Adaptive Connection Admission Control for Mission Critical Real-Time Communication Networks.pdf

Adaptive Connection Admission Control for Mission Critical Real-Time Communication Networks B. Devalla, A. Sahoo, Y. Guan, C. Li, R. Bettati and W. Zhao Department of Computer Science, Texas AM University, College Station, TX 77843-31 12 Abstract In this paper, we report our work on adaptive connection admission control in real-time communication networks. Much of the existing work on connection admission control (CAC) specifies the QoS parameters as fixed values and does not exploit the dynamic fluctuations in resource availability. We take an innovative approach: First, we allow an application to speciJj, QoS over a range, rather than fixed values. Second and more importantly, we design, analyze, and implement CAC modules that, based on QoS specified over a range, adaptively allocate system resources to connections. Delay analysis is an integral part of connection admission control. Our adaptive CAC uses an efficient delay analysis method to derive a closed form solution for end-to-end delay of messages in a connection. Our adaptive CAC improves system pe$ormance in terms of the probability of admitting connections and the QoS offered to the connections. 1 Introduction In this paper, we report our work on an adaptive approach to connection admission control for providing effective and efficient connection admission control for mission critical applications distributed over real-time communication networks. These applications typically consist of a set of tasks executing on different hosts, exchanging messages over a high-speed network to co- operatively accomplish a common mission critical objective. Examples of such applications include supervisory command and control of defense systems, manufacturing plants, etc. The success of a distributed mission critical application thus crucially depends on the ability of the underlying network to guarantee upper bounds on message transfer delay. Connection-oriented communication is wel