Work As Play
“Research is like play. What makes things tick? Why do things behave the way they do? It’s satisfying when you figure out the answers.” – Donald Towsley
Well known nationally and internationally in the computer networking field, Towsley's expertise spans a wide range of activities from stochastic analyses of queueing models of computer and telecommunications to the design and conduct of measurement studies.
A large part of what Towsley does is to figure out how the Internet works from a mathematical perspective. He develops mathematical models in order to understand the different phenomena at play in networking processes. If a good model of a system or process can be developed, then protocols can be created to solve a certain problem or to enable that better mousetrap.
Take watching a movie on Netflix as an example. When the service first became available, there were long wait times as the movie loaded. Once started, viewers had to put up with freezes and changes in quality. Towsley’s curiosity led him and his colleagues to study the problem to determine how to alleviate these negatives by coming up with a more robust network design. Their models captured the “loading and freezing” phenomenon, enabling them to predict conditions under which it would arise. “We then were able to develop new algorithms and network protocols to prevent these things from happening,” says Towsley.
Some of Towsley’s most widely utilized research arose from figuring out how Transmission Control Protocol (TCP), the workhorse of the internet, functioned. Just about any data that you receive on your laptop uses TCP, says Towsley. It is a network-friendly protocol that evaluates how a data transfer performs, speeding it up or slowing it down, and improving transfer quality and efficiency. “For many years, no one knew how TCP worked. We came up with the first model of its behavior. The key to figuring this out was to think of data on TCP like sending fluid down pipes and to make the pipes larger or smaller depending on whether you think conditions are good or not so good within the Internet. We came up with a way of describing these fluid flows. It was a very accurate model,” says Towsley.
Because Towsley’s theory utilized differential equations, the language of control theorists, his control theory colleagues jumped in and began designing new mechanisms that could improve TCP. “We developed a mechanism that runs in routers that controls how these pipes increase or decrease in diameter, providing better performance and user experience. A modified version of our technology was adopted two years ago as the industry standard and is included in the latest cable modems,” notes Towsley.
Utilizing path diversity to improve network performance is another area of research Towsley has tackled, developing the initial theory on how to extend TCP to use multiple paths. “If you are going to the CNN website, for example, you will likely have one path to get there and your data is transported on that one path. But in reality, there are many paths that can be taken to access and deliver content. You could connect to several servers and stream portions from each of them. If one path fails, you can use others. So it makes sense to spread the data traffic across the paths,” says Towsley. Technology based on his theory has been standardized. “If you have an iPhone it is in your iPhone. It will eventually be prevalent all over the Internet,” says Towsley.
Though smooth data transmission and quality of experience for movie-watchers is useful, there’s much more at stake for research on networks, says Towsley.
“Robustness and security are two of the most important issues facing networks. There have been numerous denial-of-service attacks over the last few years, the most notable at Sony Studios, Google, and Amazon. We need to better understand network traffic patterns, network faults and failures, and how to deal with these attacks,” he notes.
Path diversity can be a solution for network security as well, notes Towsley, helping to mitigate attacks. “When an attack is sensed, network algorithms can shunt it to other places. You must have a good mathematical understanding of how networks behave so you can see how they are going to react to these different scenarios,” says Towsley.
Lately, Towsley has joined forces with UMass colleague Dennis Goeckel and former PhD student Boulat Bash to investigate an intriguing new area of network security that blends technology with social science: covert communications. While encryption can make a message undecipherable to an unintended observer, covert communications ensure the observer is not even aware of the transmission.
“We came up with the fundamental limits of how two people can communicate over a wireless medium without a third party even being award of it when one party is not allowed to freely communicate,” says Towsley. He and colleagues are also looking at how people can hide information in images, such as those on Facebook. “There are people, some of them undesirables, sharing information this way. We are looking at what you can or can’t do without being detected and what we can or can’t do to prevent this from happening,” says Towsley. This turned out to be a very interesting mathematical modeling problem, notes Towsley, and where much of his love of the work comes in. “I didn't have all the tools I needed to do this work, so I ended up learning a lot from my colleagues which made it very interesting. It’s fun and satisfying to master new techniques,” says Towsley.
The real beauty of this research, adds Towsley, is that it not only attacks an important societal problem, but it also opens up a new area of information theory. “Already, numerous research groups are focusing on the implications of our results and generalizing them to a greater variety of systems. It’s very exciting.”
So what’s next for Towsley? His curiosity has led him to new research investigations in quantum computing and quantum communications. “It’s a very rich, fertile area. Almost everyone has focused on quantum communication over a single link. In the future, there will be a need for an infrastructure to support distributed quantum computing and quantum communication. There isn’t anyone working on that yet. It is inevitable that we’ll need to develop the theoretical foundation from which we’ll be able to do good engineering. My hope is to borrow what I know from classical networks to apply here,” says Towsley.
In all of this, Towsley’s ultimate goal is “to have fun, to learn, and to play with others whether they be students or colleagues. My experience has been that the whole is much greater than the sum of the parts.”
Karen J. Hayes '85