Preview

Open Education

Advanced search

Job system generation in grid taking into account user preferences

https://doi.org/10.21686/1818-4243-2016-3-4-8

Abstract

Distributed computing environments like Grid are characterized by heterogeneity, low cohesion and dynamic structure of computing nodes. This is why the task of resource scheduling in such environments is complex. Different approaches to job scheduling in grid exist. Some of them use economic principles. Economic approaches to scheduling have shown their efficiency. One of such approaches is cyclic scheduling scheme which is considered in this paper.

Cyclic scheduling scheme takes into account the preferences of computing environment users by means of an optimization criterion, which is included in the resource request. Besides, the scheme works cyclically by scheduling a certain job batch at each scheduling step. This is why there is a preliminary scheduling step which is job batch generation.

The purpose of this study was to estimate the infl uence of job batch structure by the user criterion on the degree of its satisfaction. In other words we had to find the best way to form the batch with relation to the user optimization criterion. For example if it is more efficient to form the batch with jobs with the same criterion value or with different criterion values. Also we wanted to find the combination of criterion values which would give the most efficient scheduling results.

To achieve this purpose an experiment in a simulation environment was conducted. The experiment consisted of scheduling of job batches with different values of the user criterion, other parameters of the resource request and the characteristics of the computing environment being the same. Three job batch generation strategies were considered. In the first strategy the batch consisted of jobs with the same criterion value. In the second strategy the batch consisted of jobs with all the considered criteria equally likely. The third strategy was similar to the second one, but only two certain criteria were considered. The third strategy was considered in order to find the most favorable combinations of criteria couples.

The experiment showed that the second approach showed the best scheduling results. This may be explained by a smaller degree of competition for similar resources in the case of different criteria combination. Experiment results for criteria couples depend on the considered couple. The best scheduling results for this strategy were shown by criteria couples with job runtime and cost minimization criteria and job completion and cost minimization criteria.

The results of the study are a part of course “Computing Systems” at the department of Computing engineering at National Research University “Moscow Power Engineering Institute”. Further studies will be dedicated to the search of optimal correlation between user criteria in one batch.

About the Authors

D. M. Yemelyanov
National Research University “Moscow Power Engineering Institute”
Russian Federation

Candidate of Engineering Science, senior lecturer at Computer Engineering department,

Moscow



P. A. Potekhin
National Research University “Moscow Power Engineering Institute”
Russian Federation

graduate student at Computer Engineering department, 

Moscow



V. V. Toporkov
National Research University “Moscow Power Engineering Institute”
Russian Federation

Doctor of Engineering Science, professor, head of Computer Engineering department,

Moscow



References

1. Foster I., Kesselman C., Tuecke S. The Anatomy of the Grid: Enabling Scalable Virtual Organizations // International Journal of Supercomputer Applications. – 2001. – Vol. 15, N 3. – P. 200–222.

2. Garg S.K., Buyya R., Siegel H.J. Scheduling Parallel Applications on Utility Grids: Time and Cost Trade-off Management // Proc. of the 32nd Australasian Computer Science Conference. – Wellington, 2009. – P. 151–160.

3. Adaptive Computing on the Grid Using AppLeS / F. Berman et al. // IEEE Transactions On Parallel and Distributed Systems. – 2003. – Vol. 14, N 4. – P. 369–382.

4. Эволюция системы метакомпьютинга X-Com / Вл.В. Воеводин и др. // Вестник Нижегородского университета им. Н.И. Лобачевского. – 2009. – № 4. – С. 157–164.

5. Preference-Based Fair Resource Sharing and Scheduling Optimization in Grid VOs / V. Toporkov et al. // Procedia Computer Science. – 2014. – Vol. 29. – P. 831–843.


Review

For citations:


Yemelyanov D.M., Potekhin P.A., Toporkov V.V. Job system generation in grid taking into account user preferences. Open Education. 2016;(3):4-8. (In Russ.) https://doi.org/10.21686/1818-4243-2016-3-4-8

Views: 1069


Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.


ISSN 1818-4243 (Print)
ISSN 2079-5939 (Online)