Research On Scheduling Strategy And Communication Optimization Technology Of Distributed Docker Cluster

With the rapid development of cloud computing technology,applications become increasingly large-scale,increasingly complex user needs,iterative updating increasingly frequent.The urgent need for the development of cloud computing,rely on LXC container technology support Docker technology came into being.Docker's powerful cross-platform support and efficient use of resources and isolation,providing a sustainable deployment of test methods and much higher than the traditional response to the speed of virtualization,the complete release of the lightweight virtualization Power,making the utilization of computing resources raised to a new level.Although there are many advantages of Docker,but it is widely used in large-scale distributed system,there are two difficulties.Firstly,with the increasing scale of Docker applications,distributed Docker clusters are being used in more and more actual production situations.Unlike a traditional cluster,a Docker cluster contains a large number of physical machines.On a single physical machine,multiple instances of the associated Docker containers are deployed,making the cluster more complex.How to improve the resource utilization of the whole cluster is the focus of this paper.How to implement the container task scheduling efficiently is the key problem in the large-scale distributed Docker cluster.Based on the task scheduling strategy of Swarm cluster management tool,which is provided by Docker,this paper proposes a new distributed Docker cluster scheduling strategy based on load balancing and exception recovery.And the simulation experiment is carried out in the LAN environment.The experimental data prove that the scheduling strategy has better performance in improving the resource utilization and operating efficiency of the whole cluster compared with other strategies.Secondly,relative to the CPU,memory and other traditional resource isolation,Docker is not perfect in the network isolation.The four kinds of network communication modes offered by the government are too simple and difficult to expand and lack the quality control of network service,which make it difficult to apply to the network bandwidth demand complex and network setting dynamic change application scenario.When multi-application hybrid deployment,can't be in accordance with the user's desired priority for more granular resource dynamic control.This paper will address the above scenarios,the Docker used in thetraditional NAT communication on the basis of adding dynamic address translation and adaptive bandwidth control and other functions,proposed a new container communication model.It is proved that compared with the traditional NAT,the communication model can make the Docker technology have a better performance in a wider range of application scenarios.