Research On Key Technologies For Data Access In Intelligent Terminals Of The Next Generation Broadcasting Network

With the promotion of the development of the next generation broadcasting net-work (abbreviated in NGB), and gradual development of the triple-play services, vari-ous interactive applications on open service platform will put the end-users of broadcast-ing network from passive information recipients into a mixture of information producers and consumers, and this transform raise special demands for data storage and retrieve.Based on the users’strong demand for a convenient and reliable way to share data with each other, and the development status that most intelligent terminals of the broad-casting network struggles with limited storage and computation resources, it is a desired low cost and effective solution to adopt the centralized storage pool with highly scalabil-ity and characteristics of on-demand, which follows the trends of information industry. Under this background of demand, it is important to establish a storage service which takes the characteristic advantages of NGB.On the other hand, high-definition video services with assured quality, which is built-up during the digitalization transform of the broadcasting network, are character-istic services which differentiate the broadcasting network from the Internet and the broadband telecommunication network. Distinguished from other type of data, high-definition videos are mainly enjoyed via live broadcast, requesting or recorded type. When recorder applied, the huge amount of recorded high-definition video files is stored in the local storage of high-definition media player. Because of their reasonable price, hard disks are adopted as the most common storage medium for home intelligent ter-minals. Hard disk drives have been running into a flatter performance promotion stage these years, on the contrary, the geometrical growth of the definition and bit rate of video requires faster read/write speed of hard disk. It is obviously important to exploit the total performance of disk through file system built upon and drop down the proces-sor utilization when parallel read and write requests are directed to a single disk, thus ensure well user experience.The mainly contributions and innovations of this dissertation are shown as follows.1. Based on the higher and allocatable download bandwidth in broadcasting net-work, an enhanced distributed data deduplication protocol-delta reference is proposed. In this protocol, the storage server searches for similar file for those digests of chunks received from the client, and replies with similar chunks, delta compression is applied at the client side to reduce the total volume of data trans-mitted to the server. By shifting part of volume to the higher bandwidth stream with a slightly increased total transferred volume, delta reference significantly decreases transfer time and results in faster data synchronization.2. An effective directory index framework-HB-Index, which takes the advantages of hash table and B+-tree in combination, is proposed. In this framework, inode corresponding to a given file name can be found via a hash table in O(1) time complexity, and only inode numbers of subfiles are kept in the parent directory file using a compact B+-tree, thus creation and deletion of file can both be completed in O(logd N) time complexity. Benchmarks in real environment reveal that HB-Index can easily handle directories with large amount of subfiles, and it results in a well overall performance when huge numbers of files reside in a single file system.3. A new disk file system-Boaz is designed for fast sequential read/write of high-definition video. In this file system, metadata and data are placed contiguously in separate regions, appropriate size of cluster is adopt for the assurance of min-imum sequential read/write bandwidth and maximum disk utilization ratio. The relatively larger cluster size reduces the metadata update frequency and consump-tion of computational resources as well. Moreover, an expected file size based pre-allocation strategy is applied to make metadata of files more compact and de-lay the aging process of file system. Benchmarks in high-definition player reveal that Boaz results in faster speed and lower CPU utilization than other counterparts in sequential read/write and parallel access tests.These improved methods proposed here against key issues about storage of termi-nal data in the next generation broadcasting network may provide inspiration for related subsequent research. Further consummations could provide some practical solutions for the demonstration work of NGB.