The following text is copyright 1999 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

Convergence from the other side

by Scott Bradner

At some random networking conference over the last year or so I signed up to get one of the magazines designed for the traditional telephone companies. I've now started getting it and I've been surprised at the amount of familiar information in it.

"America's Network" bills itself as covering "technology for the public network since 1909." It is definitely a magazine for telephone carriers and their suppliers. It has articles about telephone carrier topics such as reusing the old digital loop carrier cabinets, 100,000 of which are scattered around the landscape, and telephone billing systems. But each of the issues I've received since my subscription started a few months ago has had a number of Internet-related articles. For example, the November 15th issue has articles on the IETF's MPLS technology alongside a research report on the future of wireless telephones.

MultiProtocol label switching (MPLS) is in the final stage of being approved by the IETF as a Proposed Standard. Its origins were in Cisco Systems Tag Switching and was initially targeted at providing Internet service providers (ISPs) the ability to do traffic engineering. Traffic engineering in an ISP is the ability to direct IP traffic through paths in the ISP backbone that normal IP routing would not have chosen. For example ISP traffic from Boston to San Francisco might normally be routed through Chicago. If the ISP links through Chicago get overloaded and the ISP has excess capacity in a fiber link through Cincinnati, the ISP can use MPLS to direct the Boston to San Francisco traffic through the Cincinnati link. This function is just what some ISPs who have an underlying ATM network have been doing with ATM virtual circuits. MPLS allows non-ATM-based ISPs to do traffic engineering.

Later on it became clear that this same traffic engineering could be used to help provide better quality IP service for specific applications if the determination of what MPLS path to use was based on what application was being run rather than what city the traffic was coming from. Most US ISPs are focusing on the use of MPLS for non-QoS traffic engineering.

The articles in America's Network on the other hand focus on the QoS aspects. The same technology being looked at from a different vantage point. Most of the IP related articles in the magazine are from this different vantage point, a telephone company vantage point. More and more telephone people are participating in the IETF so some of this other view is now being incorporated into the IETF work. But at times this can be a very different viewpoint indeed since the architectural and management assumptions that underlie the phone networks and the Internet are so very different. It will be interesting to see if we can keep true to the Internet model while learning from the phone input.

disclaimer: Harvard has made a science of having different management assumptions for each of its schools but the above observation is mine.