Wireless Network Utility Maximization

Dan C. O'Neill
Stanford

Wednesday, May 27, 2009 at 1:00pm
Engr IV Room 57-124

Abstract
WNUM, Wireless Network Utility Maximization, extends the NUM approach to wireless networks in random time varying environments. The goal is to find useful, tractable methods to mange network resources. The approach builds on techniques used in stochastic optimization and stochastic approximation to find optimal network management and control policies that manage network resources. WNUM does not assume prior knowledge of the underlying random distributions, but rather learns necessary channel characteristics through sampling. The talk will outline the WNUM approach, discuss convergence properties, and describe two applications. The first, finds approximately optimal control policies in interfering MANET's. The result is a description of the rate, reliability, delay, tradeoff for the system. The second application applies WNUM to manage buffer stability in wireless video applications with constraints on buffer underflow and overflow.

Biography
Dan C. O'Neill is currently Assistant Consulting Professor at Stanford University. His research interests are in the areas of network management and control using techniques from stochastic optimization, ADP, and stochastic approximation. An additional interest in practical stochastic techniques to mange power and energy usage in data center (servers and routers) applications. He is involved in several wireless projects and is a PI on projects with DARPA and AFOSR. Prior to this he was Senior Director at SUN Microsystems, President and CEO at Clearwater Networks, General Manager of the Integrated Processor Division at National Semiconductor, and VP and General Partner at Merrill Lynch Venture Capital. He received his PhD degree in Electrical Engineering from Stanford University and his MBA (management science) from UC Berkeley.