Research On The Resource Scheduling Strategy Based On Joint Optimization For The Streaming Media Edge Cloud

There are more or less limitations on scalability, reliability, service encapsulation capability for the traditional streaming media system based on Content Delivery Network (CDN), Peer-to-Peer (P2P) or cluster techniques. In cloud computing, there are greater improvements on resource virtualization, elastic scalability, and service reliability. Then, as a natural product of combining streaming media and cloud computing, the streaming media cloud can better solve the above problems. What’s more, considering the higher requirement for the real-time performance, the streaming media edge clouds are usually deployed at different local regions. Therefore, the requested video contents are pushed to the Internet edge, the user requests can be timely responded and the traffic on the core network can be reduced.Streaming media edge cloud is a kind of resource-intensive service system, and the most important are the bandwidth and storage resources. Similar to the traditional system, resource scheduling is still an important problem for the streaming media edge cloud. Specially, when there are only finite cloud resources, due to the skew and dynamic fluctuation of request pattern, the allocation for the bandwidth resources and storage resources could be unreasonable. Traditionally, session migration and video deployment are proposed to reallocate the bandwidth resources and storage resources, respectively. However, traditional researches about two methods are independent of each other. Because factual request pattern usually changes gradually, simplex session migration or video deployment cannot get a good trade-off between scheduling effect and cost. Meanwhile, deleting some replicas usually makes some sessions disconnected during video redeployment. Therefore, aiming at the above problems on the joint optimization of resource scheduling, the main content and achievements can be described as follows:1) By using the advanced cloud computing and OpenFlow techniques, a novel streaming media edge cloud architecture is proposed. Providing the streaming service in the application layer and optimizing the forwarding paths in the network layer would be separated with each other, and the service would also be more transparent.2) Using some new features in the OpenFlow techniques, a network-layer session migration method is proposed, that can separate the application-layer and the network-layer status information synchronization from each other.3) Considering the dynamic fluctuation of video popularity, a three-phase integrated scheduling strategy is proposed. In detail, on the basis of the original video popularity, a static deployment strategy is used for initializing the resource allocation. When the popularity among different videos has some minor variations, a session migration strategy would be adopted, and both load distribution and video popularity are used. Moreover, a strategy named "lazy migration" and a multi-step migration algorithm based on branch-and-bound, are adopted to take charge of the migration executing. When the video popularity incurs a great change, a video redeployment strategy is adopted. Besides the acceptance ratio, the adaptability for the dynamic fluctuation and the deployment cost are optimized together. Moreover, a session migration strategy based on load balancing is also proposed. On the other hand, a gradual deployment strategy is adopted to further control the deployment cost.4) Finally, aiming at the resource scheduling problem in the streaming media cloud, a numerical simulation platform is bought forward. Different scheduling strategies could be used in our simulation platform, which can improve the fairness among different scheduling strategies.