SCS

Abstract

Simple, distributed and iterative algorithms, popularly known as message-passing, have emerged as the architecture of choice for a variety of networks. They have been surprisingly effective despite their simplicity. In this talk, I will try to argue in favor of such algorithms by discussing an example from wireless communication networks. Specifically, I will discuss the design of an efficient medium access algorithm for wireless networks using queue-sizes. Here nodes wish to transmit without interfering with each other while maximizing utilization of the wireless medium. To minimize co-ordination cost, solutions implemented in practice are based on `random access'. However, they perform quite poorly as proved in theory and observed in practice. I will present an `adaptive' random access algorithm that is provably efficient in both asynchronous and synchronous setup. This work draws insights from the classical variational principle, mixing times of Markov chains and reversibility. The talk is based on joint work with Jinwoo Shin, MIT.