Split agent-based routing in interconnected networks

Abstract

Adaptive behaviour of swarm-based agents (BT Technol. J. 1994; 12:104–113; AAMAS Conference '02, Melbourne, Australia, Month 1–2, 2002; Softcomput. J. 2001; 5(4):313–317.) is being studied in this paper with respect to network throughput for a certain amount of data traffic. Algorithmically complex problems like routing data packets in a network need to be faced with a dynamically adaptive approach such as agent-based scheme. Particularly in interconnected networks where multiple networks are participating in order to figure a large-scale network with different QoS levels and heterogeneity in the service of delay sensitive packets, routing algorithm must adopt in frequent network changes to anticipate such situations. Split agent-based routing technique (SART) is a variant of swarm-based routing (Adapt. Behav. 1997; 5:169–207; Proceedings of 2003 International Symposium on Performance Evaluation of Computer and Telecommunication Systems—SPECTS, Montreal, Canada, July 20–24, 2003; 240–247.) where agents are split after their departure to the next node on a hop-by-hop basis. Packets that are delay sensitive are marked as prioritized which agents recognize-as being a part of a packet- and try to influence the two-way routing tables. Thorough examination is made, for the performance of the proposed algorithm in the network and the QoS offered, taking into account a number of metrics. It is shown that the split agent routing scheme applied to interconnected networks offers a decentralized control in the network and an efficient way to increase overall performance and packet control reducing at the same time the packet loss concept. Copyright © 2004 John Wiley & Sons, Ltd.