Building for the Cloud: Lessons Learned at Heroku

The Heroku platform grew from hosting 0 applications to more than 115,000 without ever buying a server. We’ll talk about the overall architecture of the platform and dive into some of the operational consequences of those architectural choices.

Some examples of the types of things that will be covered include:

• Dynamic infrastructure. What are the pitfalls of having a constantly changing set of nodes running your infrastructure?
• The noisy neighbor problem. How do you handle having widely variable performance from identical nodes?
• How much dogfooding is too much? When your platform is the easiest way to deploy and application, it’s tempting to deploy parts of your infrastructure on your platform. Circular dependencies can be nasty, where do you draw the line?
• Handling failures. Why everything should be as disposable as possible.
• Networking hacks. How do you host SSL sites when you can only get one IP address per server?