Research And Implementation Of Container Cloud Resource Scheduling Method For Multi-Dimensional Computing Framework

Docker container virtualization technology is a virtualization solution that shares the operating system kernel.The cloud platform based on Docker technology is gradually becoming the core of the next generation cloud computing.In the container cloud,the resource scheduler plays a key role in the performance and resource utilization.Kubernetes container orchestration engine has become the most widely used container cloud platform scheduler with its powerful service discovery,cluster monitoring and error recovery capabilities.However,Kubernetes does not consider the multi-computing framework resource scheduling problem,resulting in low resource utilization and poor load balance when the container cloud cluster processing big data problems.For the insufficient performance of Kubernetes resource scheduling under multicomputing framework,this thesis proposed a MRWS(Multidimensional Resource Weights Scheduling)scheduling method,and developed a container cloud platform Paladin for multi-computing framework based on open-source container Open Shift.Designed and implemented new scheduling methods on the platform and performed scheduling performance tests.The main contributions are as follows:· A container cloud resource scheduling method MRWS for multi-computing frame-work is proposed.The method comprehensively considers the resource require-ments of the container to be scheduled,the CPU of the physical node,the memory,the disk,the network bandwidth idle rate,and the number of deployed container applications,and comprehensively scores based on the multi-dimensional resource weighting parameter.· Designed a method to automatically solve MRWS weight parameters.For the weight parameters in the MRWS scheduling method,the FAHP(Fuzzy Analytic Hierarchy Process)is used to automatically model the cluster resources and solve the container application multi-dimensional resource weight parameters.· Designed and implemented the MRWS scheduling method on the container cloud platform and performed performance analysis.Firstly,a large-scale scheduling simulation is performed on the container cloud simulation platform Container-Cloud Sim,and compared with the Random,First Fit,and Kubernetes scheduling algorithms.The new scheduling method can improve the cluster load balance by about 40%.Then based on Open Shift,the container cloud platform Paladin for multi-computing framework was developed.Finally,the multi-computing framework resource scheduling performance test is performed on the platform,and the new scheduling method improves the task processing performance by about 20%.