A research group at Dublin Institute of Technology hopes to speed up Internet and computer communications using advanced mathematics, writes Dick Ahlstrom.
Having to deal with heavy traffic flows isn't only a problem for the gardaí. Computer-to- computer communications also experience their traffic jams, making the Internet move at a snail's pace.
A new research group set up at Dublin Institute of Technology in Rathmines Road hopes to break this data logjam, helping data to flow more smoothly and getting more out of existing communications networks. It is all about "traffic engineering", according to Prof John Lewis,who heads the research group.
Lewis is the principal investigator at the DIT's Communications Network Research Institute (CNRI), set up with funding from Science Foundation Ireland. Its main goal will be "the basic science and technology associated with communications networks", says Lewis. "Our first concern is with the Internet."
Internet service providers have for some time attempted to solve an overload in data traffic by adding more and more "bandwidth", communication links that could take higher data volumes. This hasn't worked well,however, and the traffic jams will only get worse in the future.
"At the moment the Internet is highly unreliable," says Lewis. "In order for it to be a success for the service providers there has to be some economic discipline in it."
More bandwidth means more money thrown at the problem. Lewis's approach leaves the bandwidth the same but makes it work more efficiently through traffic engineering.
"Our work is mathematically based. We are providing a mathematical base for traffic engineering," he says. "Traffic engineering is used to describe the improvement they are trying to bring into the Internet."
The goal is to get a better match between an actual bandwidth offered and a given service requirement.
While gardaí watch the number of cars on the road, Lewis's team studies the number of data packets moving down a communications link. Instead of taking an exact count, it uses advanced mathematics to sample the volume and make assumptions about traffic.
The aim is to replace "hard guarantees by statistical guarantees" about the amount of traffic, he explains. "There are well-known examples of this in the telephone system."
A given phone network might need 1,000 lines to provide an absolute 100 per cent guarantee that there will be no blocked lines because of traffic, he says. The same-sized network only needs 64 lines if the demand is for a 99 per cent guarantee. "There are huge economies of scale when you bring in statistical guarantees and that is what traffic engineering is all about."
The approach is based on "Erland Theory", he says, developed by a Danish mathematician during the 1920s who was attempting to model customer behaviour.
Network providers wanted to use mathematics to understand what their over-burdened networks were doing, developing huge models that looked at thousands of parameters.
"Clearly that wasn't going anywhere," says Lewis. "The breakthrough came with my group in the Dublin Institute for Advanced Studies in 1993."
The institute had joined with Cambridge University and communications company Telia in an attempt to crack the problem. They developed a mathematical method to measure traffic directly over a given time period. It worked because they were measuring directly rather than trying to model.
"We showed it was possible to measure the essential characteristics of the traffic online and at real time. We were sampling these rates and then that enables us to estimate what is known technically as the rate function." Because it is real time, the service provider "can get a whole new view of traffic engineering."
The research group at DIT is now doing a "systematic development of traffic engineering from a mathematical point of view" that can be used in any situation were data is being moved about.
While this research work is relatively close to market as it progresses, the CNRI is also involved in quite exotic research in the area of quantum computing. This involves using the quantum states of atoms to store and process information, a technology that would deliver unprecedented computing power.
"This is real blue-sky stuff, but when the time comes and it becomes a commercial reality it will be important for Ireland to have this expertise," says Lewis.
This work will progress with an expert in quantum information theory, Prof Christopher King, who came here from Northeastern University, Boston to join the CNRI team.