Research On The Energy-efficient Routing Algorithm For Data Center Fat Tree Network

As a commercial computing mode, cloud computing, which provides a convenient way of utilizing the Internet, is becoming more and more popular. In cloud computing, shared resources can be allocated according to needs. Data Center, as the infrastructure platform of cloud computing, is growing exponentially in recent years. At present, traditional electronic exchange data center is facing many problems, such as complex wiring, low bisection bandwidth and high energy consumption. In this thesis, we focus on the energy consumption of data center network.Firstly, we introduce the current native and abroad research on data center energy consumption in recent years. The components of energy consumption in data center is discussed. Also, we classify and summarize the existing data center network and server energy-saving technologies in detail. In this thesis, an energy-efficient routing algorithm base on traffic restriction is proposed. This energy-saving routing algorithm jointly considers both the network equipment's power consumption and working hours. In the proposed algorithm, flows will be divided into the network-limited flows and the serverlimited flows according to the number of the flows passing through a switch port. For the network-limited flows, we adopt flow balance routing algorithm to choose a path with low load and to avoid network congestion. For the server-limited flows, we adopt flow aggregation routing algorithm to completely use the network resources, thus the proposed algorithm reduces the energy consumption of the network. Finally, we build the energy consumption model based on the Fat Tree network. Compared with the existing energyaware routing, the flow-limited energy-aware routing is demonstrated to save more energy.Compared with the traditional electronic exchange data center network, optical interconnection data center network has a considerable advantage. An optical/electrical hybrid Fat Tree network is a form of common network architecture. Based on optical/electrical hybrid Fat Tree network architecture, we build the energy consumption model. The model takes the power consumption of the electrical switches, optical switches and the corresponding optical transmission equipment as inputs. Through respectively analyses on the energy consumption of optical network and electrical network in the optical/electrical hybrid Fat Tree network architecture, we can estimate the whole energy consumption of the network. According to the simulation analysis, compared with traditional electronic exchange Fat Tree network, the optical/electrical hybrid Fat Tree network has a good energy-saving property under different traffic loads and different network scales.