contributed articles

Doi: 10.1145/1538788.1538809

Network software adapts to user needs and
load variations and failures to provide reliable
communications in largely unknown networks.
BY eRoL GeLenBe
steps Toward
self-Aware
networks
tHe inForMation needed

to route packets in large
networks and in networks in which nodes join and
leave the network frequently or move in and out
of wireless range of each other can change more
frequently than the rate routing information is
updated throughout the network. In such a system
it becomes necessary to allow individual nodes to
proactively discover the presence of other nodes,
links, and paths (as needed and on demand), leading
to the design of self-aware networks. here, I focus on
experimental and theoretical research concerning the
technical steps leading to these networks.

The Internet Protocol offers an orderly update of its status based on the shortest-path algorithm, ¹⁸ Distance Vector, ²² and Link State²³ techniques so routing algorithms operate seamlessly, despite changes in network topology and conditions. however, as computer networks become extremely large, the information available concerning the network state becomes uncertain. Link state changes

are more frequent in larger networks, increasing the overhead and delay due to updates throughout the network. Consequently, information about the network state, including connectivity, condition of nodes, traffic conditions, and quality of service (QoS), propagates more slowly than rate changes occur. The need to convey time-sen-sitive information (such as voice and media) also motivates investigation of routing techniques based on user requirements and the network’s instantaneous state. Thus it is preferable that nodes discover the network state autonomously, without having to rely on an overall scheme that updates routing tables systematically throughout the network. Information updates can be initiated by the nodes that need this information at the time it is needed, rather than throughout the network and when changes occur.

We use the term “self-aware network,” or SAN, ¹¹ for a system consisting of nodes that can join and leave the network autonomously and discover paths when the need to communicate arises. The nodes in a SAN should sense the status of other nodes, links, and paths, including traffic level and congestion, so as to update their own relevant information about the paths they need to use, based on criteria specific to their own needs. Each connection may then use paths that optimize the connection’s own QoS criteria, rather than a common criterion (such as the shortest path) for all connections. These needs might include user QoS requirements, or performance, reliability, security, defense against attacks, ^{9, 24} and power utilization. ¹² A SAN can be a wired, wireless, or a peer-to-peer system. A wireless ad hoc network is a practical example of a SAN that responds to time-varying conditions related to the mobility of nodes and changes in the conditions of wireless links (such as noise and physical obstructions). Networks that must operate autonomously and remotely (such as sensor networks) also benefit from self-aware capabilities.

Research on effective SAN architec-

DRAWING B Y EMMA McNALLY