NeTS: Small: Network Failure Recovery Employing Network Coding

远程通信网络中使用的设备包含克服故障的机制，但是，这导致使用备用容量（冗余）来实现相对较短的恢复时间（约50毫秒）。因此，一些提供商实现了专有的网络故障恢复技术，其冗余程度较小，恢复时间较长。对于使用路由器和互联网协议的网络来说，从网络故障中恢复是特别困难的。本研究利用网络编码技术在未来网络中实现无命中（对用户透明）的服务恢复。该研究正在调查实际网络的拓扑和代码设计算法，以了解其性能，比较新方法与传统技术的恢复时间和额外容量需求，并发现实现新技术的新协议。在这项工作中，使用网络编码的故障恢复设计将分两步实现。在第一步中，对于给定的网络，将使用线性程序设计故障恢复网络。在第二步中，将确定所需的伽罗瓦字段大小。新的无碰撞恢复技术将在备用容量、恢复速度和操作复杂性方面与使用不相交路径的传统故障恢复技术进行比较。这项新技术还将使网络中的丢包恢复成为可能，而不是像传统的那样在端到端基础上进行。一种用于无线网络的特殊实现也在开发中。这项研究的成功结果将对未来的网络产生广泛的影响。

Facilities used in long-distance communications networks incorporate mechanisms to overcome failures, however, these result in the use of spare capacity (redundancy) to achieve relatively short restoration times (about 50 ms). Therefore, some providers have implemented proprietary network failure restoration techniques with a smaller degree of redundancy and with a larger restoration time. Restoration from network failures is particularly difficult for networks that employ routers and the Internet Protocol. This research is applying using network coding techniques for hitless (transparent to the user) restoration of services in the networks of future. The research is investigating topology and code design algorithms for actual networks, to understand their performance, to compare restoration time and extra capacity requirements of the new approach with conventional techniques, and to discover new protocols to implement the new technique. In this work, failure recovery design using network coding will be implemented in two steps. In the first step, for a given network, a failure recovery network will be designed using a linear program. In the second step, the needed Galois field size will be determined. The new hitless restoration technology will be compared with conventional failure recovery techniques using disjoint paths in terms of spare capacity, restoration speed, and operation complexity. The new technique will also enable packet loss recovery in the network, as opposed to on an end-to-end basis as performed conventionally. A special implementation for wireless networks is also under development. The successful outcome of this research can have a broad impact on future networks.