APPROXIMATE PERFORMANCE ANALYSIS OF REAL-TIME TRAFFIC OVER HEAVILY LOADED NETWORKS WITH TIMED TOKEN PROTOCOLS

This paper studies timed token protocols with respect to real-time pac ket traffic in local area networks (LAN's), such as FDDI and Token Bus , employed in distributed control systems, Typically, in such systems, three classes of packet traffic are encountered, The first class cons ists of packets cyclically generated by data acquisition tasks. The se cond traffic class is represented by packets generated in a random man ner by control tasks and sporadic events. Finally, the third traffic c lass represents nonreal-time packet streams such as, for example, file transfers. To evaluate protocol performance, three performance measur es are taken into account with respect to randomly generated realtime traffic: the mean waiting time, the blocking probability, and the prob ability that accepted packets will wait for sen;ice no longer than a s pecified time limit. In order to determine the last performance measur e, a two-moment approximation of the waiting time distribution is appl ied. All three performance measures are evaluated at the beginning of the heavy network load region. Two examples of numerical calculations compared with computer simulations done for FDDI-II and Token Bus netw orks are given.