The Dynamic Delay Scheduling Algorithm Based on Task Classification

MapReduce is currently the most mainstream parallel computation model to deal with large-scale datasets, and as for a crucial module of MapReduce, the task scheduling has important research meanings. However, there are two mainly problems with existing delay scheduling algorithm: (1) the theoretical assumption of that all tasks are short tasks is limitary, and when nodes process tasks of different lengths, performance of this algorithm will decline, (2) all tasks are based on permanent waiting time and that cannot meet the needs of different users. In order to solve these two problems, this paper comes up with The Dynamic Delay Scheduling Algorithm Based on Task Classification (TCDDS). The TCDDS algorithm divides all tasks into different categories by using fuzzy mathematics and gives different waiting time to different categories tasks, thus the response time of the whole job will be reduced and the performance of this algorithm will be improved.