Data center task scheduling method based on dynamic temperature prediction model

Abstract

The invention discloses a data center task scheduling method based on a dynamic temperature prediction model. The method comprises the steps of firstly initializing parameters of a temperature prediction model; obtaining information such as actual computation node inlet temperature and CPU (Central Processing Unit) use rate; updating the parameters of the temperature prediction model through an updating algorithm; scheduling a batch of incoming tasks. The data center task scheduling method based on the dynamic temperature prediction model overcomes the disadvantages and defects of the current static temperature predication model applied to a task scheduling method, and is applicable to the online task scheduling of a computation intensive data center. The data center task scheduling method based on the dynamic temperature prediction model has the advantages that since factors of temperature and air flow are considered during task scheduling and the parameters of the temperature prediction model are dynamically adjusted according to the feedback information of a temperature sensor, the highest inlet temperatures of all computation nodes are enabled to be as low as possible, the cooling efficiency of a cooling system is finally improved, and the goal of greatly saving energy is achieved.

Description

technical field [0001] The present invention relates to a task scheduling method on a large-scale cluster such as a data center, and in particular to a task scheduling method that considers the influence of temperature and airflow on the cooling effect of the cluster by establishing a more accurate temperature prediction model at the software system level. The cooling efficiency of the cooling system can be improved to achieve the purpose of greatly saving energy. Background technique [0002] In the era of big data, all data centers in the world consume about 1.5% of the world's electricity consumption, and the heat dissipation energy consumption of the data center cooling system accounts for 50% of the total energy consumption of the data center. The cooling efficiency of the cooling system is very low due to the "hot spot" phenomenon caused by factors such as heat back flow. The task scheduling method that considers the temperature factor is mainly to avoid hot spots fro...

AI Technical Summary

Problems solved by technology

However, with the operation of the system, the ambient temperature change caused by the heat dissipation of the computing equipment will have a greater impact on the parameters of the model, gradually reducing the accuracy of the temperature model prediction, so it is necessary to adjust the parameters of the static temperature prediction model in real time

Images