Research On Virtual Network Mapping Based On QoS

Network virtualization technology can support multiple heterogeneous virtualnetworks (VNs) to coexist on a shared infrastructure in an open, controllable, andefficient way. The VN mapping problem is the key step to achieve this technology.Recently, the traditional network architecture is meeting the technical bottlenecksbecause of the difficult deployment in network security, reliability, service model, andthe new application. In order to overcome the current ossification of the traditionalnetwork, cloud network architecture was proposed as the next generation networkarchitecture. Compared to the traditional network architecture, cloud networkarchitecture can support various new network applications, which has more computingand storage resources simultaneously. With the increasing application types in cloudnetwork, different kinds of VN applications require various quality of service (QoS).This thesis analyzes the existing VN mapping models and algorithms. firstly. Thedefects of the present VN mapping algorithms have been pointed out respectively.Based on the research of cloud routers and cloud switches, the basic characteristics ofthe routing node resources and the link resources is analyzed. And the VN mappingproblem under the cloud network architecture is formulated. Considering the variousapplications demands in QoS, a QoS guaranteed scheme for VN mapping problembased on the cloud network architecture is conceived. The proposed scheme canguarantee the delay requirements of different applications. And the genetic algorithmand the greedy algorithm have been used to solving the VN mapping problem.In order to evaluate the algorithm performance under different networkarchitectures, we use the MFC class to develop a virtual resources schedulingsimulation platform in VC++6.0environment. Then, we finished the debugging,running and results output for different virtual resources configuration. The features ofthis platform are configurable network parameters, data processing, charting, algorithmperformance analyzing and VN monitoring.