Efficient Global Network Resource Pre-allocation in SDN based Cloud Centers

In cloud centers, tenants share the network resource and the network mostly adopts shortest-path-first way to transmit data, which will cause unbalance of link load and high probability of link congestion. These will negatively affect the quality of service(QoS) and the profit of cloud providers. To address this problem, this paper proposes a network resource pre-allocation model, which is applied in cloud centers enabled by Software-Defined Network(SDN) technology. Based on the pre-allocation model and tenants' bandwidth demands, cloud providers can assemble enormous different network resource allocation strategies. As finding the optimal strategy in the solution space takes a long time, this paper proposes a network resource pre-allocation method based on genetic algorithm. The method can be deployed in parallel and scalable way to achieve availability and scalability which is required by the networks that are large-scale and have different topologies. The simulation proceeds under a network topology with high-bisection bandwidths and the results show that our method can help cloud centers to achieve load balance and high utilization of network, meanwhile provide more bandwidth guarantee for tenants.