Research And Application Of Microservice Failure Avoidance Technology In Federated Cloud Platform

The battlefield or disaster relief environment includes a large number of information data processing tasks.The traditional cloud computing platform is too far away from users to meet the real-time task requirements in the unstable network environment.Federated cloud is an emerging cloud computing paradigm,which can manage multiple distributed clusters to provide services to users nearby,and the clusters scale can be dynamically adjusted according to the task demand.With feature of loose coupling,extensibility,and lightweight,the microservice architecture is more appropriate service deployment model in resource-poor scenario.However,due to the complex dynamic environment of federated cloud platform and the interaction requirements of microservice applications,cloud services are at serious risk of failure.The existing cloud platform mainly adopts redundant deployment or elastic scaling according to load to avoid services failure,without considering the dependence between microservices and the dynamics of clusters.The previous method has the problem of service fluctuation and failure response delay,so the reliability of microservice applications in the federated cloud cannot be guaranteed.This thesis is aim to study how to avoid frequent failure of microservice in federated cloud and provide highly reliable microservice applications.First of all,in the service deployment stage,this thesis designs a two-level reliability evaluation model considering the impact of the node and sub-clouds on services,and proposes a reliable redundant service placement method.This method defines the service placement problem as a graph node mapping problem,and searches a high survival rate service placement strategy according to the reliability of nodes and service fault tolerance.Secondly,in the service running stage,this thesis proposes a microservice dynamic adjustment framework based on MAPE-K control model,which collects service running data regularly and captures abnormal services in advance for adjustment.The framework adopts an elastic scaling method based on dynamic window,which can improve the stability of services when dealing with changing traffic through the adjustment strategy of fast expansion and slow contraction.At the same time,considering the difference of network latency between different sub-clouds and the interaction requirements of microservices,this framework designs a service migration strategy based on affinity,which deploys highly dependent services in close locations to reduce access latency between microservices.Finally,this thesis implements a federated cloud platform system with fault avoidance modules,and verifies the system functions and methods.Experimental results show that the proposed method can effectively avoid microservices failure and improve the reliability of microservice applications.