Research On In-network Caching And Evolution Technology Of Named Data Networking

A data-centric future network architecture,Named Data Networking(NDN),has been developed in recent years and rapidly becoming a hot research topic in both academia and industry.Taking advantage of name-based routing and in-network caching,NDN breaks IP-based connections which bind content with its host,and achieves efficient data distribution through the caches equipped in routers.These merits upgrade current Internet from "communication tube" to "content repository".Considering the limitation of cache capacity,access speed and energy consumption,efficient content allocation schemes need to be explored to maximize caching benefits.In order to verify the caching strategies,the most effective way is to run the proposed mechanisms in real-life networks.However,NDN research is still on the theoretical stage,an incremental evolution approach is required to implement NDN in the existing IP-dominating network.For this purpose,this dissertation investigates the caching schemes and evolution techniques of NDN,and makes the following contributions:1.This work builds the tree-based caching model within an ISP domain and proposes two online coordinated caching schemes: AsympOpt for single-gateway ISP and EMC for multi-gateway ISP,respectively.The schemes precisely choose the caching location for every content within the network by taking a variety of factors into consideration,including content popularity,requests distribution in different branches,caching redundancy and overhead.Evaluations show the proposed schemes can significantly reduce both the ingress/egress traffic and the content download delay.2.This dissertation reveals that the Interest aggregation effect of PIT is very weak(when CS uses LRU policy),based on which the preliminary energy consumption model of NDN is built.Further,this work proposes an energy-efficient caching heuristic named EEC,which weighs content by energy-oriented caching benefits and utilizes a windowbased dilution structure to keep the timeliness.Simulation results indicate that EEC can generally reduce network energy consumption up to 70%.In addition,this work explores the energy consumption trend as in-network cache capacity increases and presents suggestions for making an energy-efficient router cache size.3.This work investigates incremental deployment of NDN in Ethernet.To support both IP and NDN traffic,this work designs the architecture of Dual-Stack switch,lays out three NDN deployment scenarios,examines the technical issues involved in each scenario and presents solutions.In particular,in the hybrid network consisting of Ethernet switches and Dual-Stack switches,this work proposes heuristics to optimize the placement of DualStack switches.Through utilizing links which are not on Ethernet Spanning Tree,the heuristics can balance traffic load,reduce data download delay and content server pressure with only a small budget of NDN-enabled nodes.