Service evolution in a nomadic wireless environment

Abstract

In this paper, we present and analyze a framework for self-evolving autonomic services in a wireless nomadic environment. We present a disconnected network architecture, where users mobility is exploited to achieve a scalable behaviour, and communication is based on localized peer-to-peer interactions among neighboring nodes. Service management is achieved by introducing autonomic services, whose operations are based on a distributed evolution process. The latter relies on the concept of mating, i.e., the exchange of information (e.g., code, parameters, data) among service users, which collaborate to enhance their fitness, defined as the ability of the actual service to fullfill the environmental features. The core of the evolution process is given by the service mating policy, which defines the way the running services should be modified when mating with other users. We introduce a general framework for analyzing service mating policies and exploit results from martingales theory to study their convergence properties. In particular, we introduce two optimal policies, clone-and-mutate and combine-and-mutate, and analyze their convergence times through extensive numerical simulations, addressing the impact of various parameters (number of nodes, users speed, mobility pattern). © IFIP International Federation for Information Processing 2006.