数据命名网络上的一种存储高效、可并行的数据名查找方案.pdfVIP

Abstract Named Data Networking (NDN), as one of the most important proposals for future Internet architecture, is naturally suited to support Internet ’s functionality transition: from host-centric transmission to content-centric communication. Different from today’s TCP/IP network architecture, NDN uses content names as identifiers of data packets instead of IP addresses, thus routers in NDN route and forward packets based on content names. Since content names have variable and unbounded lengths, a NDN name routing table could be much larger than an IP routing table, which makes efficient storage of a name routing table and fast name lookup in NDN as challenging tasks. To overcome these challenges, this paper proposes an efficient parallel name lookup scheme. The main work and contributions of this article are as below: (1) A component hash method is proposed to reduce memory requirements, which takes Multiple-stride Aligned Transition Array (MATA) as a main data structure. (2) Based on the Split Character Trie model in IP network and the component hash method, Split Name Trie (SNT) is proposed. It splits a name routing table into two small ones to reduce memory requirements. When operating a name lookup, SNT first splits a name into two parts and then looks up them on the two small name tables independently. SNT can reduce memory requirement of a routing table and guarantee the name lookup speed. (3) To verify the functions of the above two techniques, several experiments are designed and implemented. The results show that SNT can improve general memory efficiency on NDN name lookup by 23% ~ 49% 。 Key words: NDN, Nam