Donnet, Benoît[Université de Liège - ULg > Dép. d'électric., électron. et informat. (Inst.Montefiore) > Algorithmique des grands systèmes >]
du 21 mai2012 au 25 mai 2012
[en] LISP ; Locators ; Measurement
[en] In the Internet, IP addresses play the dual role of identifying the hosts and locating them on the topology. This design choice limits the way a network can control its traffic and causes scalability issues. To overcome this limitation, the Locator/Identifier Separation Protocol (LISP) has been introduced. In LISP, the addresses used to identify end hosts (i.e., identifiers) are independent of the addresses used to locate them (i.e., locators). LISP maps identifiers into a list of locators and provides a mean to transport the packets with the appropriate locator. A key feature of this separation is that several locators can be associated to a given identifier, leading to more control for an end-site on the path selection to reach a given destination.
In this paper, we show that the choice of the locator can have an impact on the performance and the reliability of the communication in a LISP environment. To this aim, we build a mapping between identifiers and locators as if LISP were deployed today. In addition, we extensively collect delay data between locators and demonstrate that the locator selection for a given identifier prefix impacts the performance of the LISP path in 25% of the cases. Finally, we measure the locators availability over time and demonstrate that it remains quite stable.