Scheduling method and system suitable for mixed architecture container cloud

Abstract

The invention provides a scheduling method and system suitable for a mixed architecture container cloud, and the method comprises the steps: 1, filtering out a working node which is not matched with an application mirror image architecture in a pre-filtering stage of scheduling; 2, in the scoring stage of scheduling, according to the hardware performance and the allowance, scoring and sorting are carried out on the K8S working nodes at the same time; and step 3, in a score evaluation stage, according to the information of the scheduled K8S working nodes of the application, adjusting a score sorting result to obtain an optimal application scheduling result. According to the method, the K8S working nodes which do not conform to the mirror image architecture of the application are filtered out through the filter, it is ensured that the application is scheduled to the K8S working nodes matched with the mirror image architecture of the application, and the problem that in a mixed architecture scene, the application is scheduled to the working nodes with the unmatched architecture, and consequently starting fails is solved.

Description

technical field [0001] The present invention relates to the technical field of node scheduling, and in particular, to a scheduling method and system suitable for hybrid architecture container clouds. Background technique [0002] The traditional K8S cluster multi-copy application orchestration has weak copy control ability. In the mixed architecture scenario, the K8S cluster working node architecture is different and the performance is different. According to the current scheduling algorithm, it is difficult to achieve a relatively consistent performance of each copy of the application. [0003] Patent document CN108108245B (application number: CN201711364950.2) discloses a hybrid scheduling method and system for wide-node scientific workflow on cloud platforms. The method includes: when each scheduling stage is initialized, the task is constructed into a task multi-layer clustering tree structure by using the key-energy algorithm transformation; in the global initialization...

AI Technical Summary

Problems solved by technology

[0004] However, most current images only support a single architecture, but the scheduling algorithm does not recognize the image architecture. In a mixed architecture scenario, there is a high probability that the image and the worker node architecture will not match.

Images