After compare and analysis the advantage and disadvantage of the default speculative task scheduling algorithm on the Hadoop platform and LATE scheduling algorithm in heterogeneous environment

an improved speculative task scheduling algorithm based on Hadoop computer cluster is proposed. The algorithm dynamically adjusts and updates the proportion of stages in the Reduce task based on node history information

smoothes the real-time task progress rate to improve the forecast accuracy of the remaining completion time

and applies MCP model to verify the effectiveness for backup task. The experiment analysis show that the algorithm can improve the success rate of backup task and reduce job completion time.