Research On Key Technologies Of Quality-of-service In Next Generation Internet

The most important research focus in the field of Next Generation Internet (NGI) is to research and construct new network architectures to meet the requirements of users and services in the future Internet, in which QoS (Quality of Service) guarantee is also one of the most fundamental network services. A series of new characteristics and enhanced functionalities are required in the future Internet, which include higher forwarding speed, larger address space, and abilities of self-management and self-adaptation. Increasingly diversified service requirements and more complex dynamic network environments of the future Internet make it a challenging task to design efficient QoS management architectures and high performance QoS guarantee mechanisms in the next generation Internet.In this dissertation, several key technologies of QoS guarantee in the next generation Internet are studied in detail. For requirements of high speed packet processing in the QoS management of future Internet, high performance QoS guarantee mechanisms are studied in the dissertation, and novel high speed packet classification algorithms are proposed. To address QoS issues in the increasingly complex and dynamic environment of the future network, autonomic attributes are introduced into the architecture design of the next generation Internet and its QoS management framework. Novel solutions of QoS management and context-aware autonomic QoS mechanisms (packet marking and queue management) are proposed. The main contributions of the dissertation include:(1) A high speed packet classification algorithm, ERFC (Enhanced Recursive Flow Classification), is proposed to solve the problem of high space and preprocessing complexity in the original RFC (Recursive Flow Classification) algorithm. In ERFC, a hash-based aggregated bit vector scheme is exploited to speed up its preprocessing procedure, and a compressed and cacheable data structure is introduced to decrease total memory requirement and improve searching performance. Evaluation results show that ERFC provides a great performance improvement over RFC in space requirement, preprocessing time and search speed.(2) A novel decision tree packet classification algorithm based on Efficient Multiple Bit Selection (EMBS) is proposed. In the proposed algorithm, the process of building the decision tree is transformed to a sequence of prefix bit selecting procedures. A Performance Estimate Function (PEF) and an efficient bit selecting algorithm are exploited to determine which selection will lead to an efficient decision tree with high performance. EMBS algorithm is capable of handling range match fields, is suitable for IPv6packet classification as well as IPv4. Evaluation results show that EMBS provides a great improvement over recent decision tree based algorithms in both space requirement and searching performance.(3) Some key issues in the autonomic DiffServ QoS management architecture are studied, which include QoS management framework, context awareness and dissemination, and design of autonomic QoS mechanisms. In this dissertation, a context-aware packet marking (CAPM) algorithm is proposed. CAPM collects various types of contexts, such as service status, network conditions, and semantic priorities of packets in service flows. Based on context analyzing and autonomic feedback controlling, it is capable of adaptively adjusting its behavior to provide better QoS guarantee. Simulation results show that, CAPM provides better transmission quality for multimedia streaming video services than traditional packet marker, and significantly improves user’s quality of experience.(4) In the autonomic DiffServ QoS management framework, a context-aware queue management (CAQM) mechanism is also proposed. CAQM is capable of collaborating and exchanging contexts with CAPM, and provides differentiated packet processing based on semantic contexts. In CAQM, a multi-level sub-queue structure is used to manage different packet priorities in aggregated flows. In order to improve the transmission quality of critical data packets of service flows, the congestion control of CAQM is performed based on packet priorities which imply service semantics. Simulation results show that, collaborating with CAPM in autonomic DiffServ network environment, CAQM will further improve the transmission quality and user experience of streaming video service.