The Network Resource Allocation And Management To Guarantee Qos Based On SDN

The traditional QoS mechanism is designed for the Internet,based on the Internet's fully distributed,hop-by-hop routing architecture.It lacks a global view of the overall network resource distribution such that it is difficult to be realized in the real networks and Internet.SDN(Software Defined Network)has the characteristics of centralized control.It can easily deliver QoS policies through centralized controller,and can realize centralized management and control of all network devices and the whole network traffic.It can implement flexible QoS service policy selection and QoS policy consistency..Therefore,once the QoS mechanism in SDN is put forward,it gets the attention of academia and industry,and becomes a research hotspot.With the cloud computing,mobile Internet,Web 2.0 and other emerging businesses to flourish,to ensure data center network and its application quality of service has become the industry consensus.In the data center network,many well-known application forms,such as Web search,advertising,recommendation systems have similar business flow characteristics,that is,long flow and short flow mixed business flow.The short flow is sensitive to delay,but the bandwidth is not high;the bandwidth of the long flow is high,but it is not sensitive to the delay.The delay of the short flow directly affects the quality of the returned result and the final economic benefit.So how to protect the delay of the short flow has been a hotspot.In the first part of this paper,we propose a QoS mechanism to guarantee short flow delay in SDN.For the current network node output port processing capacity with non-preemptive features,that is,at a time can only handle a packet,and can not be preempted.This mechanism combines hierarchical queue management and dynamic routing mechanisms.When the output port is occupied by the long-flow packets,the routing scheme is adopted,and the optimal route is selected according to the occupancy of other output ports.When the output port is not occupied by the long-flow packets,the hierarchical queue management is adopted to ensure the priority of the short flow.Simulation results show that the proposed scheme can significantly reduce the transmission delay of short flows.Video streaming media has become the main application of Internet.In order to provide better network service to video users,it is an important task for network providers to understand and guarantee the service quality and user experience of video streaming.As playing video,rebuffering will directly affect the user's viewing experience,thus affecting the user to watch the video frequency and duration,rebuffering events are often used as a measure of the main indicators of user experience.In view of the fact that DASH(Dynamic Adaptive Streaming over HTTP) is the main transmission technology of video streaming media,the second part of this paper proposes a method to predict the video card according to the QoS parameters of DASH video stream.The method uses the machine learning technology to establish the classification model by using the traffic data in the two directions between the video client and the server to predict whether or not the video is rebuffering.The results of prototype experiments show that the accuracy of our method is 98% and the missed rate is only 3%.