USING SIMULATED DATA TO TEST FOR THE EFFICIENCY OF MATCHMAKING ALGORITHMS (FIRSTMATCH, MINIMUM DIFFERENCE AND MINIMUM DISTANCE ALGORITHMS)

¹Odiketa Juliet C. and ²Aremu D. R.

¹Department of Computer Science, The Federal Polytechnic, Idah, Kogi State

²Department of Computer Science, University of Ilorin, Kwara State

E-mail: chiomie@yahoo.com, draremu2006@gmail.com

ABSTRACT

Resource allocation is useful in grid like environments where access to grid based components is needed. Resource allocation is needed when there are many tasks, which are not accomplished due to the lack of resources. Similarly task allocation is needed when there is abundance of idle resources with no tasks to use them up. Grids are a set of interconnected nodes. Agents run on these nodes.  Nodes could have either resources or tasks available with them and allocation of resources and tasks needs to be done using agents. In this situation sub-optimal allocation of resources and tasks are encountered because some nodes have too many tasks and some nodes may have extra resources. The question is how resource allocation and task distribution can be carried out in a grid based environment. There is need to do task re-distribution and proper resource allocation so that one node will not be overburden with many tasks. Matchmaking is done to rebalance the workload. This paper uses simulated data generated using Java programming Language to determine the efficiency of matchmaking process using three matchmaking algorithms (first match, min difference and min distance) based on some optimization criteria that are suitable for solving this allocation problem. These algorithms are simulated to determine their efficiency in terms of match time, number of matches, task usage and resource utilization. The result shows that first match function is the simplest and fastest algorithm and the minimum difference function is more efficient in terms of resource utilization.