Track: Systems that never stop

Abstract: Erlang was designed in 1986 at Ericsson for building Carrier Grade Telecoms systems. Erlang has a number of unique features that are explicitly designed into the language to simplify programming non-stop systems. For example, Erlang code can be changed on-the-fly so that systems can be upgraded without taking them out of service. Such systems have been deployed by Ericsson and are routinely upgraded with zero loss of service. The largest system written in Erlang is the Ericsson AXD301 (an asynchronous transfer mode switch) which has outstanding in-service performance and which has achieved "nine nines" reliability. In this lecture I'll start with an overview of the techniques we use to achieve high reliability. I'll show how these relate to language primitives in Erlang and to a number design patterns that we use to build highly reliable systems.