Internet Scalability (Part 2)

In this lecture Florin continues with the explication of the possible solutions.

-Evolutionary Internet Architectures (Location/ID Split): IP addresses have two complementary roles (Identifier and Locator). Change of locator results in change of identifier thus breaking ongoing flows. The solution is to separate the two functions

– Host-based approach: The principal characteristics of this are that it converts packets such that transport layer is exposed only to identifiers and Locators are present just at network routing level, Host manages locators and Host obtains Loc/ID binding.

– Network-based approach: Hosts are unchanged; each host has a stable IP address. IP address used as an identifier and The ID is not globally routable.

LISP-> It requires:  No changes at the end-hosts, a few network equipment to be changed and incremental deployment. Now, exists two addresses, the end-host ID address space (EID) and the routing locators address space (RLOC), due to the hosts and majory of routers are unaffected, no changes within the core networks, the mapping systems needs to be added and the introduction of tunnel routers, which serves to don't advertise their EID prefixes into BGP anymore.

One EID can be associated with more than one RLOC, so we can stablish priorities and weights. The benefits of LISP are: decrease of the Default Free Zone routing table, proper multihoming support,  no changes at the hosts and only few routers need to be changed, but also has some issues: the path reachability problem, mapping system scalability, deployment of scenarios, etc. The first can be increased with the LISP specific mechanisms, the second with the LISP Tree what is based on the DNS ideas and the third with draft-jakab-lisp-deployment.