SNTS: sensor network troubleshooting suite

  • Authors:

  • Mohammad Maifi Hasan Khan;Liqian Luo;Chengdu Huang;Tarek Abdelzaher

  • Affiliations:

  • University of Illinois at Urbana Chamapign, Department of Computer Science;University of Illinois at Urbana Chamapign, Department of Computer Science;University of Illinois at Urbana Chamapign, Department of Computer Science;University of Illinois at Urbana Chamapign, Department of Computer Science

  • Venue:
  • DCOSS'07 Proceedings of the 3rd IEEE international conference on Distributed computing in sensor systems
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Sensor network troubleshooting is a notoriously difficult task, further exacerbated by resource constraints, unreliable components, unpredictable natural phenomena, and experimental programming paradigms. This paper presents SNTS (Sensor Network Troubleshooting Suite), a tool that performs automated failure diagnosis in sensor networks. SNTS can be used to monitor network conditions using simple visualization techniques as well as to troubleshoot deployed distributed sensor systems using data mining approaches. It is composed of (i) a data collection front-end that records events internal to the network and (ii) a data processing back-end for subsequent analysis. We use data mining techniques to automate failure diagnosis on the back-end. The assumption is that the occurrence of execution conditions that cause failures (e.g., traversal of an execution path that contains a "bug" or occurrence of a sequence of events that a protocol was not designed to handle) will have a measurable correlation (by causality) with the resulting failure itself. Hence, by mining for network conditions that correlate with failure states the root causes of failure are revealed with high probability. To evaluate the effectiveness of the tool, we have used it to troubleshoot a tracking system called EnviroTrack [4], which, although performs well most of the time, occasionally fails to track targets correctly. Results show that SNTS can identify the major causes of the problem and give developers useful hints on improving the performance of the tracking system.