A Novel Scalable IPv6 Lookup Scheme Using Compressed Pipelined Tries

A Novel Scalable IPv6 Lookup Scheme Using Compressed Pipelined Tries Author: Michel Hanna, Sangyeun Cho, and Rami Melhem Publisher: NETWORKING 2011 Presenter: Han-Chen Chen Date: 2011/5/11

Introduction • IPv6 packet forwarding • IPv4 trie based packet forwarding algorithms port to IPv6 • Trie height • Memory • TCAM • High power consumption • Poor scalability to long IPv6 prefixes • Hash-based hardware engines • Constant search time • Collision chaining causes unbounded memory access time

Introduction • The outcome of any IP lookup operation is an interface (or output port) number. • Real-life backbone routers contain a relatively small number of output ports • The routers assign many next hop addresses to one output port. • This means that the number of unique next hop information is greater than the number of output ports.

Introduction

Multibit trie (a) Multibit trie for Figure 1(a) with strides {3, 2, 2}. (b) Its leaf-pushed trie.

Uncompressed multibit trie -2 : with child link -1 : null >0 : port number

Compressed multibit trie Uncompressed trie Compressing the leaves Inter-Node Compressed Trie

Performance Evaluation (1/5) Statistics of IPv6 tables on June 2010. H is the number of output ports.

Performance Evaluation (2/5) Number of INCT trie levels vs. total memory (KB) for Linx2 and Eugene1

Performance Evaluation (3/5) Memory (KB) of INCT(7) vs. Uncompressed(7) trie

Performance Evaluation (4/5) Memory (MB) of Lulea(6)/Tree Bitmap, INCT(6), MIPS(57) and Binary INCT(57). MIPS & Binary INCT {8,1…56} Lulea & INCT {16,16,8,8,8,8}

Performance Evaluation (5/5)

A Novel Scalable IPv6 Lookup Scheme Using Compressed Pipelined Tries