Middleware for decentralised fault tolerant service execution using replication in pervasive systems

Devices in pervasive systems are resource constrained, heterogeneous, mobile and personal. The devices may seek or provide such services as data compression, encryption, image analysis, query processing and other types within a local, but dynamic network. In order to enable sharing of services and resources, it is necessary that the middleware facilitate collaboration amongst devices. Resource and service sharing in such environments is a challenge due to device mobilities, heterogeneity and link failures. In this paper, we propose a middleware that enables decentralised decision making for fault tolerant service execution in pervasive systems. Opportunistic contacts between pairs of devices are exploited to locate service-device mappings and initiate replicas of the required service. Redundancy is an unnecessary, but unavoidable consequence of any replication based fault tolerance scheme. One of the objectives of the proposed middleware is to minimise the number of replications. Simulation results show that the proposed scheme has lower time overhead as compared to existing schemes.