基于支持向量机的任务调度模型

The Tasks Scheduling Model Based on Support Vector Machine

云计算框架大大改进了并行算法的实现难度,但是大部分算法有其局限性.介绍了MapReduce(映射化简)的基本实现原理和调度模型的缺陷,提出了基于支持向量机的的MapReduce进化算法,并给出了基本模型及实现.运用Hadoop云计算平台进行了仿真验证,实验结果表明,基于支持向量机的MapReduce计算框架在候选云节点的调度分配的准确性上有明显提高,并且加快了数据迭代的效率.

The cloud framework reduced the difficulty to realize parallel algorithm. But most of the algorithms have the defects. The fundamental and scheduling model of MapReduce are introduced.The evolving algorithm is proposed based on Support Vector Machine.The basis model and realization are built.By simulation realization on Hadoop cloud computing platform, compared with the traditional scheduling algorithm, experimental results show the new theory based on the SVM improve the accuracy of cloud candidate point scheduling and improve the speed of data iteration.