|There are three ways of
looking at research and education networks like Orion, CA*net4,
Internet2: One is as a cheaper
alternative to a commercial service.
Where the activities of the Internet go on as usual allbeit with a
smaller universe of players. The
second is as a place for experiment to get a glimpse of what a possible
future would look like. It becomes a experiment with new ways to use the
network with the hope that the understanding thus obtained will help shape
the future of the real and universal Internet. Of couse the third way of
looking at the network is as a research entity in and of itself. Are there interesting ways to move bits
around, faster, more predictably and in greater volumes.
|Looking at the R&E
networks as a cheaper and better alternative to commercial networks has some
built in limitations. The major one is
the limited reach of the network. While
organizations like ORION and CANARIE keep on touting the great reach of their
networks, when compared with the commercial Internet the reach is
paultry. Even as 100s of schools are
added we have the government claiming that all (100s of thousands) schools
are connected to the Internet. Of
course the commercial Internet is often too slow or variable to do things
like video conferencing or other time sensitive operations. I think that Skype is showing
otherwise. I regularly use VC from
home and I've seen commercial connections into the PCO that are better than
we often get over the R&E networks.
And if you really want good audio and video just rent 3 to 6 ISDN
lines and you'll probably do much better than any H.323 video
connection. So in this scenerio are we
just claiming to be cheaper? (and perhaps a little easier -- it is a real
pain to get those ISDN lines).
Consider the subsidies that have gone into making those lines
cheaper. $110M for CANARIE over 5
years, ??? for ORION, ??? for the internet2.
Are they really that much cheaper?
|When we view the network as
a testing place for new applications we pull further away from the existing
commodity network. This is ideal for
projects that include only a few end-points but lots of time-sensitive data. Due to the limited connectivity of the
R&E networks it is a harder sell for an application that is stressed by
having millions of users of relatively small portions of the network. So when physicists need to work out how to
efficiently send huge amounts of data from a collider in Switzerland to a
computational farm in Calgary and then back to storage in France, the R&E
networks are the obvious first choice.
When a researcher in Edmonton is trying to work out just how rooms
that are sensitive to and react to their occupants located in rooms Edmonton
and Victoria and then write a development language for making such scenerios,
again the R&E networks are perfect.
If you want to stress test a new Voice-over-IP system by providing a
free trial version to millions of users, you've probably come to the wrong
|The last view abstracts the
applications into entities that need to move bits of data from one place to
another (or multiple places to multiple places) within certain time
contraints and variability. This is
often a "crash and burn" network that carries no
"production" traffic. It is
research into the ways that networks could work differently from how they do
today. Some of this occurs in the
existing R&E networks and you can see some of it with the experiments in
setting up "light-paths" between two points or between two
cooperating networks. Typically the applications are secondary here and may
be only glimpsed at since it isn't quite clear yet what the opportunities are
that will be enabled by the new networking technologies that are being