An analysis of reconvergence delay when using BGP-LS/PCEP as southbound protocols
Aim of this paper is to explore latency in reconvergence after link failure when using BGP-LS/PCEP as southbound protocols in IP/MPLS network with centralized controller. Controller in network is PCE element in charge of Traffic Engineering management. Scenario in this paper describes an IP/MPLS network with four routers, fours links and four tunnels passing through multiple links. In case of a link failure, affected tunnels must be rerouted to alternative links evenly distributing traffic across the links. Therefore, it is necessary to have a central controller that will have a complete knowledge of the network in order to evenly and efficiently redistribute traffic. Our IP/MPLS network contains multi-vendor devices and do not support OpenFlow protocol hence as a southbound protocols standardized BGP-LS and PCEP are used. BGP in general is designed as a very slow protocol. Additional, BGP-LS has to wait for the internal routing protocol on the routers to update the Traffic Engineering Database before the BGP-LS signals change event to its peers. Therefore, we will explore some possibilities of reducing the latency in reconvergence after link failure.