FFRD: Fragment forwarding and reassembly decoupling based chunk transmission in NDN

Abstract

In-network caching is an inherent feature of Named Data Networking (NDN), and the basic data unit of naming and caching in NDN is called “chunk”. However, a chunk needs to be further fragmented into fragments when its size is larger than the link layer's Maximum Transmission Unit (MTU). Furthermore, fragments also need to be reassembled into the original chunk at intermediate routers so that subsequent requests can be satisfied by the cached copy. The current NDN design adopts a coupled hop-by-hop reassembly mechanism where fragments can be forwarded to the next hop only if the chunk has been fully reassembled, which leads to a significant end-to-end delay when large chunks are transmitted due to the processing delay at intermediate routers. In this paper, we propose a reliable and fast chunk transmission protocol based on Fragment Forwarding and Reassembly Decoupling (FFRD) at intermediate routers in NDN. In FFRD, fragments are forwarded to the next hop upon being received and reassembly occurs after all fragments are received. Meanwhile, FFRD can timely detect and recover packet losses at intermediate routers to minimize the transmission delay. The simulation results show that FFRD can significantly reduce chunk retrieval delay and decrease end-to-end Interest packet retransmission times, especially over lossy networks with non-negligible packet losses.