The need for a name to MAC address mapping in NDN: towards quantifying the resource gain

Abstract

In this paper, we start from two observations. First, many application scenarios that benefit from ICN involve battery driven nodes connected via shared media. Second, current link layer technologies are completely ICN agnostic, which prevents filtering of ICN packets at the device driver level. Consequently, any ICN packet, Interest as well as Data, is processed by the CPU. This sacrifices local system resources and disregards link layer support functions such as wireless retransmission. We argue for a mapping of names to MAC addresses to efficiently handle ICN packets, and explore dynamic face-based mapping schemes. We analyze the impact of this link-layer adaptation in real-world experiments and quantitatively compare different configurations. Our findings on resource consumption, and reliability on constrained devices indicate significant gains in larger networks.