Self-stabilizing with service guarantee construction of 1-hop weight-based bounded size clusters

Abstract

This paper makes contributions in two areas. First, we introduce an extended approach of self-stabilization, called self-stabilization with service guarantee. A self-stabilizing system tolerates transient faults: it automatically recovers to a correct behavior after a stabilization period. However, during stabilization periods, no property on the system behavior is guaranteed. A self-stabilizing protocol with service guarantee quickly provides a useful minimal service, and it maintains the minimal service during stabilization despite the occurrence of some disruptions. To illustrate our approach, we propose a clustering protocol (called SG-BSC), that builds 1-hop, bounded size and weight based clusters. SG-BSC protocol is self-stabilizing with service guarantee: it quickly reaches, in 4 rounds, a safe configuration from any arbitrary one. In a safe configuration, the following useful minimal service is provided: “each node belongs to a cluster, and each cluster has a leader and at most SizeBound ordinary nodes that are at most at distance 1 from their cluster-head”. The convergence to a legitimate configuration (optimum service) is done in at most 7∗N2+4 rounds, where N is the number of nodes. We prove that any self-stabilizing protocol building weight-based clusters requires O(N) rounds to stabilize. Once the optimum service is reached, any cluster-head has the highest weight in its cluster, and the number of clusters is locally optimal. During the stabilization period, the minimal service is preserved; so, the hierarchical organization stays available throughout the entire network. Simulation results show the interest of self-stabilization with service guarantee compared to self-stabilization.