On 22 Sep 2004, at 08:13, Richard Naylor wrote:
Mind you as we all step up to 10GE and the 100GE starts to appear this will be trivial.....
In some parts of the world networks have been running at over 10G for years, and this kind of traffic load *is* commonplace and trivial. You don't get to sell residential 50Mbit/s VDSL service as a standard product without building a backbone that can carry heavy traffic loads reliably. The usual response to the argument that experimental networks are important for high-speed experimentation is that commercial networks are now, and have been for some time, much faster than the experimental, academic networks. The periodic i2 "internet land speed record" press releases seem amusing when you compare them to production networks in the US and Europe (never mind South Korea) which need to load-balance multiple STM-64s in order to keep traffic flowing over the Atlantic. The key question is why it is considered more cost-efficient to subsidise a restricted-access network than to allow the universities to buy equivalent access to the commercial Internet, and to run their experiments across that. (And if the answer turns out to be "the price is too high, because there's no damn competition" then perhaps that tells you something.) Joe