A Hash-Free method for FIB and LNPM in ICN programmable data planes

• Published in: 2022 International Conference on Information Networking (ICOIN) pp. 186 - 191
• Main Authors: Rosa, Eduardo Castilho, Silva, Flavio de Oliveira
• Format: Conference Proceeding
• Language: English
• Published: IEEE 12.01.2022
• Subjects: Data models; FIB; Hash functions; ICN; Memory management; Name Lookup; NDN; Programmable Data Plane; Simulation
• DOI: 10.1109/ICOIN53446.2022.9687201

One core component of Named-Data Networking (NDN) is the Forwarding Information Base (FIB). FIB makes the NDN a global network where hierarchical names are routed by performing a Longest Name Prefix Matching (LNPM) at each node. One of the challenges in NDN is how to perform the LNPM so that even up lookup speed and memory consumption. Even though some initiatives speed up the LNPM by using P4 in programmable data planes, they all rely on hashing. However, given the huge namespace that NDN is subject to, the collisions caused by the hash functions may increase memory consumption significantly. This work introduces a FIB model for programmable data planes that are hash-free and balance lookup speed and memory consumption. Instead of using one large FIB, we segregate it into smaller ones and perform the LNPM using packet recirculations. Preliminary simulation results have shown that, compared with the state-of-the-art, our method consumes three times less memory at the expense of an increase in latency by a factor of two when the traffic intensity is up to 32% of the baseline.