The Research Of Key Technology Of Virtual Network Resource Scheduling And Collaborative Optimization

With the advent of the information age, the demands of Large-scale data storage, computing and network services have increased rapidly. The data center as a platform for data storage and web services has attracted more and more attention. With the development of the server virtualization technology, the data center network has become the bottleneck of the improvement of the utilization of physical resources and the reliability of the data center. Researchers have made many researches on data center network to make full use of the network resource and provide minimum bandwidth guarantees. Researches do research on importing wireless, optical hybrid switching and other network media to data center; according to the drawback of the traditional data center network, researches explore new network architectures and data center network topology for the purpose of improving the reliability, availability and utilization of the network. Taking the poor utilization of network resources into consideration, people invert the technology of network virtualization, the resource allocation of the virtual network, packet scheduling and so on to implement the sharing of network between multi-tenants and the high utilization of the network resource.Researches indicate that the internal traffic reaches about 80% of total traffic in datacenter, and the throughput of the internal traffic became critical to the network performance. Many researchers focus on data center network virtualization to improve the efficiency of communication between VMs and the high utilization of network and resources proportional allocation between tenants. However, the traffic of VMs is dynamic and unpredictable, and it’s necessity to adjust of the bandwidth dynamically. There are a large number of switches in data center and the utilization of the device directly affects the efficiency of the data transmitting. At the guidance of the end to end principle, the switches ignore the difference between flows and regard the flow as the a undifferentiated binary data stream, which results the variety of network devices resources in switch could not be fully utilized at the same time. The bottleneck of a kind of resource would decrease the utilization of the switch, which affectes the throughput of the network.. The paper made the following research on these two problems.First, according to the statistics of the datacenter network traffic characteristics from other researches, the actual demand bandwidth to a VM is changeable, and now the datacenter just set a maximum bandwidth to a VM. The bandwidth of a VM cannot be changed dynamically according to its actual needs, which loss the flexibility of network bandwidth management. After taking the consideration of the utilization of the network resources and VM bandwidth guarantee, we propose a algorithm to adjust the bandwidth of a VM dynamically according to the needs. With the help of the SDN, we implement the algorithm to test the effectiveness of the algorithm and found it’s effective.Second, as different network flows consume different proportions of switch resources, as scheduling flows according to the resource consumption of network flows, we propose a collaborative multi- resource scheduling mechanism to optimize switch resource management. By monitoring the status of the network device resource utilization when the network congest and feedback to the sender, to slow down the sender’s forwarding packets speed if it consumes the bottleneck resource by monitoring the status of the network device resource utilization. Through the buffer queue management, network flow adjusted to different sequence to improving resource utilization of network devices and the efficiency of network transmission.