Design And Implementation Of Network Resource Management AndConfiguration System Based On ContainerCloud Platform

The cloud platform relies on virtualization technology to provide massive shared resource management and elastic scaling. In the past few years, virtual machine has played an important role of cloud platform infrastructure, enabling isolation and management of physical resources. However, it imposes extra performance degradation by introducing additional control layer. Container is a similar but much more lightweight solution. Compared to virtual machine, container is less resource-consuming and starts/stops with milliseconds, thus it has been suggested as a better solution for application packaging and deployment in the cloud. This offers both opportunities and challenges for container. Currently there is no mature network, security or storage solution for container, so it may take a long time to truly build a container-based cloud platform.This thesis analyzed the current network challenges of container-based cloud platform. According to a survey on existing network solutions for container, we find that the capabilities of them in network isolation, resource management and elastic network scaling aspects are still very limited, and cannot completely meet the network demands of cloud platform.In view of the network problem of present container-based cloud platform, this thesis puts forward three specific solutions as follows:1 Take advantage of network virtualization technology to divide the underlying physical network into multiple virtual network for tenants to use, each tenant can get an independent virtual network which is isolated from each other. Besides, the virtual network is more elastic and task migration on the virtual network is also more convenient.2 By using Linux provided network traffic control interface, we can manage the network resource usage of each container, fill in the gap that existing network solution cannot control network resource usage of containers. In addition to the resource management, this scheme also supports client monitoring the history of each container’s transmit/receive network traffic and real-time network transmit/receive speed.3 By using distributed configuration management and autonomous configuration based on cluster membership information, the new scheme is fast and fault tolerant compared to the existing manual or semi-automatic solution. It reduces the unnecessary manual intervention and meets the requirement of large-scale network configuration for cloud platformThis thesis not only puts forward the technical solution to solve the problems, but also implements them as a new network configuration and resource management software. We test this tool on the real container-based cluster, and find it meets the network isolation and all the other requirements and can help to manage network resources on cloud platform in a better way.