Issue No.04 - April (2005 vol.6)
Published by the IEEE Computer Society
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MDSO.2005.23
For electrical engineering professor Sriram Vishwanath, measuring a network's capacity is the first step to improving it.
W e can all relate to losing a cell phone connection, getting "kicked off" the Internet while surfing, or waiting for what seems like an eternity for a file to download. What many of us might not realize, however, is the extent to which many everyday technological problems can be attributed to wireless capacity limits.
According to Sriram Vishwanath, an assistant professor with the University of Texas at Austin's Wireless Networking and Communications Group, the problem is twofold. "We don't know what the capacity of our wireless system is," he says. "And, even if we did our current system is operating far under capacity, and we don't know exactly how to improve it."
Earlier this year, the electrical engineering professor was awarded a US$400,000 US National Science Foundation Early Career Development (CAREER) grant to measure the maximum capacity of wireless networks. And, by many accounts, Vishwanath has his work cut out for him.
"Determining the capacity of wireless channels is a highly challenging problem," says Andrea Goldstein, an associate professor of electrical engineering at Stanford University.
Still, Goldstein, who is familiar with Vishwanath's research methodology, is optimistic. "He has taken a very creative approach to solving capacity problems, using new insights, different ways of looking at problems, and powerful mathematical tools to tackle them," she says.
Vishwanath believes that to formulate workable solutions, you must clearly identify the problems—the first of which is more of a communication issue than a technology one.
"The word 'capacity' [when referencing wireless networks] means a different thing to researchers working at each layer of the protocol stack, with each set of researchers unaware or only partially aware of the others' definition," he says. As a result, there are varying notions of capacity, each suited to a particular scenario.
Vishwanath also cites the over-simplification of lower layers when modeling the system. "The higher layers tend to abstract out the lower layers and assume 'suitable' models for them, the models going further and further away from reality the higher you climb," he explains. "Solutions become model dependent, and although these abstractions are useful for a first-order analysis, they do not 'add in' the complexities of the lower layers in a step-by-step fashion. Nor do they attempt to remain consistent with the notion of capacity at the lower layers."
Finally, Vishwanath says the difficulty in measuring wireless capacity can also be attributed to too much intra-layer research and not enough crosscutting research that merges theoretical concepts from one domain with more applied concepts from the other.
Vishwanath's approach to investigating wireless network capacity is to define and conquer.
"The first step… is defining a good notion of capacity and determining it is what this project's initial thrust is," he says. "Once we know what capacity is, my project and techniques will also produce algorithms and techniques that allow a system to get close to it in performance, and then we can design our system with the full knowledge of both capacity and techniques to approach it, leading to a much smarter design."
Vishwanath says he will use information theory, optimization theory, and coding theory to carry out this research. "I will rely on information theoretic tools for a lot of my analysis," he says. "However, information theory just lets you know what the limits are. We need optimization theory and techniques to find ways of achieving those limits. Once you have come up with the optimum solution, you have to implement it in software and hardware; that is where coding theory comes in. In essence, coding theory brings everything closer to practice."
Vishwanath isn't working directly with wireless providers on this project. "The first few years will be in a computer lab, working out the math and testing it with simulations," he explains. "I would like to see my ideas and algorithms eventually be implemented in a laboratory setting, but that is at least a couple of years down the line."
Vishwanath already has early progress to report. "We have made some headway with simple networks recently," he says. "These are three-node networks, and I'm happy to report that we understand them better today. They are a small but important stepping stone."
Whatever the project's outcome, however, identifying and testing the limits of wireless capability is a trend—one that's here to stay, says industry watcher Leif-Olof Wallin.
"All past experience indicates that the industry will come up with new solutions that will address known limitations and the 'barrier' will continue to be pushed for at least the coming two decades," Wallin says. "People will try hard to achieve higher speeds than what's projected as possible; it will become a challenge to break the barrier."
"My aim is towards [setting] an academic standard," Vishwanath says. "An industry standard would be a dream come true, but I'm not having those lofty dreams… yet."