Bell Labs/Columbia U. Bell Labs/Columbia U. Bell Labs/Columbia U. Resource reservation needs to accommodate the rapidly growing size and increasing service diversity of the Internet. This memo first defines the scaling problem in today's Internet backbone, and briefly discusses several existing resource management approaches. Then we will present a distributed approach and introduce a protocol, called the Border Gateway Reservation Protocol (BGRP), for inter-domain resource reservation that can scale in terms of message processing load, state storage and control message bandwidth. The main idea of our approach is to build a sink tree for each domain network. Each sink tree aggregates reservations from all data sources in the network. Sink tree initiation, maintenance and termination involve only backbone border routers. Within each domain, the network service providers manage network resource and direct user traffic independently. At the border routers, the service providers can use BGRP to setup domain-level reservation trunks base on bi- lateral agreement. Since routers only maintain the sink tree information, the total number of reservation states at each router scales, in the worst case, linearly with the number of domains in the Internet. For bandwidth reservation, BGRP relies on differentiated services for data forwarding. As a result, the number of packet classifier entries is small. To reduce the protocol message traffic, routers may reserve domain bandwidth beyond the current load so that sources can join or leave the tree or change their reservation without having to send messages all the way to the root for every such change.