Self-stabilizing Distributed Algorithms for Networks

Abstract

Many essential fundamental services for networked distributed systems (ad hoc, wireless or sensor) involve maintaining a global predicate over the entire network (defined by some invariance relation on the global state of the network) by using local knowledge at each of the participating nodes. The participating nodes can no longer keep track of even a small fraction of the knowledge about the global network due to limited storage. We need a new paradigm of localized distributed algorithms, where a node takes simple actions based on local knowledge of only its immediate neighbors and yet the system achieves a global objective. Self-stabilization is a relatively new paradigm for designing such localized distributed algorithms for networks; it is an optimistic way of looking at system fault tolerance and scalable coordination; it provides a cost effective built-in safeguard against transient failures that might corrupt data in a distributed system. We introduce self-stabilizing protocol design with the example of a total dominating set in a network graph and discuss some open problems.