rcallon@PARK-STREET.BBN.COM (Ross Callon) (12/11/87)
Don; In response to part (3) of your message: The BBN Butterfly gateways do NOT use a simple hop count for routing decisions, for essentially the reason that you gave. Instead we use a cost which is administratively set as a function of the network characteristics. Each gateway-to-gateway link may in principle be given a different cost. For example, the cost (over the Arpanet) between gateways at BBN and MIT could be administratively set to be different than the cost between gateways at BBN and ISI. In practice, we tend to assign a single cost for all links over a single network. As an example, we currently use a single fixed cost for links over the Arpanet, which is greater than the cost which we use for links over LANs. The routes calculated by the Butterfly SPF algorithm minimize the sum of these costs, and do reflect differences in network characteristics. A link up/down algorithm is used to determine whether each gateway-to-gateway link is operating successfully, but as long as the link is up its metric is stable. We have been having good success with operational use of this algorithm since it was installed about a year and a half ago. The IS to IS routing proposal which is being progressed in ANSI X3S3.3 (and is being taken to the next ISO network layer meeting in Guernsey) also uses administratively set costs. Here again a link up/down protocol is used to determine whether a link is operational. You asked if there is a "published" method which is better than hop count. We are intending to publish a description of our SPF, but haven't done so yet. The ANSI proposal is semi-available. The most recent version was passed out at the last X3S3.3 meeting, and should be mailed out to the X3S3.3 mailing list soon. Ross