ANALYTIC EVALUATION OF CONTENTION PROTOCOLS USED IN DISTRIBUTED REAL-TIME SYSTEMS

The probability of a station failing to deliver packets before their d eadlines, called the probability of dynamic failure, P-dyn, is an impo rtant measure for the communication subsystem of a distributed realtim e system. Another closely-related performance measure is the epsilon-b ounded delivery time, T-epsilon, which is defined as the least time ne eded to deliver a packet with probability greater than 1 - epsilon. Us ing P-dyn and T-epsilon, we comparatively evaluate four contention pro tocols often used in distributed real-time systems: (i) the token pass ing protocol and its priority-based variation (called the token schedu ling protocol), and (ii) the P-i-persistent protocol and a priority-ba sed variation thereof. The communication subsystem equipped with diffe rent contention protocols is modeled first as embedded Markov chains. Then, we derive the probability distributions of access delay, from wh ich P-dyn and T-epsilon can be calculated. The blocking probability, Q (i), can also be derived from the access delay distribution. These mea sures are derived first under the assumption of a single buffer at eac h station. The single-buffer model is then extended to the multiple-bu ffer case. The effects of buffer size on P-dyn, T-epsilon and Q(i), an d the performance improvement with multiple buffers are analyzed over a wide range of network traffic.