引入蒙特卡罗模型的闪存转化层管理算法研究

Solid state storage is important storage medium, which has so many advantages such as non-volatile, high-speed read/write performance, low power and small form-factor. The Flash Translation Layer algorithm is the key to affect the performance of solid state hard drive. With the development of Flash memory technology, Flash Translation Layer management algorithms has some challenges. Under the limited buffer space, read and write performance of solid state drive is poor; the block is frequently exchanged when data of read and write is frequent; system overhead affects the efficiency of garbage collection. According to feature attribute of Flash memory chip, this project will in-depth research Flash Translation Layer management algorithm on mapping, wear-leveling and garbage recycling process, and will propose the multi-objective optimization function and numerical solution method of address mapping efficiency description, establish a new type of wear-leveling algorithm based on data block activity model, design dynamic garbage collection strategy based on the I/O processing ability, to improve the performance of Flash Translation Layer management algorithm, to ultimately improve read and write performance of solid state storage drive , and to extend the service life. Monte Carlo method is introduced to model the block activity parameters of short time data and calculate the probability of garbage block in this project. It is useful to enhance the core competitiveness of China in the field of solid state storage, and it is theoretical foundation for localization of solid state drive.

固态存储，具有非挥发性、快速读/写、低功耗及易携带等优点，已经成为一种重要的存储媒质。在固态存储核心技术中，闪存转化层管理算法是影响固态盘性能高低的关键所在。随着闪存芯片技术的发展，现有的闪存转化层管理算法存在：有限缓存空间下，固态盘读写性能差；损耗均衡时读写频繁的数据被频繁交换；系统开销影响垃圾回收效率等问题。本课题根据闪存颗粒的特征属性，通过深入研究闪存转化层管理算法的映射方式，均衡和垃圾回收过程，引入蒙特卡罗法进行短时数据块活跃度参数建模和垃圾块产生概率的计算，拟采用描述地址映射效率的多目标优化函数及数值求解方法，建立基于数据块活跃度模型的新型磨损均衡算法，设计依据I/O处理能力的动态垃圾回收策略，完善闪存转化层管理算法，提高固态存储盘的读写性能，延长使用寿命，提升我国在固态存储领域的核心竞争力，为固态硬盘国产化做好基础理论研究。