Research Of The Important Technology Of Flash File System Based On Large Capacity NAND Flash Memory

Flash memory is commonly used in embedded system which storage media with the advantages of small size, high capacity, low cost. It is first to be known spread through the U disk. Now, flash memory is widely used in our daily lives with the growing popularity of mobile phones, digital cameras, PDA, portable game consoles and other consumer electronic products. The flash memory is divided into two categories, NOR-type and NAND-type. NOR-type flash memory has a separate data bus and address bus, reading speed is quite fast, and it is suitable for storing program code, but in the NAND-type flash memory, data and address using the same bus, it read slowly, but the erase speed is quite fast, and it is suitable for storage of the large files and data. With the increased capacity of flash memory, flash file system requirements are getting higher and higher, its storage management affects the performance and efficiency directly. Currently, the file system which is available for NAND-type flash memory has two types. The one is centralized index and the other is designed for NAND flash file system. There are lots of flash-based file systems to be used in flash memory now, such as JFFS/JFFS2, YAFFS, LFM, UBIFS and so on. Flash file system includes following aspects of the key technologies: (1) data storage structure. (2) The file management system. (3) The file system load. (4) The garbage collection and wear leveling approach.This paper is based on the work of some of the important technology of the flash file system. Firstly, it take deeply research and analysis with the development of new flash file system UBIFS, and introduce the hierarchy structure of UBIFS in detail, and introduce the document management and garbage collection logs and other important technologies. They analyze and compare difference in performance between UBIFS and existing JFFS2 flash file system from the internal structure and data storage. And then it point out the superiority of the UBIFS performance. However, by analyzing the several important technologies of UBIFS file system, we also found out two major limitations in the wear leveling management of the UBI subsystem. Firstly, they are not well adapted the problem which is the erase block Data frequently updated parts caused by the block that erased too many times. Secondly, it can not be effectively achieved to move the data of the hot and cold between the erase block. In response to these problems, we propose a new algorithm which is based on the temperature of physical erase block and the age of the logic erase block, that is called LTPA (leb temperature peb age) algorithm. LTPA algorithm maps the hot logical block to the young physical block, and the temperature of the cold logical block is mapped to an old physical block. The temperature of the logic block is high or low depends on the written frequency of the data, and the physical block size is determined by the erase counts of physical block. We take the partial methods to predict the operating time of the temperature of the logical block in the logical erase block layer. At the same time it use a circular queue approach to manage the free physical block allocation strategy, in the circular queue, it use the binary search method, and it achieve the conversion of hot and cold data, it take better effect results for the wear leveling. The simulation algorithm is taken between UBI and LTPA algorithm which is compared the differences of the wear leveling performance. It proves the superiority of the LTPA algorithm in wear leveling application.Finally, UBIFS has a poor garbage collection strategy, we point out a new design method for the garbage collection, this method consider the dirty blocks of the program which focuses on the effective ratio of data and dirty data. It manages the garbage collection that is based on the circular queue. The policy will make management more efficient when it is applied to the UBIFS flash file system.