网络视频存储系统的设计与实现

发布时间：2018-06-03 16:09

本文选题：网络视频存储 + 任务线程池　；参考：《华中科技大学》2013年硕士论文

【摘要】：从技术层面看，视频监控技术经历了第一代模拟监控、第二代数字监控，发展到了目前的第三代网络监控阶段。网络视频存储系统，具有访问客户端多，图像清晰度要求高，数据流稳定，存储时间长等特点，而且要求回放检索的时间颗粒度小。因此，如何使大量视频数据及时无误的录入和快速准确的检索是网络视频存储系统要解决的关键问题。为解决上述问题，提出了一种基于“时间索引+数据块”键值（key-value）存储模型，采用类ext2文件系统方式组织管理数据，每个逻辑卷对应一个块组，在逻辑卷上存放时间索引和数据块。为了保证负载均衡和系统超载时仍能正常工作，采用任务线程池的方式管理工作线程，同时保证存入的优先级高于检索任务的优先级。利用视频流顺序存放和读取的特点，每一路数据流在内存中开辟一个双循环无锁缓冲区，，来提高存储的效率。通过对网络视频存储系统的测试，结果表明系统时间索引与数据块能保持一致。性能分析表明，常规配置的存储服务器能够实现秒级快速检索和支持40路以上高清视频流并行录像。
[Abstract]:From a technical point of view, video surveillance technology has experienced the first generation of analog monitoring, the second generation of digital monitoring, to the current third generation of network monitoring stage. The network video storage system is characterized by many clients, high definition, stable data flow and long storage time, and the time granularity of playback and retrieval is small. Therefore, how to make a large number of video data timely input and fast and accurate retrieval is the key problem to be solved in the network video storage system. In order to solve the above problems, a key-value-based storage model based on "time index data block" is proposed. The data is organized and managed by ext2 like file system, and each logical volume corresponds to a block group. Store time indexes and data blocks on logical volumes. In order to ensure that the workload balance and system overload can still work properly, the task thread pool is used to manage the worker thread, and the priority stored is higher than the priority of the retrieval task at the same time. Based on the characteristics of sequential storage and reading of video streams, each data stream creates a double-loop unlocked buffer in memory to improve storage efficiency. By testing the network video storage system, the results show that the time index of the system is consistent with the data block. The performance analysis shows that the normally configured storage server can quickly retrieve and support more than 40 parallel video streams in high definition video stream.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333

【参考文献】