Distributive parallel computing platform and method based on CPU (central processing unit) core management

Abstract

The invention discloses a distributive parallel computing platform based on CPU (central processing unit) core management. The distributive parallel computing platform comprises a data node server, a dispatching node server and a computing node server, wherein the data node server comprises at least one computer with CPU multi-core capability, and is used for storing historical results and database service nodes, the dispatching node server comprises at least one computer with CPU multi-core capability, and is used for computing task dispatching and management, the computing node service comprises at least one computer with CPU multi-core capability and is used for processing data submitted by users and invoking third-party computing program to participate in the computing, and the third-party core computing program is deployed on the computing node. The distributive parallel computing platform based on CPU core management has the beneficial effects that the CPU multi-core technology is sufficiently utilized, the computing efficiency is greatly improved on the multi-task processing, computer resources are sufficiently utilized, a multi-core processing unit has better performance and efficiency advantages that a single-core processor, and the multi-core processing unit can become a widely adopted computing model.

Description

technical field [0001] The invention relates to the field of power system simulation distributed parallel computing, in particular to a distributed parallel computing platform and method based on CPU core management. Background technique [0002] The distributed parallel computing platform is to realize the distribution of computing tasks, task scheduling, summary of computing results, error handling, etc. in a multi-computer environment. interaction between. [0003] The hardware of parallel computer group has developed by leaps and bounds in the past three years. The parallel computing stand-alone hardware has jumped from the previous stand-alone two-way dual-core to the current four-way six-core. Twenty-four cores. With the construction of the D5000 platform project and the Shandong cloud simulation pilot project, the distributed parallel computing platform is clearly integrated into the D5000 platform and the cloud simulation platform as a basic function, and plays a k...

AI Technical Summary

Problems solved by technology

[0005] Judging from the current application performance, the advantages and disadvantages are obvious: whether the online running distributed parallel computing platform is periodic computing, event or manual trigger computing, the computing tasks are relatively saturated with computing resources, and the operation is efficient and stable; online and offline research The total number of CPU cores in the state parallel computing management cluster is less than the total number of tasks, that is, when tasks are saturated, the operation is efficient and stable; when the online and offline research state distributed parallel computing platform cluster tasks are not saturated, the operation is stable, but the computing resources The utilization rate is low, and the current distributed computing platform does not take into account the advantages of CPU multi-core processing, and the processing efficiency of multi-task computing is low, and the stability is not high

For example: the number of calculation failures is 40, the number of computing blades in the cluster is 52, and each blade has 8 cores of computing resources. The rate is only 9.6%; the time-consuming barrel effect of cluster computing on the distributed parallel computing platform is obvious, and the total time spent on analysis and calculation of a single batch of tasks (stages and rounds) is equal to the longest time-consuming calculation time in a round

Images