低时延通用NFV故障恢复与弹性伸缩方法研究

如何保证和提高NFV环境中软件中间盒的可用性是一个开放的挑战性难题，也是NFV实用化道路上最急待解决的问题之一。故障恢复与弹性伸缩是提供可用性保证和提高可用性的有效手段，但目前尚未解决通用性、低时延流迁移和低时延流状态管理三个关键问题。为解决上述问题，本项目通过研究一个基于通用中间盒包处理模型的流状态复制方法和一个基于中间盒链的可扩展流迁移方法，实现在不修改软件中间盒源代码的前提下，当中间盒因故障不能提供服务时，可快速、低开销、正确地恢复服务，以提供可用性保证。通过研究一个基于容器组的本地流状态访问方法和一个基于大多可靠多播的远程流状态访问方法，实现中间盒在数据中心中按需即时迁移、扩张与缩减，以提高可用性。本项目研究具有重要理论研究意义和实用价值，将从理论与系统设计实现角度，为解决NFV可用性问题提供有力支撑，同时有助于云南楚雄彝族自治州的高水平科技人才培养与云数据中心新兴产业培育。

How to ensure and improve the availability of soft middleboxes in NFV is an open challenge, which is also a critical problem to be solved urgently on the road of the practicality of NFV. Fault recovery can provide availability guarantees and auto-scaling is the most effective means to improve availability, but the state-of-the-art still cannot solve three challenges: general-purpose, low-latency flow migration and low-latency flow state management. This project proposes to research a flow state replication method based on a general middlebox packets processing model, a scalable flow migration method based on the middlebox chain, by which faulted software middleboxes can be fast, with low-overhead and correctly recovered with their source codes unmodified. By studying a local flow state access method based on the abstraction of container group, a remote flow state access method based on a mostly reliable multicast communication model, we expect to get the answer of how to instantly migrate, scale out or scale in a software middlebox on demand. The research of this project has essential theoretical significance and practical value. The research findings would provide a strong support for addressing NFV availability issues from the perspective of theory and system design. At the same time, as for Yunnan Chuxiong Yi Autonomous Prefecture, this project will be beneficial for its training of IT talents and its cultivation of the cloud industry.