Lightweight tracing for wireless sensor networks diagnostics
Wireless sensor networks (WSNs) are being increasingly deployed in various scientific as well as industrial domains to understand the micro-behavior of physical phenomena. WSNs are highly susceptible to post-deployment failures due to their in-situ deployments in harsh environments e.g., volcanoes. The traditional tools and techniques to handle such failures are inadequate because of inherent resource constraints of WSNs. The lack of diagnosis tools for post-deployment failures hinders the more widespread adoption of WSNs. ^ An execution trace containing events in their order of execution can play a crucial role in postmortem diagnosis of these failures. Obtaining such a trace, however, is challenging due to stringent resource constraints. In this dissertation, we propose an efficient distributed control-flow tracing technique for WSNs based on two key observations. First, WSN executions are highly repetitive, and second, WSNs exhibit restricted communication patterns. Our distributed control-flow tracing combines three novel lightweight techniques: (1) an efficient interprocedural control-flow encoding technique that generates a succinct control-flow trace of all events happening within a node, (2) a generic hybrid trace compression technique that significantly compresses traces, and (3) a space efficient message tracing technique that is amenable to compression. We show our tracing technique's effectiveness through failure case studies and efficiency through measurements and simulations.^
Patrick Eugster, Purdue University.
Engineering, Computer|Computer Science