Research On Mechanisms And Methods Of Resource Adaptation For Smart Identifier Network

Smart Identifier NETwork (SINET) is a novel future Internet architecture, besides solving the problems of scalability, mobility and security, it can also support the intelligent resource adaption inherently through the embedded wisdom, therefore leading to higher network utilization, lower energy consumption and fundamentally realizing the energy-efficient "Green Network". As part of the SINET theory, targeting at implementing the inherent wisdom in SINET, this thesis aims at studying its smart features, manly focus on status-awareness, dynamic resource adaption and hibernate device selection. The main contributions of the thesis are as follows:1. As a foundation for the implementation of the embedded wisdom, the status-aware resource adaptation scheme is proposed. Firstly, the network components can collect network status dynamically, and maintain the status, through embedding the status-aware module into the basic elements, resources managers and forwarding nodes. Secondly, the smart cooperative control protocol is designed for the adaptation layer, which helps controlling the component layer to perform basic data processing, and obtain status. Thirdly, a status-based routing algorithm is proposed in the adaptation layer, for reallocating network resources according to the fluctuation of environmental conditions in SINET. Lastly, the feasibility of our proposed scheme and protocol is demonstrated through the implementation performed at the SINET prototype system. The proposed scheme optimizes the utilization of resources and enhances the user experience.2. Two alternative status message transmission approaches are proposed for further improving the status-aware resource adaptation scheme, namely Inquiry and Report. The first mechanism shows stronger real-time capability, since its resources manager inquiries the forwarding nodes actively, and obtains the static state of the global status, however it involves greater extra overhead on control. The second mechanism has nearly no additional controlling overhead, since the forwarding nodes provide a dynamic state of the local status, but its measurement accuracy is influenced by the type of traffic. A proof-of-concept implementation is built, and their real-time performance and overhead are tested under different scenarios. Finally, we make performance evaluation based on the real data traces from Tier-1 ISP. Considering the tradeoff between overhead and accuracy, we recommend the Report approach as the status message transmission mechanism in SINET.3. A traffic-aware smart path allocation method is proposed, for supporting dynamic resource adaption. SINET has an advantage of being aware of the accurate traffic matrix. Making use of the traffic matrix as a constraint, we model on load balancing routing problem, and transform the primal problem into dual problem by Lagrange function. Then we propose a traffic matrix based load balancing routing algorithm in order to achieve the dual problem optimization goal. Finally, we conducte the experiment simulation and implement the method at the SINET prototype system. The results show that, the method can better reduce congestion and realize load balancing, and its overhead is extremely limited, entirely within the acceptable range.4. An approach focusing on simultaneously optimizing the network power consumption and cooperative cache is proposed, for supporting hibernate device selection. The forwarding nodes not only forward data, but also cache and sleep In SINET. Therefore, a multi-objective mixed integer linear programming model is built for achieving the duple target of energy efficiency and better cache performance. The cache optimization objectives are converted into the constraints, which lowers the difficulty on this problem. And a hibernate node selection algorithm is proposed to optimize the resource adaptation. Finally, based on three real ISP topologies, we conduct a comparative evaluation on this proposed algorithm and the other three, through the simulation of real flow. The results show that, the proposed algorithm has better energy efficiency, and the more redundant links, the better the energy efficiency is. Besides that, the approach has nearly no impact on the remaining active network during low and moderate traffic periods.