A Survey of Fast-Recovery Mechanisms in Packet-Switched Networks

Abstract

In order to meet their stringent dependability requirements, most modern packet-switched communication networks support fast-recovery mechanisms in the data plane. While reactions to failures in the data plane can be significantly faster compared to control plane mechanisms, implementing fast recovery in the data plane is challenging, and has recently received much attention in the literature. This survey presents a systematic, tutorial-like overview of packet-based fast-recovery mechanisms in the data plane, focusing on concepts but structured around different networking technologies, from traditional link-layer and IP-based mechanisms, over BGP and MPLS to emerging software-defined networks and programmable data planes. We examine the evolution of fast-recovery standards and mechanisms over time, and identify and discuss the fundamental principles and algorithms underlying different mechanisms. We then present a taxonomy of the state of the art, summarize the main lessons learned, and propose a few concrete future directions.

Improving the Resilience of Fast Failover Routing

“…Failures are common in computer networks [32] and hence fast reroute mechanisms have been studied comprehensively over the last decades, we refer to a recent survey for an overview [8]. In general, for rapid connectivity restoration, the computation & distribution of new routes is too slow [29], but even convergence mechanisms in the data plane such as DDC [29] can introduce non-trivial delays [5].…”

“…Fast reroute [8] is an attractive solution used in most modern communication networks today to quickly reroute traffic around failed links (henceforth also called fast failover). The fast reroute mechanism is implemented in the dataplane, relying only on local information for fast decision making, and hence avoiding the overheads and delays usually involved in control plane mechanisms (such as path reconvergence, link reversal or the notification of a centralized controller) [8].…”

Randomized Local Fast Rerouting for Datacenter Networks with Almost Optimal Congestion

“…Motivated by measurement studies of network-layer failures in datacenters, showing that especially link failures are frequent and disruptive [13], the problem of designing resilient routing mechanisms has received much attention in the literature over the last years, see e.g., [3,4,[6][7][8][9] or the recent survey by Chiesa et al [5].…”

“…First consider the case of some node w having in-deg(w) ≥ log 5 n in T R . Fix this node w. In this case there must be a set of log 5 n nodes v such that edges of form (v, w) lie in G t log . We now set F v = {d} for each of these nodes v. By Observation 4, each of these log 5 n edges will appear in G F α with independent probability at least 1/ log 4 n. From a Chernoff bound application, it follows that at least log n/2 of these edges will indeed appear in G t log w.h.p.…”

“…To meet the resulting stringent dependability requirements, datacenter networks implement fast failover mechanisms that enable routers to react to link failures quickly and to reroute flows in a decentralized manner, relying on static routing tables which include conditional local failover rules. Such local failover mechanisms in the data plane can react to failures orders of magnitudes faster than traditional global mechanisms in the control plane which, upon failure, may recompute routing tables by running the routing protocol again [5,11,17].…”