Reliable upgrade of group communication software in sensor networks

Abstract

Communication is critical between nodes in wireless sensor networks. Upgrades to their communication software need to be done reliably because residual software errors in the new module can cause complete system failure. We present a software architecture, called cSimplex, which can reliably upgrade multicast-based group communication software in sensor networks. Errors in the new module are detected using statistical checks and a stability definition that we propose. Error recovery is done by switching to a well-tested, reliable safety module without any interruption in the functioning of the system. cSimplex has been implemented and demonstrated in a network of acoustic sensors with mobile robots functioning as base stations. Experimental results show that faults in the upgraded software can be detected with an accuracy of 99.71% on average. The architecture, which can be easily extended to other reliable upgrade problems, will facilitate a paradigm shift in system evolution from static design and extensive testing to reliable upgrades of critical communication components in networked systems, thus also enabling substantial savings in testing time and resources.