Automating Wireless Network Management: Lessons from Managing Wireless LANs and Sensor Networks
Supported by NSF CNS-0746841
Wireless networks, by virtue of their untethered nature, ease of set-up, and mobility support, have been heartily welcomed into our daily work and lives. Crucial to the continuing success and even wider deployment of wireless networks is automatic management of these networks.
The primary objective of this project is to automate wireless network fault management. In particular, we investigate two fundamental questions of fault management in wireless networks: (1) What are suitable management architectures and what data/measurements need to be collected for effective management? (2) What are suitable techniques for fault management? To be focused, this research mainly investigates two “extreme” forms of wireless networks -- wireless LANs (WLANs) and sensor networks.
For managing WLANs, we proposed a novel WLAN management architecture that uses high-level measurements at a single centralized wired monitor to guide low-level measurements at distributed air monitors inside the network. The centralized monitor is at an aggregation point of a local area network. It captures all packets coming into and going out of the network, and hence is at an ideal location for monitoring the health of the WLAN. The distributed air monitors can collect detailed physical- and MAC-layer information that is critical for detailed understanding of the network and fault diagnosis. Under this architecture, we have investigated rogue access point detection, health monitoring, sniffer channel assignment, and more recently network performance of smart handheld devices.
For managing wireless sensor networks, we have been developing novel fault management techniques under three monitoring architectures: end-to-end monitoring, air sniffing, and localized monitoring. Under the end-to-end monitoring architecture, we have developed a novel approach that carefully combines passive end-to-end measurements and active measurements to localize faults. Under the air-sniffing architecture, we have rigorously quantified the capability and fidelity of mote-class sniffers for sensor network monitoring, formulated and solved a k-monitoring problem for placing sniffers, and investigated using air sniffers to passively monitor delays and detect abnormal delays without the need of clock synchronization. Under localized monitoring architecture, we have investigated neighbor discovery in multiple packet reception (MPR) networks, and failure node detection in mobile sensor networks.
- Delay monitoring for wireless sensor networks: An architecture using air sniffers
Wei Zeng, Jordan Cote, Xian Chen, Yoo-Ah Kim, Wei Wei, Kyoungwon Suh, Bing Wang and Zhijie Jerry Shi.
Elsevier Journal of Ad Hoc Networks, accepted.
- Leveraging cloud infrastructure for troubleshooting edge computing systems
Michael Fagan, Mohammad Maifi Hasan Khan and Bing Wang.
Proceedings of IEEE Conference on Parallel and Distributed Systems (ICPADS), December 2012.