Approximate MVA for client-server systems with nonpreemptive priority

A new approximate algorithm for the Stochastic Rendezvous Network (SRVN) model with nonpreemptive priority scheduling is introduced in this paper. SRVN is a performance model for client-server systems with synchronous communication which is different from Queueing Network models in two ways: it allows for nested services, and offers two phases of service (the first executed while the client is blocked and the second in parallel with the client). Earlier SRVN solutions have used a kind of approximate MVA based on heuristic assumptions to determine the queues properties at the instants of service request arrivals. More recently a new strategy called "Task-Directed Aggregation" (TDA) was introduced for the derivation of the arrival-instant probabilities equations for FIFO servers. The present paper applies TDA to nonpreemptive priority scheduling, thus demonstrating the value of this new strategy for models with no product-form solution. Experimental results show that the accuracy of the algorithm is good if the server is not saturated, and if a reasonable fraction of the load is available for the low-priority clients. The accuracy of the algorithm is consistent with results known for QN priority approximations.<
>