A Microservice-Oriented Resource Management System Based on Reinforcement Learning

Abstract

A microservice-oriented resource management system based on reinforcement learning, including: a state transformation unit, a resource allocation decision-making unit, and a management method evaluation unit. The state transformation unit collects the current operating status information of the data center, executes the actions of the resource allocation decision-making unit, and The resource demand relationship between microservices is transformed into a resource bipartite graph, and the resource allocation decision-making unit derives the performance characteristics of microservice resources through bipartite graph neural network modeling, and obtains a feasible resource allocation plan through the sensor network, and evaluates the management method The unit adopts the standard gross product index that reflects the benefits of data center resources to evaluate the pros and cons of the resource allocation plan, and realizes the optimization of the resource allocation plan by comparing the GNP of the system under different configurations of resource allocation decision-making units. In the case of limited resources, the present invention can, on the one hand, ensure the time delay requirements of microservice applications, and on the other hand, can maximize the use of data center resources, thereby improving the execution capability and overall performance of the data center.

Description

technical field [0001] The invention relates to a technology in the field of computer information processing, in particular to a microservice-oriented resource management system based on reinforcement learning in a data center. Background technique [0002] As an emerging software architecture, microservices emphasize the splitting of traditional large single application services into many small services with irregular characteristics and dynamic changes. Although the microservice architecture has been widely used, the resource management system under the condition of microservices faces two challenges: first, for many microservices, it can identify their irregular characteristics, which includes obtaining multi-dimensional resources of different microservices requirements and the importance of judging the resource allocation results of different microservices to the performance of the entire application, etc.; second, in a sudden and dynamically changing execution environme...

AI Technical Summary

Problems solved by technology

[0003] Aiming at the defects of resource waste and application performance degradation in the existing data centers in the above two situations, the present invention proposes a microservice-oriented resource management system based on reinforcement learning, and uses reinforcement learning to obtain irregular and dynamically changing microservices. According to the characteristics of the service, a resource allocation mechanism that is aware of the characteristics of the microservice is formulated, so that in the case of resource constraints, on the one hand, the delay requirements of the microservice application can be guaranteed, and on the other hand, the resources of the data center can be maximized, thereby improving Ability to execute and overall performance of the data center

Images