Research On Technology Of Service Self-adaption Reconfigurable Service Carrying Network

With the new services rising and application scales extending rapidly, the Internet has benefited everyone‘s life. Due to the thin-waist construction of IP carrying mode and its ossified construction, it has been overburdened over the years. With the high-speed development of Internet information industry, traditional network architecture not only cannot satisfy the carrying demands of constantly expanding user services through upgrading and expanding technologies, but also can not satisfy the requirements of universal inter-connection, heterogeneous convergence, manageability and controllability, etc.Combined with the fundamental technique research task in the ―Reconfigurable Information Communication Substrate Network Architecture‖ project belonging to the National Basic Research Program of China(973 Program), the reconfigurable information communication substrate network architecture is designed to bridge the gap between network capability and service requirements through network construction self-organization, the function self-adjustment and service self-adaptation, with which, the network can adapt to the varying requirements. So, on one hand, the reconfigurable network needs to satisfy service requests and provide customized services, on the other hand, it can optimize the network resource configuration according to the dynamic change of the network service and significantly improve the resource utilization of the network.Based on the above-mentioned considerations, this paper is centered on strengthening dynamic service provision capacity of network, targets on the optimal matching between diversified service carrying requests and network resource provisions, and provides diverse network services by constructing the reconfigurable service carrying network(RSCN) dynamically. It researches on the key operational mechanism of the RSCN which supports the dynamic adaption between the service and resource. Its main works and achievements are outlined as follows:1. Aiming at eliminating the defect of neglecting dynamic change of the service request while constructing the carrying network, we propose a Dynamic Service Request-oriented RSCN(DSR) algorithm based on the requirements of applications. This algorithm defines the construction expenditure of the RSCN based on the dynamic characteristics of service request and targets on minimizing the construction expenditure, designs the corresponding construction algorithm of the RSCN according to the type of service request and its changing state. The simulation results show that the proposed algorithm can satisfy the requirements of applications, achieve higher success ratio and gain higher revenue/cost ratio comparing with the ex-isting algorithms.2. Considering the competition of various applications sharing the limited substrate network resources, an algorithm of Request Competition-based RSCN(RCR) is proposed. This algorithm conceives an auction-game model by introducing game theory. By means of periodical mapping, it clusters the service requests in a time period, effectively balances the tradeoff between the maximized construction effectiveness and the minimized construction request and the waiting time. Meanwhile, this algorithm perceives the service change in time, dynamically adjusts the RSCN construction, and finally reaches the profit maximization of the whole system. The experiment results show that comparing with the ViNEYard and G-SP algorithms, the proposed algorithm has relatively gained higher revenue of the RSCN even though the special setting for the experiment gets lower success rate.3. To effectively utilize the limited substrate network resources, an algorithm of Dynamic Topology Awareness-based RSCN Reconfiguration(DTAR) is proposed. This algorithm considers the number of shortest paths of nodes or links as measurement index of resource criticality, prioritizes mapping on the non-critical resources which satisfy the service request. It can effectively avoid the emergence of the substrate network bottleneck. It detects the critical resources by means of dynamically cognizes the states of critical resources, reoptimises the RSCN according to different service requests, and effectively improves the success rate. Experimental results show that comparing with the existing algorithms, the proposed algorithm achieves higher success ratio, and gains higher revenue/cost ratio and load balance for substrate network.4. Considering the negative effects of substrate network failure, we propose a Self-healing RSCN algorithm based on Network Resource Dynamic Awareness(NRDAS). This algorithm extracts network resource constraints by dynamic awareness of substrate network resources, and reduces the RSCN construction cost effectively. We introduce the Q-Learning method and build a model of Self-healing RSCN. It can make a dynamic regulation once network congestion or failure occurs, and realize the effective carrying for different requirements of applications. The simulation results show that the proposed algorithm can reduce the resource consumption of RSCN and achieve better performance on service adaptation and failure recovery.