The topology component of protocol performance

Abstract

The paper demonstrates that the performance of MAC-level protocols is not very meaningful, when considered alone. The true object of performance analysis should be the pair (T,P), where T stands for network topology, and P for MAC-level protocol. The paper examines the performance of two MAC-level protocols: DQDB and LCSMA-CD/U/P. As both of them are capacity-1 protocols, they achieve the same maximum throughput; thus, the focus is on the packet access time and on fairness. The investigation shows that either protocol can be shown as greatly unfair or totally fair, depending on the topology that it operates on. As a byproduct, it is shown that the UU-BUS topology is quite interesting and merits further studies.<>