Powerful algorithms for queueing simulation (PAQS).

Abstract

Queueing theory models have been widely used in several industries, such as manufacturing, maintenance, computer systems, transportation, telecommunication, etc., in order to build high-performance systems that respond to customer’s demand in a reasonable time and cost-efficient manner. This research project addresses basic aspect of queueing analysis related to simple single-node systems, where customers arrive to a multi-server system according to a known distribution, wait in line, if needed, get served based on another well-determined distribution in a first-come, first-served manner, and then depart the system. Our proposed PAQS software aims to implement efficient and effective algorithms for simulating single node queues generally denoted as G-G-s. PAQS is sought to utilize state-of-art technique for generating the arrival and service time variates and determining the run length necessary for an accurate output. In particular, we improve the efficiency of the simulation via a fast sorting technique. Our simulation methodology is suitable for analyzing high-variability queues that have been recently observed on many internet servers.