Research On Service Carrying Network Embedding Technology In Reconfigurable Network

With the extension of Internet scale and emergence of new network services, it is difficult to satisfy the various requirement of new services with the fixed network architecture. To solve this problem, the Reconfigurable Network introduces the notion of network virtualization which logically divides the network providers into infrastructure providers and service providers, and provides the adaptive basic network services to users by constructing Service Carrying Network (SCN). Indeed, the SCN embedding scheme has an important impact on the utilization efficiency of network infrastructure resource.This paper is supported in part by the National Basic Research Program of China (973 Program) —"Research on Reconfigurable Information Communication Basic Network". Since the SCN construction is an important research item in that project, this paper researches the SCN embedding technology in different network environments, and designs the supporting mechanism for the SCN embedding. At first, a resource monitoring mechanism and a service clustering algorithm are proposed. Then, three SCN embedding algorithms are designed for the different network management models, which include:centralized model in single domain, distributed model in single domain, distributed model in multiple domain. The main achievements and contributions are outlined as follows:1. In order to efficiently generate the SCN embedding scheme, the management node needs a real-time resource monitoring mechanism. For this requirement, a network resource monitoring framework is designed in this paper which reduces the traffic loads of state information notification by setting monitoring agents. To obtain the best monitoring agents deploying scheme, this framework formulates the issues of deploying monitoring agents as 0-1 programming problem, and solves it by an advanced quantum genetic algorithm. In addition, a dynamic monitoring agents reassigning policy is designed for reducing the impact of topology change on the stability of the monitoring system. Simulations demonstrate that the framework does not need artificially adjusting the number of monitoring agents, and has outstanding performances on communication cost.2. In order to construct the adaptive SCNs for various types of services, it is necessary to cluster the services based on their QoS requirement. By detailed analysis of the requirement of services clustering and the characteristics of clustering algorithms, BIRCH algorithm is chosen for the services clustering. Since this algorithm has a poor performance in identifying clusters of arbitrary shapes, an advanced algorithm named Advanced Split BIRCH (AS-BIRCH) is designed. Experimental results demonstrate the effectiveness of AS-BIRCH in services clustering.3. In order to improve the effect of transmission quality optimization in the Reconfigurable Network, a SCN embedding algorithm for services aggregation is proposed following a new principle-the service aggregation embedding principle. This algorithm tries to embed the SCN requests belonging to the identical type onto the same group of substrate infrastructures, which enables more services to benefit from optimization. Simultaneously, it achieves outstanding performance on acceptance ratio, load balance, and average length of virtual links. Furthermore, a reconfiguration algorithm is developed, which is able to reduce the number of overloaded nodes and congested links by migrating virtual nodes and links with lower reconfiguration cost.4. If the management node is broken or the management model is distributed, only distributed SCN embedding algorithm can be used to embed SCN. But the existing distributed embedding algorithm has the problem of high communication cost and high embedding cost of virtual links. At first, the difference between centralized embedding and distributed embedding is analyzed. And then, the paper depicts the character and application of distributed embedding. A distributed SCN algorithm based on negotiation is designed. This algorithm implements the embedding process by the nodes negotiation in a limited scope, and makes improvement in reducing communication cost. Moreover, this algorithm introduces the conflict avoidance mechanism for supporting parallel process. Simulation experiments demonstrate that the proposed algorithm can achieve better performance on efficiency of resources utilization and communication cost.5. Inter-domain embedding is able to offer more substrate resources for SCN, but it is also necessary to consider the revenue allotment. Therefore, an inter-domain virtual network embedding policy for revenue maximization is presented to satisfy the selfishness of network providers. This policy controls the process of inter-domain embedding by auction, then formulates the local embedding problem as an integer programming model and solves it by the genetic algorithm. Evaluations show that this policy not only increases the revenue of network providers, but also reduces the embedding cost.