Programming Distributed Erlang Applications: Pitfalls and Recipes

Paper in proceeding, 2007

We investigate the distributed part of the Erlang programming language, with an aim to develop robust distributed systems and algorithms running on top of Erlang runtime systems. Although the step to convert an application running on a single node to a fully distributed (multi-node) application is deceptively simple (changing calls to spawn so that processes are spawned on different nodes), there are some corner cases in the Erlang language and API where the introduction of distribution can cause problems. In this paper we discuss a number of such pitfalls, where the semantics of communicating processes differs significantly depending if the processes reside on the same node or not, we also provide some guidelines for safe programming of distributed systems.