DefinePK

Abstract

Applications to be solved in parallel fashion use distributed environments. Generally, these kinds of applications are large computational projects of high complexity. The distributed environment is a massive pool of heterogeneous resources which could be utilized by the applications. Grid computing environment is one such environment that aggregates networks, computers, servers, applications, programs and the users. The co-ordination of resources is the mandatory aspect of distributed computing environment and this is achieved by efficient scheduling. An effective scheduling is very significant as it encounters high performances. A high quality scheduling is meant for its low cost, earlier completion, accurate results; and obviously that might be the key expectation of the clients too. These can be met only when resources are properly utilized by the applications. Parallel applications are illustrated mathematically as Directed Acyclic Graphs (DAGs), which help us to understand the dependencies and data mobility in the real applications. In this article, we present a scheduling algorithm that motivates quick and quality schedules, which in turn encourages reduction of makespan time, increases speed up and best rate of result recurrences. The algorithm called PD-SECT (Priority on Dependency and Start, Execution, Communication Time) is based on list heuristics. The algorithm accomplishes the stages of (i) task selection, (ii) resource selection, (iii) mapping the chosen task and resource. The priority in selecting the tasks and resources is based on the following criteria, (i) task selection is based on population of inter dependent tasks; (ii) resource selection is based on resource availability with respect to the start and execution time. The proposed algorithm does not encourage insertion policy as the tasks are encouraged to pack consecutively without idle slots. On comparisons with algorithms of its kind, this algorithm arrived at best results when scaled from smaller to bigger task graphs. The algorithms are implemented in GridSim simulator, which gives a feel of real time environment.