Research On Fast Name Lookup In Forwarding Plane Of NDN

With the high frusion of mobile communication and Internet, Mobile Internet will become the foundational platform of information transfer in future communica-tion. However, many problems still require to be further investigated, especially in future Internet architecture research field. Our work focus on a novel Internet architecture called Named Data Networking (NDN). Especially, the efficient storage structures and fast name lookup algorithms of Pending Interest Table (PIT) and Forwarding Information Base (FIB) are studied to realize an enhanced forwarding plane of NDN and meet the requirements of future Mobile Internet. Some innovative works are carried out below.Firstly, a novel data structure named as Mapping Bloom filter (MBF), which is a modified data structure of Bloom filter, is presented. MBF can support querying and mapping the set elements in the memory by exploring a Mapping Array (MA). That is, MBF makes up for the shortage of Bloom filter, which cannot be directly implement-ed as the summary. The evaluations indicate that MBF has the efficient performance as the summary and can be deployed on SRAM, RLDRAM and DRAM.Secondly, an inproved storage structure named as MaPIT are proposed based on MBF for the forwarding plane of NDN. Through the multistage storage structure, the MaPIT can realize the fast name lookup, reduce the frequency of memory access, and minimize the on-chip memory consumption. Meanwhile, the off-chip memory consumption has been further minimized over the traditional PIT by redesigning the name lookup algorithm. It allows MaPIT to use SRAM as on-chip memory and satisfy the requirements of current network. The experiment results demonstrate the effectiveness and efficiency of the proposed MaPIT.Finally, a novel storage structure named as MaFIB is proposed by exploring MBF for the forwarding plane of NDN. MaFIB achieves the forwarding information retrival for variable-length and hierarchical names based on the traditional FIB in TCP/IP network. The evaluations indicate that MaFIB can greatly reduce memory costs and achieve on-chip memory deployment on SRAM. Furthermore, MaFIB has the excellent probability of false positive performance, which can meet the requirements of current network.