Traffic Engineering for LISP-Enabled Networks

Abstract

Inter-Domain Traffic engineering in the Internet faces serious limitations because of the current IP routing and addressing architecture. This coupled with Border Gateway Protocol’s (BGP’s) way of selecting performance-blind paths forces ISPs to de-aggregate IP prefixes to control the flow of packets between ASes. Advertising such de-aggregated, surplus prefixes for local benefits is causing the routing table of the Default Free Zone (DFZ) to grow rapidly, which contribute to routing scalability problems. Recently, in order to address this scalability issue, LISP (Locator/Identifier Separation Protocol) has been proposed, which separates an address space into a non-routable End-Point Identifiers (EIDs) and a routable Routing Locators (RLOCs), where each EIDs can be associated to more than one (multiple) RLOCs. In this work, we discuss two optimization models for traffic engineering in LISP-enabled network which exploits the route diversity or the path diversity the LISP inherently provides by introducing the concept of grouping multiple RLOCs with traffic proportioning or load-balancing as the optimization criterion. We compare the models to the base case that identifies with the current routing architecture (i.e. no proportioning). Through our study, we observe that LISP-based traffic engineering with multiple RLOCs offers noticeable benefits compared to when we do traffic engineering without proportioning demands to multiple RLOCs, except when the network is uniform in terms of load and capacity.