A novel architecture in high availability (HA) systems

High availability is one of the most important features in modern complicated computer and communication systems, especially for backbone network devices, e.g. core routers, most of which can benefit from the hierarchical concurrent redundant backup models based on the inherent parallel architecture to achieve high dependability. We study the structure of redundant backup policy from the viewpoint of decreasing the time for takeover. Unlike the traditional approaches common in use, which typically deal with the recovery strategies located in different levels separately, our model assume an overall centralized error detecting component to improve the recovery efficiency. We propose a novel mechanism, crash-at-the-first-error (CAFE), and discuss the performance it can achieve.