Research On The Model And Algorithm For The Load Capacity Of Video-on-Demand

With the development of online Video-On-Demand (VOD) service, in a large number ofuser environment and high concurrent flow, to ensure that demand quality and improve thesystem load capacity is the inevitable trend of development of multimedia. From theperspective of system analysis, load problems usually include system architecture, thecontents of the video codec, transport protocols, access control, storage scheduling. And inrecent years, the rapid development of cloud computing technology provides us with newresearch methods. Therefore, combined with cloud computing technologies to design a newVOD model and improve its load capacity. Become a challenging, but also has greatsignificance.In this paper, we combined with the actual experience on the construction of VODproject. We analyzed the different typical VOD cluster architecture, access control and storagescheduling design ideas, implementation, deployment method and its performance advantagesand disadvantages. Around the issue of large-scale VOD load capacity of the Internet. Use ofvirtualization and cloud storage technology, want to optimize and improve the load capacityof the large-scale VOD system. Focus on system architecture, resource scheduling, accesscontrol, cloud storage and other aspects. These innovations are briefly summarized as follows:(1) We designed a full virtualization-based VOD services cluster model. According to themonitoring data of the virtual node, the cluster system could dynamic schedule the physicalresources. Compared to traditional physical deployment mode, greatly improving theutilization of hardware resources and the expansion of the cluster load-supporting capacity.(2) We designed a dynamic cluster scheduling management strategy based on resourceprojections, and in accordance with the actual demand on hardware resources (determined bythe priority of the application and the actual needs of the user-specific business) dynamicallyadjusts allocation ratio of resources between virtual machines. So that we can build a scalablevirtual server cluster to achieve the pre-allocation and on-line flow of resources, enhance theefficiency of resource use, and load balance virtual machine. Compared to the EDF, Creditalgorithm, the algorithm can be achieve maximize hardware resource utilization, improvevideo on demand service load capacity.(3) We introduced ant colony clustering to the video popularity analysis algorithms, trying to make early irregular clustering, according to the clustering results to the formation ofa hot video collection. Then give priority to control the access of these collections. Thismethod not only to avoid the uncertainty of α value of the zipf algorithm, and also more realguarantee of coverage of the target video hot content.(4) Designed an access control algorithm based on video popularity. Compared totraditional robin algorithm, greedy algorithm and the other access control algorithm, thealgorithm can be possible to meet the hot resources on-demand request, but also indirectlyincrease the service load capacity of the system.(5) We proposed an improved lightweight cloud storage engine design which suitable forstorage video multimedia resources, which called "LiDFS"(Lightweight Data File System).The program is based on mainstream of open-source cloud storage system design idea, theLCA (Relaxation coupled asymmetric structure) architecture, focus on cloud storage enginearchitecture, communication protocols, improved the design of synchronization and filestorage location mechanism, compared to MogileFS, the MongoDB storage architecture has abetter rapid resource sharing capabilities and low costs.(6) Designed a corresponding “pop-priority” copy of video content schedulingalgorithm. The algorithm is based on the file popularity (file access times, request time period,frequency, etc.) to determine the file to make a mirror, and pre-migration operations, to ensurethat files are the fastest and most requested and reduce transmission distance (routing), inorder to effectively improve the response speed of the video file in order to ensure that asmuch as possible of the document request was accepted, and thus better achieve thescheduling of video services, and improve the load capacity of the entire video on demandservice system.In summary, experimental results show that the above method is preferably to improvethe load capacity of the VOD system. This work has been partially successful used in theconstruction of the Guangzhou University Digital Campus project, and achieved better results.