Researches In Resource Allocation Algorithm In Networks Virtualization

With the development of network technology, the Internet system structure hasbeen in "ossification" phase. Network virtual technology succeeds in solving thecurrent Internet "ossification" problem by constructing platform in the currentsystem structure for the deployment and research of various emerging technology.The core issues of Network virtualization are resource allocation and management.Excellent resource allocation can improve the utilization rate of substrate physicalresources and increase the virtual network acceptance rate. Effective resourcemanagement can coordinate the interests between resources suppliers and improvethe resource utilization rate. Therefore, it has great significance of designingexcellent resources allocation algorithm and effective resource managementframework.First, the thesis shows detailed analysis and comparison of existing networkvirtualization resource allocation algorithm. Due to the problems such as the poorcoordination of nodes and link allocation in existing algorithm, poor performance ofdistribution algorithm in sparse topology and so on, the thesis puts forward aheuristic network virtualization resource allocation algorithm (HVNE-Heuristicnetwork virtualization Embedded). The algorithm merges node mapping and linkmapping into a unified process, which improves the performance of general mappingalgorithm in sparse topology. Furthermore, HVNE allows multiple virtual nodes ofthe same virtual request mapping to the same physical node, saving physical linkresources. HVNE combines "the k-division of optimization" theory with thetopology segmentation theory and improves the traditional star segmentation method,which applies to large-scale network. Simulation results show that in the case ofensuring network load, HVNE can achieve great virtual request acceptance rate, highresource utilization and network profit.Second, the thesis raises a concurrent network virtualization resourcemanagement system. It can greatly coordinate resources management and resourceallocation, simplify complexity of the resources management and improve theefficiency of resource allocation. For centralized network virtualization resourceallocation, the thesis puts forward a concurrent distribution mechanism. Throughmultiple resources distributors and resources reserve and locking mechanism, it concurrently distributes resources. Furthermore, in the case of ensuringindependence among all virtual networks, the system greatly improves virtualnetwork distribution rate and the performance of system. Moreover, the thesisdesigns a complete communication mechanism and failure management mechanismto guarantee the stability of the entire system communication, normal run of virtualnetwork when at fault and improve the system robustness. Simulation experimentsshow that when the concurrent distribution mechanism has no impact on theperformance of the virtual network distribution algorithm, it can improve thedistribution rate of the virtual network, perfect utilization rate of physical node andlink and optimize the system performance.Finally, combining with the CloudSim simulation environment, the thesisdesigns and implements the dynamic resource allocation of network virtualdemonstration system. The system can form with graphics; dynamically demo theprocess of the whole network virtualization resource allocation. In addition, thesystem provides an excellent expansion interface and a universal demonstration andanalysis platform for other network virtualization resource allocation algorithm.