垃圾回收是影响闪存性能和可靠性的重要因素之一。随着重复数据删除技术和日志结构文件系统在闪存介质上的应用，本项目从闪存存储系统的整体角度出发，挖掘重复数据删除的技术特性，主要研究（1）内容感知的垃圾回收技术，将重复数据删除技术嵌入到闪存垃圾回收过程中，消除在线重复数据删除技术对高性能闪存的性能影响；（2）跨层协同垃圾回收技术，通过将日志结构文件系统、重复数据删除和闪存存储三个层次的垃圾回收操作协调一致，减轻不同层次的非协调垃圾回收造成的闪存写放大问题；（3）闪存阵列中基于数据引用率的条带化垃圾回收技术，将闪存阵列中各闪存设备的垃圾回收协调一致，并以条带作为垃圾回收的基本单位，解决垃圾回收造成的校验更新和性能波动性影响问题。本项目拟基于仿真和原型系统验证相结合的实验方法研究上述关键技术，提高闪存垃圾回收效率，进而提高闪存存储系统的性能和可靠性，进一步推动闪存存储的应用。

Garbage Collection (GC) is one of the key factors that impact the performance and reliability of flash storage. Along with the application of data deduplication and log-structured file systems on flash storage, this project exploits the specific characteristics of data deduplication to optimize GC from the overall perspective of flash-based storage systems, as follows: (1) Content-aware GC, which embeds the data deduplication within the GC process of the flash device to eliminate the performance impact of online data deduplication on high-performance flash storage. (2) Cross-layer Coordinated GC, which co-designs the GCs among log-structured file system, data deduplication, and flash device to alleviate the write amplification induced by the independent and uncoordinated GCs among the three layers. (3) Reference-count based and stripe-oriented GC for flash-based disk arrays, which coordinates the GC processes among different flash disks and makes a stripe as the minimum erase unit within flash-based disk arrays, to solve the GC-induced parity update and performance influence problems. By experiments based on both the simulation and prototype implementation, these key techniques could improve the performance and reliability of flash-based storage systems by improving the GC efficiency, thus further strengthening the applications of flash storage.