CASPIAN JOURNAL

MANAGEMENT AND HIGH TECHNOLOGIES

Deterministic simulation model of clusters of a Grid-system for comparison of heuristics for task distribution

Read Gaevoy Sergey V., Al-Khadsha Fares A.Kh., Fomenkov Sergey A., Lukyanov Viktor S. Deterministic simulation model of clusters of a Grid-system for comparison of heuristics for task distribution // Caspian journal : management and high technologies. — 2014. — №2. — pp. 148-157.

Gaevoy Sergey V. - post-graduate student, Volgograd State Technical University, 28 Lenin av., Volgograd, 400005, Russian Federation, gaevserge@mail.ru

Al-Khadsha Fares A.Kh. - post-graduate student, Volgograd State Technical University, 28 Lenin av., Volgograd, 400005, Russian Federation, alhadsha@mail.ru

Fomenkov Sergey A. - D.Sc. (Engineering), Professor, Volgograd State Technical University, 28 Lenin av., Volgograd, 400005, Russian Federation, saf@vstu.ru

Lukyanov Viktor S. - D.Sc. (Engineering), Professor, Volgograd State Technical University, 28 Lenin av., Volgograd, 400005, Russian Federation

In paper an attempt to predict job execution on a Grid-system have been made. The Grid-system consists of computing clusters. To achieve this goal we have developed a deterministic simulation model. The cluster system is represented as a queueing system (QS) that executes incoming jobs. Each job has following parameters. The width is an amount of faculty units being required to execute the job. The length is the time the job needs to be executed. Square (difficulty) is a product of the length and the width. Unlimited queue is assumed because of the requirement to execute all jobs. That approach allows us to determine average queue length, average waiting time of the job etc. Now we investigate a method of job distribution between clusters. It is assumed that a task is executed by a single cluster. Each cluster of the GRID-system has his own jobs incoming from users, departments, some organizations etc. The connection allows clusters to execute jobs of each other. That policy makes a criteria necessary. And we need to answer the question В«What criteria must be used?В». There is neither an analytic solution nor a general answer. One has to model the actual clusters and try different job distributions. From those distributions one needs to select solutions that satisfy our requirements. The authors have created a deterministic simulation model of such calculating system, proposed some heuristics for the job distribution and compared these heuristic for a chosen example.

Key words: Грид-системы, брокер заданий, имитационное моделирование, детерминированная имитационная модель, распределение заданий, эвристики распределения заданий, системы массового обслуживания, вычислительный кластер, parallel workloads, Grid-systems, resource bro