A bottleneck-free tree-based name resolution system for Information-Centric Networking

In this paper we propose Ftree, a tree-based name resolution system in which name-records (name-to-locator mappings) are stored only at leaf level. Ftree addresses scalability and bottlenecks encountered at the root level in tree-based name resolution systems and answers some of the requirements for ICN (Information-Centric Networking). Given an object name, Ftree generates a set of keys using multiple hash functions. The name-record performs a key-based routing in a prefix tree and reaches a leaf node, becoming then its resolver. Using the same multiple hash functions, an object requester can select the key that corresponds to the nearest leaf resolver. A requester of a data object can have access to the nearest replica when resolvers hold them in addition to name-records. Moreover, and unlike existing name resolution systems, the root and high level nodes are rarely reached. An analytical study has been conducted to estimate the cost and the load induced by the lookup and registration processes. Through this study, the optimum replication factor has been computed to minimize the network traffic. Experimental evaluations of Ftree are reported, with a comparison with the MDHT, DONA and DNS-like resolution systems. Ftree outperforms these systems in terms of lookup overhead and root node loading.