Joint Routing And Layer Selecting For Scalable Video Transmission In SDN

With the rapid development of network multimedia, the video service in the net-work transmission is playing a more and more important role. However, different pref-erence of client, heterogeneous terminal equipments and varying network condition are challenging for the transmission of the video stream. Scalable Video Coding (SVC) technology can code the video stream into a base layer and several enhancement layers and it allows adjusting the number of sent video layers to solve above problems. In the traditional network, researches about scalable video streams are often the layer adap-tive policies of video streams. Because of the transparency of routers for SVC streams, it’s difficult to control the transmission path of SVC video streams in the traditional network. Software defined network (SDN) is an emerging Future Internet technolo-gy. The controller can centrally control the network and establish the corresponding forwarding flow tables in switches to control every flow. Therefore, there have been some researches to study the in-network control policy of SVC streams in SDN. How-ever, there are few papers to study the joint policy between the layer adaptation and the in-network control of SVC streams. Therefore,to sufficiently utilize the capability of network and improve the quality of SVC video, this thesis will propose a joint decision policy to dynamically adjust the number of video layers to be sent and the transmission path of each selected video layer in SDN.Firstly, this thesis proposes an OpenFlow-based Scalable Video Transmission Sys-tem. In this system, the controller collects the SVC application information as well as monitors the network state, and then dynamically calculates the number of video lay-ers to be sent, the transmission path of each selected video layer and its corresponding QoS guarantee policy. After receiving the information about the number of video lay-ers to be sent, the video streaming server immediately sends the corresponding video layers to the network. Also, the controller establishes the corresponding forwarding flow tables to control the transmission path of each video layer as well as configures the corresponding queues for QoS guarantee in switches.Secondly, this thesis formulates the joint decision problem as an MDP-based mod-el. In this model, the state is composed of the non-SVC traffics in each link among all N candidate transmission paths; the action is the vector which is composed of the path number of each video layer; the reward function considers the QoE and the loss rate of non-SVC flow; the objective is maximizing the long-term total expected reward. Be-cause of the unknown transition probability, Q-learning algorithm is employed to find optimal policy.Finally, we deploy the experiment in the simulation platform Mininet to evaluate the performance of our MDP-based transmission policy. This experiment shows that this policy can dynamically adjust the number of video layers to be sent and the trans-mission path of each selected video layer in the varying network condition. And also, we compare the MDP-based transmission policy with other two transmission policies and it shows that the MDP-based transmission policy can achieve the higher long-term reward.