CA*NET

NSFNET Transition Update -- CA*NET

Eric Carroll

CA*NET

Table of Contents

Slides from this Talk

CA*net Transition

Eric M. Carroll
University of Toronto
CA*net Networking

Timeline

Mar 1994
the change is coming
May 1994
assess the options
Jul 1994
the penny drops
Aug 1994
the Decision
Sep 1994
Planning, contract - Can we make 1 Nov?
Oct 1994
Nope. Contacts, setup, ship.

Timeline

Oct 1994
Colocation construction, circuit installation
Nov 1994
Construction complete, links up, peering starts
Dec 1994
Transition attempts, cutover
Jan 1
Transition Complete

Stan Barber's Notes

CA*NET is a cooperative of 13 providers in Canada. CA*NET had to get off the NSFNET first. The University of Toronoto is the contractor that operates CA*NET. CA*NET did not have the option of getting notifications for the shutdown of the ENSSes they were connected to or establishing the dates for such a shutdown.

CA*NET chose not to do peering over international circuits because it was going to be a real problem. It was VERY complex using invese muxes and many other parts. CA*NET wanted theses parts all under the control of one providers (more or less). It turns out that even this was not that simple, but it is much simpler.

CA*NET has a RR using the RA software. Eric is very pleased with how this works.

CA*NET did not really have the option of running on both links (like the CoREN regionals do). CA*NET had problems with their routers, but Cisco and MCI both worked hard with CA*NET to resolve these problems. Transition was completed on January 1, 1995. CA*NET is very pleased in their connectivity to Alternet and MAE-EAST.

CA*NET has a CIX connection, but Eric believes that CA*NET does not get much value from that connection. CA*NET uses OSPF internally. CA*NET injects about 6Mb/sec into MCI.

CA*NET will add two links in Chicago and one each at the other locations.

Is using IMUX worth doing? There are certain challanges and Eric has not had good experiences with it. The load sharing works almost as well.

CA*NET generated three defaults and there is always a fallover available in case a default fails. CA*NET uses MEDs. IBGP is used, but reluctantly.

ix.on is an internet exchage point that is operated by the University of Toronto.


Copyright © 1995 Stan Barber. Reproduction with attribution granted.
Academ Consulting Services
P.O. Box 300481
Houston, Texas 77230-0481
Comments via email to www@academ.com