Assessing Locator/Identifier Separation Protocol interworking performance through RIPE Atlas

Abstract

The Locator/Identifier Separation Protocol (LISP) has been proposed to tackle the BGP scalability issue encountered in legacy Internet architecture, while introducing additional benefits. The key idea of LISP is to split the conventional IP addressing space into two categories: Identifier and Locator. Identifier is used for identification and Locator is used for routing. It is currently under standardization in IETF and is deployed in the wild at the same time, thanks to two testbeds: the LISP Beta Network and the LISP-Lab project. The interworking mechanism is proposed to ensure the communication between LISP-speaking sites and legacy Internet. The performance of LISP interworking with legacy Internet is of paramount importance to promote the adoption of LISP in future Internet. The existing research efforts to evaluate LISP interworking lack of large scale experimental results. The objective of this paper is to fill this gap by providing a latency evaluation and routing path measurement about LISP interworking mechanism in real networks. The work is based on a 2 weeks measurement campaign by using RIPE Atlas, which is the largest existing Internet measurement infrastructure for both IPv4 and IPv6. Experimental results show that LISP introduces additional latency, especially for close destinations, but negligible for intercontinental long-distance destinations. The additional latency highly depends on the position of the new network element being in charge of the communication between LISP-sites and the legacy Internet. Although LISP introduces some stretch, its performance is generally stable and its latency is not very high compared to natively forwarding without using LISP.