Research Statement for Arunesh Mishra

We have a choice of a multitude of wireless technologies today, each optimized for a single and distinct service model – for example, cellular for voice and WiFi for data. The next challenge is thus, convergence, which places the user in control by allowing for a user-driven model of service rather than the technology dictating what services are available. My thesis has examined one part of this puzzle – understanding the challenges faced by voice applications when using WiFi. The focus has been on handoffs and the related latencies faced by users that are mobile within an 802.11 network. A unique aspect of my thesis work has been the design of practical system-level solutions which work transparently with the standards in reducing the handoff latencies from 1.2 seconds to under 50 ms. These solutions are based on the notion of neighbor graphs a structure that captures the topology of a wireless network from the perspective of user mobility. We believe that this concept could have interesting applications beyond handoffs, for example, in providing better quality-of-service, aiding network management and improving network security. A next broad direction of my work has been in optimizing the performance of WLANs by building practical and intelligent channel assignment algorithms. One unique contribution of this work is the concept of client-driven network management for WLANs and construction of protocol-level methods to realize this. We show that client participation in channel assignment and for network management in general, is critical for operating the network efficiently at peak performance. We have also built distributed algorithms to manage 802.11 channels in an uncoordinated wireless environment, for example, 802.11 access points (APs) belonging to individual hotspots and apartment homes which interfere and co-exist with each other. These distributed algorithms perform a fair and near-optimal division of the system’s resources using the concept of channel hopping. Finally, through careful observations and analysis we’ve also debunked the traditional approach of avoiding partially overlapped channels which have been defined by the IEEE standards body. We’ve shown that by careful modeling of the interference between two partially overlapped channels, one can utilize these channels effectively, thus improving the overall system throughput by a factor of three. With this experience, I hope to employ my skills and expertise to new and challenging research problems in the broad domain of systems, wireless/wired networking and security.