传统的存储系统架构在处理日益增长的海量视频数据时，存储和管理数据低效，难以发挥主机端计算性能和存储介质特性。本项目针对该问题，提出一种基于开放通道固态硬盘和主机感知叠瓦式(SMR)磁盘的大容量、低成本、高性能混合存储系统，围绕存储介质特性和视频存储场景，拟在主机端展开以下研究：1.研究适用于非结构化视频数据的键值存储数据组织形式和符合应用特点的冷热数据分离策略，构建弹性的数据管理框架；2.研究混合存储系统中协作式垃圾回收策略，降低垃圾回收期间外部应用请求延迟和垃圾回收代价；3.研究抑制SMR写放大现象的粗粒度缓存替换算法，优化读写命中率，降低替换开销；4.研究适配混合存储系统的并行I/O调度方案，加速应用请求响应，提高服务质量。本研究有利于实现经济、高效的海量视频数据存储，丰富大数据存储管理的解决方案，促进存储设备主控算法的国产化发展，提高我国视频存储领域的自主化技术水平。

Traditional storage system architectures are low efficient in storing and managing the growing massive video data, which is difficult to fully utilize the computing performance of host-side and the media characteristics of storage system. Aiming at this problem, our project proposes a high-capacity, low-cost, and high-performance hybrid storage system based on open-channel solid-state disks and host-aware shingled magnetic recording (SMR) disks. Based on the storage media characteristics and the video storage scenarios, our project is going to carry out the following research: 1. Studying the key-value store-based data organization which is applicable to unstructured video data and cold / hot data-oriented distribution strategy, and building a flexible data management framework; 2. Studying the collaborative garbage collection strategy in the hybrid storage system to reduce the request delays from external applications and lower the cost of garbage collection; 3. Studying the coarse-grained cache replacement algorithm that mitigates the SMR write amplification to optimize read/write cache hit ratio and to reduce replacement overhead; 4. Studying the efficient parallel I/O scheduling scheme that is appropriately adapted to the hybrid storage system to accelerate the responses of application requests and improve the quality of service. The research of our project will be conducive to realizing the economical and efficient storage of massive video data, enriching the big data solution, promoting the localization development of control algorithm in storage devices, and raising the autonomous technical level in domestic video data storage field.