An evaluation of IP-based fast reroute techniques

Abstract

Today, IP-based networks are used to carry all types of traffic, from the traditional best-effort Internet access to traffic with much more stringent requirements such as realtime voice or video services and Virtual Private Networks. Some of those services have strong requirements in terms of restoration time in case of failure. When a link or a router fails in an IP network, the routers adjacent to the failing ressource must react by distributing new routing information to allow each router of the network to update its routing table. A realistic estimate of the convergence time of a tuned intradomain routing protocol in a large network is a few hundred of milliseconds [1]. For some mission critical services like voice or video over IP, achieving a restoration time in the order of a few tens of milliseconds after a failure is important [2]. In this paper, we first present several techniques that can be used to achieve such a short restoration time. While most of the work on fast restoration has focussed on MPLS-based solutions [2], recent work indicate that fast restoration techniques can be developed also for pure IP networks. Recently, the RTGWG working group of the IETF started to work actively on this problem and several fast reroute techniques are being discussed. However, as of today, no detailed evaluation of the various proposed IP-based fast reroute techniques has been published. The goal of this short paper is to firstly provide a brief overview of fast restoration techniques suitable for pure IP networks, in section 2. Then, in section 3, we evaluate by simulation how many links can be protected by each technique in large ISP networks based on their actual topology. This coverage is an important issue as some techniques cannot protect all links from failures.