Virtualizing Closed-loop Sensor Networks: A Case Study.

Closed loop sensor networks are cyber-physical systems that establish a tightly coupled connection between computational elements and the control of physical elements. Existing closed-loop sensor networks are based on dedicated, 'stove-pipe' architectures that prevent the sharing of these networks. This paper addresses the problem of sharing these networks through virtualization. We propose scheduling algorithms that manage requests from competing applications and evaluate their impact on system utilization as compared to a dedicated network. These algorithms are evaluated through trace-driven simulations. We aim to demonstrate that the proposed scheduling algorithms result in cost savings due to shared network infrastructure without unduly affecting application utility. In our evaluations, we observe only a 20% reduction in average utility via the DSES scheduling approach.