Find Bottlenecks

Building software for cloud computing is largely about optimizing the application given the new tools available to you.

When you need to scale, your first question is: How? Do you have a lot of users, or just a lot of hits? Can users share some data, or is each user's data unique? What must happen when users hit the site, and what can you cache in advance?

If you compare how people have solved the problems of scaling common applications, such as WordPress, and online applications, such as Twitter, you'll see that there are very different problems.

Scaling on the cloud involves a lot of decoupling – making one thing work completely independently of another so that each process isn't held up by another. Sometimes, doing this will require re-working your application, but if you're lucky enough to be building from scratch, make sure you don't choose methods out of habit.

If you are scaling an existing application or building one from scratch, you will need to optimize. Optimization is one of the never-ending tasks – Google goes on about how they shave every millisecond off the page load time as possible, and if you've ever had to optimize, you'll understand this.

Make sure you have tools for optimizing and you know how to use them – each language has a host of tools for benchmarking and profiling the code. Optimization is important when you think about scaling, because if your code contains a bottleneck, the speed problem will multiply with the number of users. Plus, bottlenecks are good candidates for decoupling the application.

Building in Clouds

As soon as you worked out the quickest way to deliver a web page from your rack servers, along came cloud computing providers with thousands of servers and a whole new approach. Getting these cloud services into your toolset will save you from much of what Amazon calls the "heavy lifting."