Improving NDN forwarding engine performance by rendezvous-based caching and forwarding

Abstract Named Data Networking (NDN) is a new network architecture that shifts the communication paradigm from end-to-end communication to large-scale content retrieval. Unlike TCP/TP, NDN replaces the networking waist from IP addresses to content chunks. Data transmission in NDN is realized via name-based routing and forwarding. To carry out packet forwarding, each NDN forwarding engine maintains three data structures: a Pending Interest Table (PIT), a Forwarding Information Base (FIB), and a Content Store (CS). Since the name space of NDN is much larger than the IP address space and names in NDN can have a number of name components and be of variable length, it poses great challenges to store these tables in commercial NDN routers and meet the line-speed forwarding requirement.  In this paper, we proposed a caching and forwarding scheme based on rendezvous points. In this scheme, requests for same or similar contents are forwarded to the same router, to mitigate the explosion problem of the three data structures. To realize our detoured forwarding without disturbing NDN’s forwarding logic, we proposed a mechanism to add detoured routing capability to NDN. Furthermore, by using our RP-based caching and forwarding, each router serves only a subset of the whole name space, so we added route cache to the FIB design to speed up packet forwarding, and proposed two approaches to tackle the cache hiding problem rooted from the longest prefix matching (LPM) on names of arbitrary length: the atomic route caching that groups multiple overlapping prefixes as a unit for cache operation, and the on-the-fly route caching that caches only the most specific prefixes. Trace driven simulations using HTTP request traces from China Mobile and Nanjing Normal University(NJNU) demonstrated the feasibility and effectiveness of our design.