大容量固态硬盘部署于外存系统已成为当前趋势。然而，其物理页面数量的急剧增加将产生大量映射表项，而大量表项将造成更频繁的高速缓存缺失，进而影响访存性能；此外，大容量固态硬盘的物理页面在增大，这将产生因I/O大小与页面大小不一致引起的性能降级现象。因此，大容量固态硬盘在映射表设计、高速缓存管理和大小页混合管理中面临新的挑战。本项目针对大容量固态硬盘，开展映射表优化设计和访存优化研究。1)首先，建模分析不同映射表机制对访存性能的影响，提出新型的基于分组的地址映射表方法和相应优化设计，以减小映射表项在闪存上所需的存储空间；2)其次，研究映射表表项在高速缓存中存储的缩减方法，设计轻量级冷热缓存管理算法并研究时空感知的随机访问优化算法，通过提高映射表表项在高速缓存中的管理效率来优化访存性能；3)最后，研究大小页混合管理的映射表优化方法，并基于工作负载特点设计大小页分区动态管理技术，进一步优化访存性能。

It is a trend to deploy large-capacity SSDs in storage systems. However, more and more pages in large-capacity SSDs will cause more and more mapping entries, which leads to performance degradation because of more cache misses of mapping entries. Besides, page size in large-capacity SSDs becomes bigger and bigger, which also causes performance degradation due to the size discrepancy between I/O requests and flash page size. Accordingly, we face new challenges in FTL design which are how to design a suitable mapping table, how to efficiently manage cache for mapping entries and how to achieve high performance for hybrid management between pages of big size and small size. Motivated by this, we focus on the mapping table design and performance optimization for large-capacity of SSDs in this project. 1) We first develop a theoretical framework to analyze the influence of different mapping table schemes on the performance of SSDs. In order to reduce the size of each mapping entry, we come up with a novel set-based mapping table approach and do research work for its optimization. 2) We design a method to reduce the size of mapping entries stored in cache, develop a light-weight LRU algorithm and propose a temporal and spatial locality-aware algorithm for workloads with random access, so as to achieve efficient cache management for mapping entries. 3) At last, we optimize the mapping table design and develop a dynamic hybrid management technology for different size of pages, so as to further improve the performance of large-capacity of SSDs.