Grid Scheduling

Distributed computing is a new and promising technology, which aims to mobilize additional computational resources for scientific computations. This can help to manage huge computational tasks, which usually arisen in the areas like astronomy, meteorology, bioinformatics, etc. From the other hand, the participating in computational grids helps providers to increase an outcome from their computing facilities.

However, generally users and providers are geographically distributed; they belong to different organizations with different goals, policies and requirements. Moreover, the different participants have insufficient information about each other.

The effective orchestrating of such heterogeneous community is the duty of different middleware (broker) systems. These systems are aimed to perform various tasks, such as collecting information about participants, bringing different requirements to common standards, carrying negotiations, monitoring queues, security issues, etc.

Among the others, one of the most important tasks of any grid coordinator is an effective allocation of jobs to available resources. This is usually done by Grid Scheduling systems. Grid Scheduling (in general case) is known to be hard to solve NP-complete problem. Although different optimization heuristics might be applicable here, not all of them satisfy particular Grid Scheduling requirements (such as necessity to operate in fully-automated mode). Our current project is aimed to investigate the performance of different Grid Scheduling algorithms and find the most suitable one for the practical use.

We have evaluated our algorithms on benchmark instances, courtesy provided by Rizos Sakellariou and Viktor Yarmolenko from The University of Manchester .

Grid Scheduling Datasets
3 datasets provided by Victor Yarmolenko and Rizos Sakellariou
Instance Number of jobs Number of machines Workload Download
YAR-64-110 351 64 110% Yar-64-110.txt
YAR-64-100 340 64 100% Yar-64-100.txt
YAR-64-90 293 64 90% Yar-64-90.txt
The description of the dataset format is here: jobs_READ_ME_1ST.txt (Copyright V.Yarmolenko)

My E-mail:
Worktime Phone:   (0)115 846 8376
Address:          Yuri Bykov
School of Computer Science and IT
the University of Nottingham
Jubilee Campus, Wollaton Road
Nottingham NG8 1BB
United Kingdom

Last modified: 14 March 2009