An Erlang-based Philosophy for Service Reliability

Jamshid Mahdavi
Senior Software Engineer @ WhatsApp

"Let it crash" is a core tenet of Erlang language and programming. At WhatsApp we have a somewhat unique take on running a reliable service, and it turns out there are some interesting similarities between the Erlang language and how we think about and run the WhatsApp service.

In this talk I'll explain how we develop our server components, our deployment processes, and how we monitor, alert, and repair the inevitable failures that will happen in a service running at billion-user scale.

Talk objectives:

- Listeners will learn some of the ways WhatsApp has been able to reach billion-user scale with a very small server team - thanks to many of the same ideas that Erlang itself embodies.

Target audience:

- People interested in the nexus between software development, operations, and the Erlang language.

Slides
Video

Jamshid has been working on networking and networking products for 25 years. In his early career he participated in research on TCP congestion control and was co-author on a number of RFCs. More recently he has developed networking products for web and media streaming at several silicon valley companies. He has been at WhatsApp for the last two years and is currently responsible for their media server infrastructure.


Back to conference page