中文摘要：

针对MapReduce任务调度中任务属性取默认值的不合理性以及人为指定值的不确定性,对调度算法实现动态调整任务优先级、计算合理的Reduce任务数、明确Reduce任务启动时机等改进,达到提升任务并行度、缩短作业执行时间的目的.Fair与LATE算法改进前后的实验结果表明,基于任务属性的改进能提高调度算法性能与作业整体执行效率.

英文摘要：

Task scheduling is the core problem in MapReduce.Researches on scheduling algorithm rarely involve task attributes.Aiming at the irrationality of default values of the task attributes and the uncertainty of artificially specified values,this study made dynamic priority adjustments and calculated the rational number and the starting time of Reduce tasks,with a view to enhancing tasks parallelism and reducing the executing time of jobs.The results of the experiments based on Fair and LATE algorithms show that the improvement based on task attributes improves the scheduling algorithm performance and the overall jobs execution efficiency.