The modeling and simulation of a network controller based on router queue-size predictions

Abstract

While the current routing and congestion control algorithms in use today are often adequate for networks with relatively static topology and relatively lax quality of service (QoS) requirements, these algorithms may not be sufficient for military networks where a strict level of QoS is required in order to achieve mission objectives. Current technology limits a network’s ability to adapt to changes and interactions, and this often results in sub-optimal performance. This article develops a network controller that uses outbound router queue size predictions to optimize computer networks. These queue size predictions are made possible through the use of Kalman filters to detect network congestion. The premise is that intelligent agents can use such predictions to form context-aware, cognitive processes to managing network communication. The system shows great promise when modeled and simulated using the NS2 network simulation platform.