Fault-containing network protocols

Abstract

Self-stabilization is a simple and elegant approach towards designing fault-tolerant network protocols. While self-stabilization provides automatic recovery from arbitrary transient faults, self-stabilizing systems do not incorporate any optimization for more efficient recovery from limited transient faults, even though such limited faults are more likely to occur in practice than arbitrary faults. Fault-containing self-stabilizing protocols are proposed to efficiently contain the effects of more frequent, limited transient faults while retaining the desirable property of self-stabilization. In this paper, we propose a methodology for designing fault-containing selfstabilizing network protocols. The methodology views the properties of self-stabilization and fault-containment in separation. Since self-stabilizing protocols already exist for many important network problems, the design starts with an existing self-stabilizing protocol, and adds to it the fault-containment property. The feasibilitv of this technique is illustrated bv the design of a fault-containing self-stabilizing protocol for the breadthfirst-search (BFS) spanning tree problem. The applicability of the technique used in deriving the BFS protocol to other problems is also discussed.