Skip ring topology in FAST failure detection service

Abstract

This paper addresses the problem of communication among loosely coupled groups of nodes in distributed systems. We describe a novel proposal of logical communication topology based on skip list data structure. We enhance this structure to make it more resilient to failures. Its good self-stabilization characteristics are shown through extensive simulation experiments. We present this new concept in the context of our failure detection service, where we use it at a local communication level. © 2008 Springer-Verlag Berlin Heidelberg.