Next generation resilient redundant router

Abstract

The need in the commercial Internet to continually improve unit capital costs coupled with the increased reliability needed to carry all service types drives carrier architectures toward ever larger routers deployed in non-redundant configurations. This requires advancements in High Availability (HA) Non-Stop operation of these systems. We propose and implement a distributed architecture for these next generation core routers by implementing N-Modular Redundancy for control process software. The system employs an eventually consistent framework based on a Distributed Hash Table database. The overall goal is a system capable of non-stop routing, meaning that a router can still process both control and data messages, even after multiple software and hardware failures. Details on the architecture are provided and an estimation of scalability, overhead, and fault recovery time is presented.