A greedy approach for resource allocation in Virtual Sensor Networks

Abstract

Virtual Sensor Networks (VSNs) envision the creation of general purpose wireless sensor networks which can be easily adapted and configured to support multifold applications with heterogeneous requirements, in contrast with the classical approach of wireless sensor networks vertically optimized on one specific task/service. The very heart of VSNs' vision is the capability to dynamically allocate shared physical resources (processing power, bandwidth, storage) to multiple incoming applications. In this context, we tackle the problem of optimally allocating shared resources in VSNs by proposing an efficient greedy heuristic that aims to maximize the total revenue out of the deployment of multiple concurrent applications while considering the inherent limitations of the shared physical resources. The proposed heuristic is tested on realistic network instances with notable performances in terms of execution time while keeping the gap with respect to the optimal solution limited (below 5% in the tested environments).