A method and system for parallel task scheduling based on CPU core number prediction

Abstract

The invention discloses a method for realizing parallel task scheduling based on the prediction of the number of CPU cores. By measuring the high computing power of heterogeneous processors, the execution effect of scheduling algorithms, and various performance indicators such as processor load balance, the application of Tianhe No. 1 The performance indicators of the designed data prediction parallel technology and scheduling algorithm are actually measured by nodes with strong computing power. It can be found that the calculation time is significantly reduced, the parallel execution time of the prediction method is greatly reduced, and the execution of the algorithm is more efficient, and the calculation The method maintains a good processor load balance and obtains a better scheduling execution result. The present invention can make full use of existing hardware resources for calculation, and proves the execution efficiency of the prediction method based on heterogeneous processor clusters and the reliability of the parallel execution of the scheduling algorithm used in operation, and also A good load balancing between processors is ensured.

Description

technical field [0001] The invention belongs to the technical field of computer heterogeneous cluster computing, and more specifically relates to a method and system for realizing parallel task scheduling based on prediction of the number of CPU cores. Background technique [0002] At present, high-performance computing research using the computing resources of supercomputing centers has been greatly popularized in China. However, there are some problems that cannot be ignored in most supercomputing centers’ scheduling strategies for tasks: first, due to insufficient task scheduling, the queuing time of jobs is too long, resulting in low scheduling efficiency; Reliable prediction of the load of the scheduling queue leads to the inability to efficiently schedule jobs that require large-scale processors to be processed in the corresponding scheduling queue, thereby increasing a lot of time overhead; third, because the scheduling strategy does not use effective The load balanc...

AI Technical Summary

Problems solved by technology

[0003] In view of the above defects or improvement needs of the prior art, the present invention provides a method for realizing parallel task scheduling based on CPU core number prediction. Due to the technical problems of long job queuing time and low scheduling efficiency, and because there is no reliable prediction of the load of the scheduling queue, the jobs that require large-scale processors for calculation cannot be efficiently scheduled to the corresponding scheduling queue The technical problem of adding a lot of time overhead due to processing, and the technical problem of serious load imbalance caused by not using an effective load balancing strategy and forming a serious scheduling performance bottleneck

Images