Font Size: a A A

Automated Management Of Cross-regional Container Networks Based On Openstack

Posted on:2020-07-23Degree:MasterType:Thesis
Country:ChinaCandidate:Z Z XuFull Text:PDF
GTID:2428330599458570Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The collaboration of cloud computing and edge computing can effectively improve the quality of services,and more and more services are being deployed to the federated cloud and edge.To ensure high availability of services,the cross-regional cloud needs to construct a large Layer 2 network and connect networks in different areas for breaking the isolation between clouds and improving communication efficiency between services.In addition,more and more enterprises adopt micro-service architecture,the number of services increases sharply,and service communication in the application layer needs to be monitored and managed.However,currently there is no full-stack management of cross-regional networks in the open source cloud computing community.In response to the above network requirements,the automated management of cross-regional networks is responsible for the connection between multiple clouds,while the Communication Management System for Service Chain(COMSSC)handles the control of service communication in the application layer.(1)A new topology is designed for cross-regional networks in the network layer,using the virtual central router to forward and route packets,reducing the update delay of adding new networks.The implementation of this new topology construction scheme has been merged into the official OpenStack community.(2)The communication management system for service chain is designed and implemented.This system stores the requests in different message queues according to the service chain to which the request belongs,constructs a control plane for communication between services,and implements functions such as monitoring and control of communication.(3)An intelligent container allocation strategy for different service chains is proposed.It counts the number of requests in real time,predicts service load through time series analysis,and adjusts the number of container instances in real time.This approach considers the competition relationship of different service chains when assigning container instances,avoiding the monopoly of resources in a single service chain.In single-cloud and multi-cloud environments,experimental results demonstrate that the communication management time spent in the test application is less than 20% of the total service response time,and the overhead is similar to the communication management solution of the Kubernetes community.At the same time,the result of measuring the container instance allocation of COMSSC on multiple service chains proves that chain-oriented communication management ensures the fairness of service chains in the single cloud environment.
Keywords/Search Tags:Cross-regional networks, Communication management, Service chain, Container, OpenStack
PDF Full Text Request
Related items