IP: the freeway is jamming up [ with an editorial comment by

[Dave Farber <farber () central cis upenn edu>]

Editorial comment:


One of the marks of success is growing pains. It hits companies and
societies. The net succeeded. With that success came the loads and the
normal delay in engineering the growth of the infrastructure. The idea that
it is the result of the privatization of the net in and of itself is wrong.
The NSF provided negligible network management in the technical sense of the
word and little in other senses. As soon as CSNet allowed "commercial"
traffic on the die was cast.


I grew up in the telecom business. We had mechanisms for measuring the grade
of service offered to the customer and standards which had to be met else
our tarrifs suffered. The Internet has none of this. There is NO EFFECTIVE
way of telling where congestion is and why, there is no effective way in the
short run to force the deployment of additional facilities in backbone
routes. There is precious little way of recovering the costs of doing that
for many of the intermediate carriers. Saying that somehow government and
the private sector should somehow coordinate is a worthwhile statement but
darn hard to do anything with in the presence of substantial profit and
business building  motives unless that translates to let the FCC do it (GASP!!)


What can be done. First what is the real problem? Can we find out where the
bottlenecks are (we do know how to do net measurement -- don't need research
for thatone ] and try to convince the responsible carriers to remove them
(after all the user's organizations do pay $s to carriers (but maybe to
different ones). The best long term bet is to create a mechanism to allow
the distribution of money to create incentives for the proper engineering of
the network.


Just for the record, the latency of the net is one small part of more
serious future problems which will not be so easy to fix.


And finally I would recommend that the scientific community stop sounding
like someone took away their toys. They have a valid complaint as, I hope,
paying commercial customers. Time to recognize that. Much better PR.


Dave Farber




Scientists Complain of Internet Traffic
They say it is slowing research 




Jeff Pelline, Chronicle Staff Writer 


The Internet is in a ``disastrous state'' and is impeding some high- level
research for scientists at big national laboratories, a group of researchers
has charged in a memo. 


The memo, complaining about ``very slow'' access and ``catatonic''
connections, was sent to the Federal Networking Council in Washington, a
government group that looks into such matters. The council's advisory
committee met earlier this week and discussed the growing concern of
Internet traffic jams. 


The memo, obtained yesterday, is one of the most heated complaints yet about
the Internet, the global computer network that is experiencing growing pains
as more and more people log on. The network, once the domain of scientists
and researchers, is becoming a mass medium. 


``The postdocs and students, located at our experiments, have found access
very difficult over the Internet during this last year so that their
effectiveness has been much decreased,'' the memo said. ``The connection is
always very slow, and often becomes completely catatonic for minutes at a time. 


``It is essentially impossible to log in and control execution of jobs
across the network; transferring graphics information is out of the question.'' 


The research involves high-energy physics experiments at the Stanford Linear
Accelerator Center at Stanford University, Fermilab near Chicago, Cornell
University in upstate New York, and the European Nuclear Energy Center in
Geneva, Switzerland. The projects are funded by the U.S. Department of
Energy and National Science
Foundation. 


The group said the ``degradation in Internet performance'' coincided with
last year's privatization of so-called Nsfnet, formed in 1986, as the
Internet's main backbone. The government's withdrawal from Internet
management has been controversial because there is no coordinating effort to
help guard against traffic jams, critics charge. 


``We are in substantial agreement that there is need for a coordinating
operation to improve the cooperation among all the private and public
networks that are important to the research community,'' the group said. 


Federal Networking Council advisory members yesterday were reluctant to
discuss the late March memo publicly, citing its sensitivity. But they said
they would continue working with the private sector to fix the problem. 


Internet ``brownouts'' are not just affecting high-level researchers. In a
letter to its subscribers, Internet service providers such as San Jose-based
Netcom On-Line Communications Inc. recently have conceded that the Internet
is suffering growing pains.