Research On Survivability Of Service Carrying Networks

Taking the network resources efficiently utilized and network services fully matched as goals, Service Carrying Networks(SCN) was proposed by the fundamental of National 973 Program ―Flexible Architecture of Reconfigurable Infrastructure‖. SCN could support the diversified and omnibus QoS requirements of the present and future business. However, the unstable physical networks restrict the popularization and implementation of SCN. With the fundamental of National 973 Program, this thesis focuses on improving the survivability of SCN in two aspects, improving the reliability of physical nodes and fault recovery efficiency.Firstly, aiming at improving survivability of SCN, a SCN structure with survivability guarantee is proposed. Secondly, an embedding algorithm is given to assigning substrate network resources to SCN components. And then, a fault recovery algorithm is proposed to recover the fault SCN when the physical fault happens. Finally the Service Carrying Network Survivability system(SCNS) is given to check the feasibility.1. A SCN structure based on node reliability is proposed. The structure takes the reliability of physical nodes as new constraints to construct SCN with survivability guarantee. In order to evaluate the reliability of physical nodes, a Node Reliability Evaluation Algorithm based on Capability and Behavior(NREA-CB) is given. NREA-CB initializes the reliability based on the capabilities and regulates its based on behaviors in time. Simulation results show that NREA-CB could response to the node faults in time and have more stable change compared with the existing algorithm. The SCN based on node reliability could achieve 20% fault number decrease during the work time. The survivability of SCN is improved by improving the reliability of physical nodes.2. A SCN Embedding Algorithm based on improved Harmony Search(SEA-HS) is proposed. SEA-HS improves the basic HS by generating the initial harmony memory based on greedy algorithm and tuning the parameters dynamically to accelerate the iterative speed. SEA-HS completes the virtual link embedding process based on node load gradient, which could not only reduce the resource cost, but also balance the load of physical networks. Simulation results show that SEA-HS could achieve 10% Request Accepted Ratio(RAR) increase and 5% Revenue/Cost rise compared with D-ViNE-LB and D-ViNE-SP algorithms. Because the number of available node decreases, the RAP will decrease to 66% when considering the constraint of node reliability.3. A SCN Fault Recovery Algorithm based on Equivalent Resource(FRA-EA) and Network Regulation Algorithm(NRA) are proposed. FRA-ER converts the recovery problem to finding equivalent resource problem and recoveries all or part of the fault SCNs by once, improving the recovery efficiency and reducing the service interrupt time. NRA detects and regulates the SCNs periodically to optimize their structures and reduce bandwidth cost, improving the carrying capability of substrate networks. Simulation results show that compared with the conventional overall re-mapping algorithm and fast recovery algorithm, FRA-EA could achieve 15% recovery time reduction. NRA could reduce the bandwidth cost of physical network, achieving 5% total bandwidth cost reduction, improving the recovery success ratio by 10%. The survivability of SCN is improved by increasing the fault recovery efficiency.