[Help] Last update at http://inet.nttam.com : Mon Aug 7 21:39:35 1995

Abstract -- The Routing Arbiter in the Post-NSFnet Service World Network and Application Engineering Track
N2: Routing and Addressing

[Previous] [Table [Next]
[Paper [Paper

The Routing Arbiter in the Post-NSFnet Service World

Manning, Bill ( bmanning@isi.edu)


In the wake of a fundamental change in the US-based Internet architecture, the Routing Arbiter (RA) is providing a stabilizing influence by maintaining coherent routing on a global scale. This is provided through the development and promotion of a globally coordinated Internet Routing Registry (IRR), the use of Route Servers (RS) at the National Science Foundation (NSF) Network Access Points (NAPs) and other exchange points, and the development and promotion of new protocols, tools, and operational practices that will facilitate the continued growth of the Internet.

The RA is addressing the general problem of route construction and dissemination. Work is in progress on the next phase of a syntax for the specification of route selection criteria, the development of new routing protocols (IDRP, SDR, ERP), and exploring the limits of existing routing engines. The start is the gobally coordinated IRR, where users can exercise the new syntax and get direct feedback on their expressed preferences for route construction. The IRR has replaced the old Policy Routing Database (PRDB) that was maintained by MERIT. The IRR is a loosely coupled series of RIPE-181 databases with some extensions to support better synchronization.

The syntax is an outgrowth of the RIPE PRIDE tools. It has been expanded to support new features and protocols which will be needed to support new features such as resource reservation and source routing. There are also features to facilitate "what-if" trials without actually changing the entries in the IRR.

The IRR will feed data to the Route Servers, which will disseminate customized "views" of the Internet to participating RS users, based on the preferences that are registered in the IRR.

By concentrating this large amount of information in the Route Server, we have been able to test the limits of existing routing engines and have identified a number of possible approaches that will be needed to cope with the growth of the Internet. All of these provide additional levels of abstraction, which is a proven technique to facilitate scaling.