Using Simulation Modeling to Estimate Time Characteristics of a Distributed Computing System

Abstract

In present time, distributed computing systems with parallel data processing are widely used. During development of such distributed systems problem of analyzing workload and structure parameters impact on system performance is presented. Special attention have to be pointed towards evaluating time spent by requests in queues and effects of prioritization algorithms on it.To evaluate computing system’s time characteristics different modeling methods can be used, most effective of which is simulation modeling. It is important, however, not to just conduct modeling, but to make sure that results are accurate.The purpose of this research is acquiring time characteristics of distributed computing system with use of simulation modeling with certain accuracy and estimation of modeling time required.To achieve this task simulation model was developed, which represents functioning of distributed computing systems in set period and takes into account prioritization algorithm, and workload model, which corresponds with selected computational system class. The GPSS language was used to implement the model. Based on the obtained simulation results, it is possible to compare the impact of different algorithms for processing priority requests depending on the parameters of the workload and structure. To evaluate accuracy of results regenerative method of model analysis was used, which is based on concepts of regenerative process and regeneration points.In this paper results of average time spent by requests in queue in dependence with used prioritization algorithm research are presented, as well as time required to achieve desired accuracy. Due to this, it is possible to reasonably recommend the most advantageous of the considered prioritization algorithms for implementation in the developed computing system, depending on the requirements.The presented approach to the analysis of the functioning of distributed computing systems and the assessment of their temporal characteristics will also undoubtedly be useful in the educational process for teaching students in the courses “Computer systems” and “Simulation modeling”.

Keywords

• распределённые вычислительные системы
• имитационное моделирование
• регенеративный метод
• gpss