Research On Resource Provisioning And Task Scheduling Technologies In Multimedia Cloud

With the rapid development of multimedia cloud technology,it has penetration into all aspects of people’s work and life,becomes an important research direction in the field of cloud computing,and has gained the attention of industry and academia.As a specific cloud,multimedia cloud has an important influence on the development of the multimedia industry,such as film,animation,advertising and other related fields.Due to huge data volume,high concurrency,strict real-time and highly interactive features,resource provisioning and task scheduling for multimedia cloud have become one of the most critical challenges.Efficient resource provisioning and task scheduling mechanism provides significant advantages to ensure the QoE(Quality of Experience),reasonably allocate resources to multimedia applications,and satisfy growing user demands for multimedia services.Resource provisioning and task scheduling technologies for multimedia cloud have been widely studied in recent years.However,there exist still some critical problems to be solved.First,the existing resource scheduling models and algorithms for multimedia cloud mostly focus on the QoS(Quality of Service),rather than on ensuring the QoE.Second,multimedia cloud system should respond to dynamic user requests in time and provide fair services.Third,current states of the arts neglect reliability issues,and most of researchers investigate the cost-optimal scheduling problems.Finally,because of the openness,complexity and dynamiclty of the multimedia cloud,users face the risk of security and privacy.The security method in the existing literature can not be directly applied to the multimedia cloud environment.In this dissertation,aiming at enhancing the QoE of users and improving the performance of the multimedia cloud system,we systematically study resource provisioning and task scheduling for multimedia cloud in terms of energy,load balancing,reliability and trust mechanism.The main contributions and innovations of the dissertation are as follows:1)A resource provision algorithm based on QoE and energy balancing for multimedia cloud is presented.To host large-scale multimedia applications,huge numbers of computing resources are used in multimedia cloud system,and consume large amounts of energy.For multimedia applications,once achieving the desired QoE,it is important to optimize resource utilization,reduce system energy consumption,and achieve trade-off between QoE and system energy.For video streaming applications in multimedia cloud,by analyzing objective factors,a quantitative model of the QoE for video streaming is presented.Employing Lyapunov optimization techniques,an approximate online algorithm called RPA-QEB is proposed to flexibly adjust trade-off between energy and QoE.The RPA-QEB algorithm can guarantee desired QoE and reduce energy consumption,even without any information about the future fluctuation of user demands.2)A load balancing strategy based on cooperation game for video streaming task in multimedia cloud is proposed.Load balancing has become the core issue of multimedia cloud resource management and task scheduling.Performance of load balancing strategy mainly determines the overall performance and resource utilization of a multimedia cloud system.A queuing model is introduced to characterize multimedia cloud data centers.Considering fairness and Pareto optimization features of the cooperative game,a load balancing strategy based on Nash bargaining solution called CGS is presented.With extensive simulations it is shown that the CGS algorithm can achieve load balancing among cloud data centers while at the same time guarantee the fairness of the tasks submitted by users.3)A reliability-aware scheduling algorithm for video analysis task in multimedia cloud is designed.Resource failures in multimedia cloud have become norms instead of exceptions.Failures impact on system performance,reliability and availability.If the reliability of multimedia applications can not be guaranteed,it will negatively impact on the QoE.In order to satisfy the reliability and real-time requirements for video analysis tasks,a novel resource reliability evaluation model combing with execution reliability of video analysis task is introduced.Based on the model,a reliability scheduling algorithm for video analysis task in multimedia cloud with reliability and deadline constraints called RAS_VA is designed.The simulation experiment results show that the proposed algorithm outperforms other kindred traditional algorithms,with respect to guarantee reliability and real-time requirements of video analysis task in multimedia cloud environment.4)A trust and cost-aware scheduling scheme for video transcoding task in multimedia cloud is proposed.Due to the high concurrency,compute-intensive,data volume features of video transcoding task,it is necessary to process the video transcoding tasks across multimedia cloud providers.However,it is difficult for users to identify the real performance of video transcoding services.Malicious and fraudulent behaviors of multimedia cloud service providers can extremely decrease the trust degree of multimedia services.Considering behavioral characteristics of multimedia cloud entities,a trust model combining subjective and objective trust for multimedia cloud is proposed.Based on this model,for security-sensitive video transcoding task in multimedia cloud,a trust and cost-aware task scheduling algorithm called TBSA is proposed.Compared to the classical algorithms,experimental results show that the TBSA algorithm achieves some advantages in task completion time and task success ratio.