MODELING AND ANALYSIS OF MULTIHOP AND PRIORITY RANDOM ACCESS COMPUTER NETWORKS

We present a new priority protocol for CSMA/CD networks which allows high channel utilization and low delay for the priority traffic, and a model for multihop packet radio networks which has a product form solution. Local area networks operating under the Carrier Sense Multiple Access protocol (CSMA/CD) are widely used and are being considered for more demanding applications than simple data communications. The increased use of CSMA/CD networks in these environment calls for some type of priority function; otherwise the delay requirements of new data classes may not be met. We have developed and analyzed a new Priority CSMA/CD (P-CSMA/CD) protocol which allows high channel utilization and low delay for the priority traffic. We have found closed-form expressions for the throughput, the distribution of the number of busy users, and the expected delay. We have also found necessary and sufficient conditions for network stability and exact bounds on the network performance. The Packet Radio Network provides an alternative to wire or fiber optic cable when it is either uneconomical or actually impossible to lay wires to form a network. For example, they have been considered for use in automated factories where, because of changing production layouts, wire networks are impractical. When packet radios cannot communicate directly, because of interference, range limitations, or natural obstacles, their messages must be relayed to the destination. This is called multihop communication. We have developed realistic new models for two- and three-hop packet radio networks and found a control policy (PFS-control) that yields a product form solution for these networks. The PFS-control is intuitively satisfying: the station balances the flow of packets across the network by varying its transmission rates. Since the network has a product form solution, we are able to find the joint queue length distribution in closed form and derive simple, exact expressions for the throughput and the expected delay. The PFS-control has a regular structure that may, when better understood, make it easy to find PFS-controls. Because the PFS-control greatly simplifies the performance analysis, it may eventually allow larger packet radio networks to be analyzed. There is also freedom left under the PFS-control policy which could be used for optimal throughput-delay control