Research Of Docker Container Scheduling Optimization Method

In cloud computing,virtual infrastructure based on virtual machine has been widely used in elastic cloud resource configuration.The resource scheduling unit with the minimum granularity of the virtual machine has the problems of slow scheduling time,low resource utilization,and inconsistent software stack environment.The emergence of Linux-based container-based virtualization technology developed by Google provides a new complement to the cloud computing system,consumption and delivery model.Deploying the service in a container avoids the disadvantages of excessive overhead and resource utilization of traditional web containers.However,due to the different architectures of container technology and virtual machines,traditional virtual machine-based resource management methods are not well suited for container technology.Container-based resource scheduling has become a hot topic of research.In order to solve this problem,a scheduling optimization algorithm based on Docker container is designed and implemented.(1)Aiming at the dynamic change,the uncertainty of cloud resources,and the problem of node resource fragmentation caused by the container initial deployment process,an initial deployment scheme of the container is proposed.First of all,a dynamic weighting algorithm is used to analysis the load of the node;And then an improved clustering algorithm is used to select the best deployment node for the container to be deployed.Experiments show that the initial deployment scheme can reduce the generation of resource fragments and improve the resource balance and resource utilization of nodes.(2)Aiming at the problem of response delay in container dynamic scheduling,an improved multi-index grey load forecasting algorithm is adopted to provide the basis for container dynamic scheduling timing.A dynamic scheduling algorithm based on multi-dimensional resource balance of nodes is proposed because of the imbalance of resources in each dimension of nodes.(3)Aiming at the serious energy waste in cluster and the unbalanced load caused by the dynamic change of node load,a scheduling scheme based on Docker pre-migration is proposed.The initial deployment scheme is used to select deployment nodes for containers that need to be migrated.For resource nodes with overloaded load,multi-index grey load prediction algorithm is used to select the least load container migration;for resource nodes with low load,all the containers are migrated,and then the resource nodes are closed.The experimental results show that the migration scheduling scheme can improve the resource balance of nodes,load balance of clusters and reduce the application response time.