Cross-region Container Load Balancing Technology Based On OpenStack

As an open source cloud platform,OpenStack has the large-scale service deployment capability.More and more services are deployed to OpenStack cloud in the form of containers.A single OpenStack cloud is often resource-constrained,and cloud administrators need to expand multiple OpenStack clouds in the cross regions to meet isolation,security,and high availability requirements.When OpenStack cloud user requests increase dramatically,unbalanced load distribution among cross-region containers will overload some nodes in OpenStack cloud,which will affect the performance of the entire cross-region cloud.In addition,the Nova Cell mechanism in cross-region multi-cloud expansion scenario will affect the design and implementation of multi-cloud components,and there will be data redundancy and multi-user interface problems.In response to the above-mentioned cross-region multi-cloud expansion and crossregion container load balancing requirements,a cross-region container load balancing system based on OpenStack optimized the OpenStack cross-region multi-cloud architecture at the system level and solved the cross-region container load balancing problem at the application level.The new cross-region multi-cloud deployment scheme for the scale expansion of cloud platforms in this system use Central-Local mechansim to solve the problem of data redundancy and multi-user interface without affecting the overall performance of the cloud platform.The cross-region container load balancing system that responds to peak traffic can realize the user request distributing function,and can dynamically scale container instances according to the number of user requests to avoid a single point overload and resource wasting.A bi-direction decision model for container instances auto-scaling performs traffic prediction on the service front end and load monitoring on the service back end.The front-end prediction and the back-end feedback form a closed decision-making loop to accurately control the number of container instances in the system.The experimental results in a cross-region multi-cloud environment show that the load balancing system of cross-region container based on OpenStack can perform load balancing among cross-region containers effectively.At the same time,the network performance between two cross-region containers loses only 3% compared to the network performance between two cross-region bare metal hosts.The auto-scaling algorithm designed for container instances improves the system performance(Ratio of service response time to container retention cost)by 18% compared to the resource pre-allocation algorithm.