Abstract

Packet or datagram based routing is widely used for the Internet mainly due to its robustness, ease of implementation and autonomous behavior. Hierarchical networks are necessary to provide scalability. Previous developments proposed for ATM technology to be the basis of packet switched routing protocols. This technology provides flows to reduce the number of times a packet's destination address must be looked up in a routing table thereby increasing the performance of the overall system. Basic methods to establish and maintain these flows have emerged. Clearly one cannot define a flow between every (source, destination) pair since the number of destinations (in the Internet) is prohibitively large. We provide a definition of the problem with optimality criteria, and an algorithm that autonomously establishes a dynamic hierarchy of flows over an arbitrary network according to traffic demand. If the network is of size n, our algorithm uses only O(n/sub 0/log/sub n0/n) labels at each router where n/sub 0/=e/spl ap/3. Packets are routed over a /spl lambda/-segmented flow such that the label table is accessed at most /spl lambda/=log/sub n0/n-1 times. We show that the total table look-up complexity, O(n0log/sub n0//sup 2/n), is a factor /spl Theta/(log/sub n0/log/sub n0/n) times the optimal, which makes substantial improvement of other results known to us. Our hierarchy of flows naturally exists transparently on top of the underlying hierarchical address space (IP) and requires no additional information.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call