Resource Allocation In Edge-Clouds Aided HetNets For Video Transcoding Services

With the rapid development of wireless networks and spread of handheld devices(smartphones,tablets,etc.),users have an explosive growth demand for streaming service.Due to the diversity of user equipment,the source streaming has to be transcoded as different versions.However,transcoding is a computationally expensive and time consuming process.Since the unstable of wireless network and the shortage of computational resources,providing strict time-delays requirements transcoding services for wireless user equipments(UEs)is a big challenge.In this thesis,we focus on transcoding problems in video on demand(VoD)and live streaming scenarios.The main works are following:1)For VoD scenario,we propose a mobile edge computing(MEC)based heterogeneous networks(Het Nets)framework to offload these video data from UEs to MEC servers,which is a promising technique to relief the computing burden of UEs.Different from the existing literatures,our research focuses on maximizing the amount of the offloaded videos while decreasing time-delays by jointly considering offloading decision and computational resources allocation.Besides,a queue is created at the MEC server side for each UE to store the unexecuted videos in a time slot,which is used as a punishment factor in our utility function to avoid serious delay.An Actor-Critic reinforcement learning based resource allocation algorithm is proposed to resolve the problem.Simulation results indicate that the proposed algorithm can find the best strategy to maximize the amount of videos transcoded by the MEC server while decreasing time-delays.2)For live-streaming services,a software-defined network(SDN)-based edge cloud-assisted Het Nets architecture is introduced to realize the transcoding and transmission of live-streaming,which can not only significantly reduce the communication burden of the core network,but also greatly reduce latency.We jointly consider user scheduling,transcoding policy selection,computing resources and wireless spectrum resource allocation problem to maximize the quality of live-streaming while guaranteeing time-delays requirements.Different from existing literature,to approach the real wireless environment,the available computational and wireless spectrum resources are modeled as random processes in our research.Considering dynamic characteristics of wireless networks and the available resources,the above problem is modeled as a Markov decision process(MDP).Since the action space of the MDP is multidimensional continuous variables mixed with discrete variables,the traditional learning algorithms cannot find the optimal policy.Therefore,an enhanced Actor-Critic algorithm is proposed to resolve the problem,in which both the Actor part and the Critic part are employed eligibility traces to speed up the learning process.Simulation results show the proposed algorithm has superior performances compared with the policy gradient algorithm and deep Q-learning network(DQN).