also requires you to configure WordPress to use that, and of course running multiple MySQL servers is expensive.

Now that more than one web server is running WordPress, you’ll need a way to share the filesystem between those boxes too (e.g. NFS), which is another server to setup and manage. And a new single point of failure.

Also the usual caching things (W3TC with memcache for example) have to be managed differently in a cluster with multiple web servers. That’s more servers to pay for and manage. (But usually doesn’t represent a single point of failure however; normally when caching servers are unavailable WordPress can proceed, just a lot slower, especially since it will continue to try to contact the caching servers and time out.)

Also you still have to monitor the health of your servers and perform emergency work if something goes wrong. EC2 doesn’t have WordPress tech support! Or any tech support. Whereas we have 24/7 monitoring with awake human beings and real WordPress-specific assistance.

Also EC2 has no support for security. You have to know enough to lock down the servers, the database, the installation, the filesystem, WordPress itself, your plugins, your theme, etc.. Whereas at WP Engine, that’s our problem.

In the end, you’ll find it’s a lot of moving parts, and when you add the server and bandwidth costs to the human time, it’s expensive — much more expensive than us!

We do all the above and actually quite a bit more; here’s the detail. We can afford to do all that because we amortize the cost across all our customers. It’s typically way too much for a single customer to bother with.