Cache-aware data structures for packet forwarding tables on general purpose CPUs

TL;DR

This paper introduces a cache-aware search algorithm for packet forwarding tables that improves software-based longest prefix matching performance on general purpose CPUs by optimizing cache utilization and achieving logarithmic search complexity.

Contribution

It presents a novel cache-aware data structure and search algorithm that enhances software packet forwarding efficiency on general purpose hardware.

Findings

Achieves O(log n) average search time for IP prefix matching

Reduces cache misses in software packet forwarding

Improves throughput of software-based routing solutions

Abstract

Longest prefix matching has long been the bottleneck of the Bloom filter-based solutions for packet forwarding implemented in software. We propose a search algorithm to match a destination IP address against a compact representation of the FIB table in CPU cache on general purpose hardware with an average performance target of O(log n) for an n-bit address.