Research versus reality

Researchers in the US have developed a new Transmission Control Protocol (TCP) that will allow the Internet to operate at speeds approaching 10Gbit/sec. Dave Wilson reports.

‘Damn the torpedoes, Full speed ahead!’ – Admiral David Glasgow Farragut (1801-1870).

Despite the tremendous speed advantages of today’s broadband Internet technology, not everyone in England can get it at home yet.

Take Jason Ford, our Deputy Editor. At his folks’ home in the East of England, broadband options are non-existent. Oh, sure, he could crawl the Internet from there on a 56kbit/sec line, but at the moment, it’s almost as quick for him to get on a train to London and work on a high-speed connection from there.

Thank goodness, though, that the pathetically slow roll out of broadband by our major telecommunications providers hasn’t stopped the chaps that really know a thing or two about computers from developing yet faster ways to transmit data over the Internet.

Take last week’s announcement from the chaps at the Department of Computer Science at the North Carolina State University (NCSU), for example. Those guys have now developed a new Transmission Control Protocol (TCP) – dubbed BIC-TCP – that will allow the Internet to operate at speeds approaching 10Gbit/sec – roughly 6,000 times that of DSL and 150,000 times that of current modems.

According to Dr. Injong Rhee, associate Professor of Computer Science at NCSU, the existing TCP – developed in the 1980s to allow two host computers to guarantee data delivery between them – just isn’t appropriate for today’s high speed networks. That’s because it was created back when Internet speeds were much slower and bandwidths much smaller.

And he is not the only one to have figured as much. Many other researchers have also seen the limitation of the existing TCP which needs to acknowledge packets that are sent over the network and thereby sets a limit to network throughput because it is constrained by what the researchers call ‘Round Trip Time’ constraints. And they, of course, have developed their own TCP solutions too.

Out of all of them, though, the new North Carolina TCP seems to have fared the best in a recent comparative study run by the Stanford Linear Accelerator Center (SLAC). In fact, the BIC-TCP topped the rankings in a set of experiments that determined its stability, scalability and fairness in comparison with other protocols. The study tested six other protocols developed by researchers from schools around the world, including the California Institute of Technology and the University College of London.

None of this news is going to be discussed at any great length with my colleague Mr. Ford, however. Because if I do tell him, he will then undoubtedly point his folks’ low speed iMac in the direction of NCSU’s Internet site to find out more about this rather whizzy technology.

And when it takes him 5 minutes to download the page that describes how others might soon be visualising satellite images or climate models from their computers at 10 Gbit/sec, he’s going to be a very unhappy bunny indeed. So this week, for his own mental well-being, we’re crossing him off the distribution list for the newsletter.

A full copy of the comparative study of the various TCP stacks and how they performed is available at:

http://dsd.lbl.gov/DIDC/PFLDnet2004/papers/Bullot.pdf