Research On Mechanisms For Inter-domain Traffic Engineering In Smart Identifier Network

After fifty years of development,the current Internet has achieved great success.In recent years,with the expansion of the network scale and the diversification of application scenarios,the current Internet has failed to meet the future communication needs.With this background,lots of researchers are committed to develop the future Internet architecture.In order to meet the Chinese future Internet research needs,the Next Generation Internet Research Center of Beijing Jiaotong University has proposed a Smart Identifier Network,i.e.SINET,which seeks to solve the scalability,mobility,safety,and energy saving issues of future Internet.This thesis analyzed and summarized the opportunities and challenges of SINET for achieving traffic engineering.On this basis,this thesis designed and evaluated new mechanisms in SINET for solving the inter-domain inbound traffic control problem,inter-domain outbound traffic control problem,and inter-domain traffic reduction problem by leveraging the potential characteristics of routing,forwarding,network states monitoring and in-network caching of SINET.The main work and innovations are summarized as follow:1.For solving inter-domain inbound traffic control problem,this thesis proposed four algorithms based on traffic monitoring and service size metadata.The proposed algorithms use the receiver-driven communication model of SINET,which realizes the inter-domain incoming traffic control by controlling the forwarding path of Get packets.The core idea of the four algorithms are choosing inter-domain forwarding paths of Get packets with certain probabilities.The difference is that the four algorithms use different information to update the probabilities.The first algorithm uses static probabilities.The second algorithm uses real time traffic.The third algorithm uses service size metadata.The fourth algorithm uses real time traffic and services metadata.The experiment results from SINET prototype show that the proposed algorithms can efficiently and accurately control inter-domain inbound traffic.Compared with the IP prefixes negotiation approach of current Internet,the proposed mechanism can improve the accuracy of inbound traffic control by 56%,and can efficiently handle inter-domain link failures and burst traffic.2.For solving inter-domain outbound traffic control problem,this thesis proposed a Nash Bargaining based inter-domain outbound traffic control mechanism.The proposed mechanism uses services registration messages to announce inter-domain path preferences,and uses Nash Bargaining model to negotiate the forwarding decisions of Get packets so as to achieve outbound traffic control.In the simulations,the reduction of services intra-domain transmission overheads is considered as outbound traffic control profits.Simulation results show that the proposed mechanism does not require neighboring domains to exchange local sensitive information.Compared with selfish traffic control method in a no caching case,the proposed mechanism can increase 60% of domains' outbound traffic control profit by 10%.Compared with selfish traffic control method in a caching case,the outbound traffic control profit of this mechanism is decreasing with the increasement of cache space.The implementation results on SINET prototype show that the communication overhead is 1303 Kbytes/second and the CPU using rate is 16% when services are registered with 8000 times/second,which indicates that the proposed mechanism has good feasibility and deploy-ability.3.For solving the inter-domain traffic reduction problem,this thesis proposed an inter-domain traffic reduction mechanism based on Lagrange dual decomposition and cooperative game theory.The proposed mechanism uses the in-network caching characteristic of SIENT and makes multiple access networks cooperatively decide the cached services,which further improves caching efficiency.The proposed mechanism also makes neighboring access networks share the cached services,which will reduce the inter-domain traffic and transit fee for fetching services from provider network.Simulation results show that the proposed mechanism reduces 3.77 times more inter-domain traffic and transit fee than non-cooperative selfish caching strategy.Compared with the centralized cache allocation strategy,the proposed mechanism can improve 29.6% of fairness at the cost of 9.7% of reduced inter-domain traffic and has better privacy.At the cost of increasing negotiation overheads,the proposed mechanism can run in multiple access networks in a distributed manner and have lower computing overheads and better deploy-ability.For example,when the proposed mechanism runs in 42 access networks with 5GBytes caching capacity,it introduces 2.337 Mbytes negotiation overheads.