Research And Implementation Of Microservice-based Auto Scaling On Application Cloud Platform

As cloud platforms are more and more widely used in various fields,container-based cloud platforms have gradually entered people's horizons.Due to the increasing problems in deploying monolithic applications on cloud platforms,microservice architectures that are tightly integrated with container technology have begun to appear on cloud platforms.This kind of application architecture has many characteristics,and scalability as a vital characteristic has also begun to be valued and widely studied.This paper mainly aims at the microservice-based application cloud platform.Based on reading the literature and studying the project source code,this paper analyzes the existing auto scaling solutions and clarifies the problems that existing cloud platforms need to solve during the scaling process.The specific implementation of scaling is as well as sufficiently researched.In order to optimize the node creation and image pull phase,this paper proposes an optimal design of an auto scaling solution based on microservice.The design is mainly divided into two parts.The first is designing a decentralized image pull solution in the image pull phase to reduce the pressure on the image repository during the elastic expansion process.The second is to take advantage of the high similarity of image structure in microservice applications and propose a multi-level elastic scaling method based on image preloading and multi-level nodes.By classifying images and preloading them into nodes,the pre-loaded nodes can skip the image-pull phase and start the container directly.The design of multi-level nodes ensures that the spare node that preloads images will be activated preferentially,thereby effectively avoiding the high cost of node resource allocation and image transmission,thereby greatly accelerating the elastic scaling phase in the nodes.On the basis of the above solution,a microservice-based application cloud platform is designed and developed.The optimized elastic scaling scheme is adopted in the platform and well-verified under actual scenarios.The test results show that the multi-level auto scaling scheme proposed in this paper can effectively reduce the time cost in actual scenarios during the elastic expansion phase through image preloading and design of multi-level nodes,thereby improving the scaling effectiveness under facing sudden load and strengthening the stability of the system.