Network virtualization can improve the utilization of a physical network by enabling the coexistence of multiple virtual networks supporting different services. For an elastic optical network, network virtualization would create a number of virtual optical networks (VONs) embedded in the same physical optical network. Conventionally, each of these VONs is allocated with a fixed amount of network resources. However, the traffic demand on a VON is usually dynamic, varying over time. This leads to a mismatch between the fixed resource assigned and the actual resource required by users. Therefore, some user services may be blocked and the quality of service provided by each VON may be degraded. To overcome the disadvantage caused by this mismatch, we propose a network resource-sharing scheme, called spectrum trading (ST), which trades spectrum resources between different VONs based on their actual capacity requirements, for better overall spectrum utilization. Integer linear programming (ILP) models are developed to model the different ST scenarios, and heuristic algorithms are developed to tackle ST in large networks. Simulation studies show that the proposed ST scheme is effective in significantly improving the total traffic demand carried by VONs by up to 20% in the test networks considered, while not bringing in any additional hardware cost. (C) 2020 Optical Society of America
展开▼
