Modeling CCN Packet Forwarding Engine

With the in-network caching capability embedded, packet forwarding in CCN (content-centric networking) becomes rather sophisticated. To enable wire-speed forwarding, previous works have reported exciting component-level performance achievements. However, a proper system-level model for exact bottleneck identification and accordingly performance tuning is still absent. In this paper, we build two such models dedicated to the two common implementation variants of a CCN router, i.e., the pipeline model and the run-to-completion model, respectively. By carefully investigating and analyzing the interactions between FIB (forwarding information base), PIT (pending interest table) and CS (content store) in the two models, we quantitatively identify that CS is the exact performance bottleneck of the entire CCN packet forwarding engine. This conclusion is very timely (if not too late) because in the past, researchers invest a lot of time and effort in optimizing FIB and PIT while very limited for CS. According to the mathematics, we suggest that the research community should shift more attention to CS performance tuning or totally rethink the entire packet forwarding architecture.