MobiWay应用中基于Hadoop的多目标多任务调度算法

针对Hadoop应用环境复杂、网络带宽等因素多变而影响调度算法性能的问题,提出适用于多任务多目标的Hadoop调度算法(MOSMT)。在分析已有调度算法工作原理的基础上,构建MOSMT算法的数学模型和调度策略;在负载模拟器中进行算法评估,并将MOSMT算法应用于MobiWay;对MobiWay应用中的MOSMT算法性能进行测试分析,以最少的资源和最低的时间成本完成任务的调度。通过与FIFO和Fair调度算法对比表明,该算法能够实现类似的功能,并且在处理多任务多目标时耗时更少,更为经济。

A MULTI-OBJECTIVE AND MULTI-TASK SCHEDULING ALGORITHM BASED ON HADOOP IN MOBIWAY APPLICATION

Hadoop application environment is complex and network bandwidth is changeable,which affect the performance of scheduling algorithm.To solve this problem,we propose a Hadoop scheduling algorithm(MOSMT)suitable for multi-task and multi-objective.Based on the analysis of the working principle of the existing scheduling algorithm,we constructed the mathematical model and scheduling strategy of MOSMT algorithm.Algorithm evaluation was performed in the load simulator,and the MOSMT algorithm was applied to MobiWay.We tested and analyzed the performance of MOSMT algorithm,and the tasks were completed with the least resources and the lowest time cost.Compared with the FIFO and Fair scheduler,the MOSMT algorithm can achieve similar functions,and it takes less time and is more economical to deal with multi tasks and multi objectives.