The Design And Implementation Of The Basic Integrated Support System For Distributed Resource Scheduling Platform

In a large-scale distributed platform,a large number of physical nodes need to be managed,and various distributed applications are deployed on the physical nodes.Similar to the requirements of a stand-alone operating system,distributed applications are required to be executed concurrently and isolated from each other.To make sure the applications are isolated from each other,two points should be guaranteed: the first is the network level,each distributed application needs to have an independent network space,and the runtime network environment where each distributed application is located should be isolated from each other and have independent IP addresses and the port number;the other level is the distributed process isolation of each physical node,and the physical resources owned by the runtime environment of each application should be isolated from each other.At present,there are mainly two technical means that can well support the above requirements: virtual machines and containers.Virtual machines were widely used in the early days,and its main problem was that it took up too much resources.The container is more lightweight,equivalent to an operating system process,but not a complete operating system,but the isolation of the process.With the development of containerization technology,more and more applications are deployed in containers,making application deployment and maintenance more convenient.But at the same time,how to deploy and manage containers more efficiently has become a new problem.In view of the above background,the distributed resource scheduling platform is based on Docker containers.The content of this article is the basic comprehensive support system in the distributed resource scheduling platform,including the design and implementation of the network architecture of the distributed resource scheduling platform,and distributed resources The authentication and authentication of the scheduling platform and the fault detection and recovery of the container.This thesis mainly completes the following tasks:1)Designed and implemented an overlay network based on VXLAN technology.The overlay network is built on the container host’s network,which separates the communication between containers from the physical environment,improves the flexibility of cross-host container communication;manages and allocates the IP address of the container to ensure that each container in the cluster Unique IP address;2)Designed and implemented the authentication and authorization system of the distributed resource scheduling platform.The system includes the functions of certificate registration,certificate issuance and certificate revocation,and assigns corresponding authority to each certificate to ensure the security of the cluster.The communication within the cluster adopts the method of data encryption,the DH key exchange algorithm is used to realize the key exchange between the communicating parties,and the AES symmetric encryption algorithm is used to realize the encrypted transmission of data;3)In order to ensure the high availability of the cluster,the watchdog is designed and implemented to monitor the status of the containers in the cluster,discover unhealthy containers and physical hosts in time,notify the scheduling module to generate corresponding scheduling policies,and restart or migrate the containers in time.Through the function and performance test of the basic integrated support system of the distributed resource scheduling platform,the characteristics of the network of this system in the distributed resource scheduling platform are analyzed.Finally,it summarizes the basic comprehensive support system of the distributed resource scheduling platform and looks forward to the future.