Abordagens exata e heurística para o mapeamento de redes virtuais

Abstract

Network virtualization is a technique that allows the emulation of multiple virtual networks, so that different networks coexist simultaneously on the same physical network structure. Thereby, routers and physical links can host a set of virtual networks as long as they do not have their processing and bandwidth capabilities extrapolated. Such action is beneficial by allowing the freedom upon of network protocols rigidity, facilitating the new web technologies development. In this purpose, define which physical equipment set will host one or more virtual networks is a complex task, problem known a Virtual Network Mapping (VNE). The VNE is belong to the class of NP-difficult problems, so its exact resolution may not be feasible for large scenarios. Most of literature existing works presents the virtualization only in a single network domain, this work is distinguished by presenting integer linear programming model and heuristic for the resolution of the VNE in environments managed by one or more network domains. The models dealt with present network virtualization challenges and are differentiated in the proposed objectives: maximize the load balancing, minimize the use of electricity, maximize the profit of the service providers and minimize the exchange of messages between different domains. At the end of this work, simulations are done for both the exact and heuristic approaches. Experimental results show the performance of these approaches in online, periodic and offline scenarios, using CPLEX 12.6 software in the exact model. In this work it is still proposed the application of the heuristic approach in a multidomain and online environment, through the control of an orchestrator. A comparative study regarding the implementation of three different orchestration models is presented: total knowledge, partial and without knowledge of the internal physical network infrastructure of the domains. Experimental results show the performance of the orchestrator along with the heuristic, reporting energy consumption, profit generation and message exchange metrics for the different approaches.