A Dynamic Binary Hash Scheme for IPv6 Lookup

Q. Sun1, X. Huang1, X. Zhou1, and Y. Ma1,2

1. School of Computer Science and Technology2. Beijing Key Laboratory of Intelligent

Telecommunications Software and MultimediaBeijing University of Posts and Telecommunications

IEEE GLOBECOM 2008

MDS: Maximum Disjoint Set [13] K. Zheng and B. Liu, “V6Gene: A scalable IPv6 prefix gen

erator for route lookup algorithm benchmark,” IEEE AINA 2006

Characteristic of IPv6 table

The proposed greedy algorithm to find a MDS for a given routing table

OS: Overlap Set

Overall structure

Insert procedure of DBH

Delete procedure of DBH

Most of the MDS prefix lengths are less than 64 bits

There is no prefix with prefix length less than 16 The routes peak at lengths of 32, 35, and 48 The total number of unique MDS prefix lengths in

combined distribution is 16, greatly less than the length of IPv6 address 128

Optimized searching path for current IPv6 routing table

For future IPv6 consideration According to RFC 3177 and RIPE 267, the IPv6

address blocks should be allocated to subscribes following three rules

‘/48’ in the general case, expect for very large subscribes, which could receive a ‘/47’ or multiple ‘/48s’

‘/64’ when it is known that one and only one subnet is needed by design

‘/128’ when it is ABSOLUTELY known that one and only one device is connecting

So the future IPv6 routes should peak at lengths of 48, 47, 64, and 32

Optimized searching path for future IPv6 routing table

the future IPv6 routes should peak at lengths of 48, 47, 64, and 32

Evaluation platform

Implemented in standard C Intel Pentium 4 PC 2.4GHz CPU 512M DDR333 Red Hat Enterprise Linux Advanced

Server 4.0

Lookup speed

Avg. lookup memory access number is only 2.7 for current IPv6 routing tables

Quite stable for future IPv6 routing tables

Memory and update