Performance evaluation of service execution in opportunistic computing

Abstract

Opportunistic computing has emerged as a new paradigm in computing, leveraging the advances in pervasive computing and opportunistic networking. Nodes in an opportunistic network avail of each others' connectivity and mobility to overcome network partitions. In opportunistic computing, this concept is generalised, as nodes avail of any resource available in the environment. Here we focus on computational resources, assuming mobile nodes opportunistically invoke services on each other. Specifically, resources are abstracted as services contributed by providers and invoked by seekers. In this paper, we present an analytical model that depicts the service invocation process between seekers and providers. Specifically, we derive the optimal number of replicas to be spawned on encountered nodes, in order to minimise the execution time and optimise the computational and bandwidth resources used. Performance results show that a policy operating in the optimal configuration largely outperforms policies that do not consider resource constraints.