Simulation Results for Multi-Class Multi-Server Queueing Systems with Cross-Training
datasetposted on 19.07.2019 by Andrei Sleptchenko, Hasan Turan
Datasets usually provide raw data for analysis. This raw data often comes in spreadsheet form, but can be any collection of data, on which analysis can be performed.
The JSON file contains a number of cases for multi-class multi-server systems with partial class-server assignments (cross-training). For each recorded systems all possible (feasible) class-server assignments were generated, and simulation results were recorded. The simulation results (for each assignment) include: (i) marginal probabilities that each class has the corresponding number of items in the system, (ii) class-server utilizations (utilization of server time capacity), (iii) class server distribution with percentages of class flows assigned to each server. Maximum numbers of servers and classes were limited to 6. Please note that NOT ALL possible combinations for numbers of servers and classes were generated, due to the exponential growth of possible class-server assignments. However, for each presented case (with certain numbers of servers, classes, and certain arrival/service rates), ALL possible class-server assignments were generated and simulated. The presented results allow easy analysis and optimization of systems where such queue occurs (call-centers, production facilities, maintenance systems) and benchmarking with possible optimization algorithms for class-server assignments.