Light-tree Based Multicast Flow Aggregation Scheme In Elastic Optical Datacenter Networks

With the rapid development of cloud computing,big data and the Internet,the data center has developed rapidly as a physical facility to provide storage and computing resources.The data center has become the main mode of data center interconnection.Because of the characteristics such as large capacity,low delay and low energy consumption,optical interconnection technology has been widely applied to data center network,namely elastic optical datacenter network.So far,the main load mode of business in the elastic optical datacenter network service is concentrated in the light path,namely business loading is creating an end-to-end path for each business.In order to meet the needs of network data flow grow with each passing day,how to improve the resource utilization and save the cost of network has become an important problem that the elastic optical datacenter networks faced.The elastic optical datacenter network revolves around the multicast flow and the amount of data transmitted by 80%is less than 10kb,which we call "the ant stream".With the increase of network traffic,this paper puts forward a light-tree based multicast flow aggregation scheme in elastic optical datacenter networks,by aggregating the multicast flow in a certain strategy,then transmitting through light-trees to achieve improve spectrum efficiency and reduce the network cost.Using light-tree replace light path can effectively save spectrum resources,and aggregating the multicast flow before transmission can reduce the number of transceiver,that is reduce the network cost.In our paper,the current elastic optical datacenter network is confronted with the present situation of a large number of small size multicast flows,we are committed to the study of multicast flow aggregation technology.An ILP model based on multicast flow aggregation in elastic optical datacenter networks is proposed,then put forward the concept of user-coverage and the heuristic algorithm based on user-coverage,finally we evaluated the performance of the ILP model and our heuristic algorithm.The results show that the ILP model based on light-tree and the heuristic algorithm based on user-coverage have good performance under the elastic optical datacenter networks.Compared with building light-tree directly,it can save a lot of network costs at the expense of a small amount of resources.