Avoiding Internet congestion the smart way
Imagine if motorway traffic had to be suddenly routed through busy town centres – what kind of congestion and delay would that cause? Your Internet connection could be facing a similar issue: consumers increasingly expect fast connections and high volumes of data to transfer quickly, but the Internet is primarily based on technology developed in the 1970s and it’s not always equipped to manage resources smartly and cope with more modern demands.
“You could say that the Internet is now operating well beyond its initial design,” says Dr William Donnelly, director of the Telecommunications Software & Systems Group at Waterford Institute of Technology, where a team is looking at new ways to ease the congestion.The telecommunications solution that has worked up to now is just not flexible enough to handle an explosive growth in traffic, and the traditional approach of predicting traffic and planning resources accordingly is unlikely to suffice any more as networks become more complex.
“We now have users communicating with users, users communicating with machines, and machines communicating with machines,” says Dr Donnelly. “It’s a heterogeneous environment spanning many different network technologies.”
To extend the transport analogy, currently congestion in Cork might be routed to Portlaoise - but the system has no way of knowing whether or not Portlaoise is already congested. So unless networks can be managed more efficiently, as traffic increases, congestion is going to bring everything to a halt. But how can resources be allocated more dynamically?
“If the next generation of networks are to work, they have to be self-governing, self-healing, self-optimising, and self-protecting,” says Dr Donnelly, whose group is leading a European research programme on Autonomic Network Management. “It is possible to replace the simplistic predictive approach with a more sophisticated system with the automatic capacity to balance real-time needs against real-time capacity.” Such a system can be fine-tuned to distinguish between different levels of demand, and instead of having fixed allocations it could allocate bandwidth as required on the basis of priority.
Using an autonomic management system would be a more cost-effective way to maintain a high service rather than trying to add to existing networks, where costs could spiral out of control, according to Dr Donnelly. “Service providers buy enough bandwidth to satisfy expected demand,” he explains. “Having to buy in extra capacity can eat into organisational profits, but if they can better balance demand, they can squeeze more out of the existing system.”
The emerging patterns of mobile Internet use by consumers also provide an impetus for self-regulating networks that can adapt on demand rather than trying to predict resource usage in advance. “Traffic is being driven by user behaviour, and we are getting patterns that we have never had to deal with before, such as virtual communities, geographically separate, but linked by telecommunications.” Improving the efficiency of switching would also improve the bottom line in network management: “It’s not the fibre, but the inefficient switching that is so expensive,” says Dr Donnelly. “Find different ways of getting into the network and a lot of problems would be solved.”
The Telecommunications Software & Systems Group group at Waterford Institute of Technology has spun out 12 companies since 1996 and has received funding from SFI. In 2008, Dr Donnelly was awarded funding from SFI to establish a Strategic Research Cluster, FAME, which builds on the autonomic communications management work of a previous SFI award, and includes collaboration with industrial partners in Ericsson, Cisco, IBM, Teléfonica I&D, Alcatel-Lucent and HPy and academic partners in UCD, NUI Maynooth, TCD and UCC. HEAnet support the activity with Ireland’s academic research network.
