KR20110021417A - Flash memory system and contents-based block management method thereof - Google Patents

Abstract

PURPOSE: A flash memory system and a contents-based block management method thereof are provided to manage blocks by using a contents-based list in order to reduce a mount time and efficiently manage a wear ratio. CONSTITUTION: A RAM(Read Random Memory)(120) loads a contents-based file system operated by a processing unit(110), and a flash memory(140) includes the blocks managed by the contents-based file system. If a file writing operation is requested, the contents-based file system selects one of blocks. The contents-based file system determines the kinds of the selected blocks as one of a pure block, a mixed block and a free block. The contents-based file system stores block type information and writing requested file at the block.

Description

Flash memory system and its content-based block management method {FLASH MEMORY SYSTEM AND CONTENTS-BASED BLOCK MANAGEMENT METHOD THEREOF}

The present invention relates to a file system, and more particularly to a file system for managing a flash memory.

The present invention is derived from a study conducted as part of the IT SoC core design manpower training department of the Ministry of Knowledge Economy [Task management number: C1200-0901-0002, Task name: IT SoC core design manpower training project].

The use of flash memory as a storage medium for mobile devices such as PDAs, PMPs, RFID readers, and the like is gradually increasing. A storage medium using a flash memory (hereinafter referred to as a flash storage medium) has a nonvolatile characteristic and is stronger than a hard disk. Flash storage media can operate at low power, and the access time of flash storage media is faster than that of hard disks. Small in size, flash storage media is suitable for portable devices. However, flash storage media are five to ten times more expensive than hard disks. When you want to write new data in an already existing space, the process of erasing the already existing space is performed first, and then the data is stored in the deleted space. For flash storage media, the read speed is very fast, but the write and erase speeds are relatively slow. In addition, the size that can be erased at a time is constant, and the number of times that can be erased at room temperature is determined.

As is well known, a flash memory, which is a nonvolatile memory, is largely divided into a NOR type flash memory and a NAND type flash memory. NOR type flash memory has the advantage of fast read speed, and NAND type flash memory is cheaper than NOR type flash memory and is easy to use in large-capacity embedded devices. Table 1 below shows execution time of basic flash memory operations.

Media Read Write Erase DRAM 60㎱ (2B)
2.56 ㎲ (512 B) 60㎱ (2B)
2.56 ㎲ (512 B) N / A NOR flash 150㎱ (1B)
14.4 ㎲ (512 B) 211㎱ (2B)
3.53 ㎲ (512 B) 1.2 s (128 KB) NAND flash 10.2 ㎲ (1B)
35.9 ㎲ (512 B) 201㎲ (2B)
226㎲ (512B) 2 ms (16 KB) Disk 12.4 ㎳ (512 B)
(average) 12.4 ㎳ (512 B)
(average)

1 illustrates a general flash memory structure. As shown in FIG. 1, a flash memory is composed of blocks (or called memory blocks), and each block is composed of pages. In the case of small blocks, one block consists of 32 pages, each of which is 512-bytes in size. On the other hand, in the case of a large block, one block is composed of 64 pages, and each page is 2048-byte in size. Flash memory requires garbage collection to clean up invalidated pages of blocks and allocate those block (s) to free space when free space is lacking.

Due to the characteristics of flash memory mentioned above, existing file systems cannot be directly applied to flash memory. Therefore, there is a FTL (Flash Translation Layer) that replaces the existing file system. FTL is implemented in the form of a driver that performs mapping management to make the sequential flash space look like sectors on the disk. However, since the FTL exists independently of the file system, the overhead due to the independent operation has to be taken into consideration. Therefore, researches on a flash memory dedicated file system including the FTL function in the file system have been actively conducted. Representative NAND flash-only file systems (or referred to as flash file systems) include the Journaling Flash File System (JFFS2), the Yet Another Flash File System (YAFFS), and the like. JFFS2 transforms the update operation on the flash memory into an additional operation like the Log-structured File System (LFS), which sequentially stores the flash space, and solves the problem of overwriting the flash memory.

In the case of JFFS, data is sequentially written to flash memory in log form, and a read operation is performed by searching the log in reverse order and reading the latest data. JFFS uses a structure called jffs_node as a journaling node, and the structure is quite large, 48 bytes. To reduce this overhead, JFFS2 replaces jffs_node with a structure (16Bytes) called jffs2_raw_node_ref that consists of next_in_ino, next_phys, flash_offset, and totlen values, and stores it in memory. JFFS2 also uses data compression to efficiently utilize the space in flash memory. In this way the memory usage can be reduced, for example, from 48 bytes to 16 bytes. However, when the flash memory is 128MBytes, there is a problem of allocating 222Bytes, or 4MBytes, as the space for jffs2_raw_node_ref because it may require 218 pages, and the scan time for finding the node and determining the structure of the file is long. .

In addition, 128 MBytes NAND flash memory takes about 25 seconds to mount, since the entire memory must be scanned when mounted. Since JFFS2 is designed based on NOR flash memory, there are various problems in memory usage, mount and flash memory scan time, and garbage collection time to be used as a file system for NAND flash memory.

As another file system, YAFFS is a NAND flash-only file system developed to address the drawbacks of the large memory consumption and slow mounting speed used for journaling JFFS2. YAFFS is used for flash memory with a page size of 512 bytes, with a maximum file size of 512 MB, maximum file count of 260,000, and maximum file system size of 1 GB. YAFFS2 can use flash memory with a page size of 2Kbyte, and can have a file system size of up to 8GB. In flash memory, reads and writes are performed in units of pages, and deletions are performed in units of blocks.

2 illustrates a conventional YAFFS flash memory structure.

The file data is divided into chunks that are the same size as one page, and the chunks constituting the file are stored in pages of flash memory, respectively. The chunk is divided into a data area and a spare area, and a header for managing information of a file is stored in the data area. When the header is stored, the chunk ID ChunkID of the spare area has a value of zero. The header stores the file name, file size, modification time, and parent directory pointer. If the value of ChunkID is not zero, the information stored in the data area is considered to be the data of the file. Each chunk has a spare area. The spare area of each chunk includes a block status area, a page status area, an ECC area, a tag area, and the like, as shown in FIG. The page status area contains information indicating a valid / invalid page. The tag area contains data such as the size of data contained in the page, the number of file updates, the inode identifier, and the like. In the case of file update and file deletion, pages containing updated / deleted files are invalidated. Later, during garbage collection, block (s) containing invalidated pages are deleted.

Existing file systems (eg YAFFS) scan all spare areas of flash memory at mount time. When a page with a header is found, the existing YAFFS reads the data contents of the found page and the file state stored in the spare area, and dynamically creates an open file table of the file in main memory. As a result, the mount time of conventional YAFFS is much shorter than that of JFFS2, which must read the entire area of flash memory. However, since the spare area of at least all pages must be read during the mounting process, as the size of the flash memory increases, the mount time of the existing YAFFS increases.

Existing YAFFS basically uses Garbage-Collection according to Greedy policy. The use of the greedy policy results in a lifespan (wear) problem, but in the traditional YAFFS, about 5% of the total capacity is allocated as reserved areas. Complementary methods are used to replace blocks that cannot be used because their wear levels have reached their limits. The flash memory must perform an erase operation on the used block to reuse the used block. The erase speed is considerably slower than the read / write speed, and the delete operation is performed in blocks. In addition, wear limits exist, which can reduce flash memory life without taking wear into account when deleting. Representative erase policies will include Greedy, Cost-Benefit, and CAT techniques.

The greedy scheme selects the block with the smallest valid page in the segment to be deleted. This technique has the advantage of lowering the cost of moving a valid page to another block and eliminating the need for additional computation. However, in the case of local writes in which a particular segment changes frequently, both temporally and spatially, the Greedy technique is expensive, and wear-leveling, which makes the number of erases of each segment even, is not considered at all. .

The cost-benefit technique is an erasure policy that selects and deletes a segment that profits more than the cost (the cost of deleting a page + the cost of moving a valid page to another). The cost is the time required to delete and move, and the profit is the additional capacity after the erase process.

In Equation 1, u is the ratio of valid pages in the block and (1-u) is the ratio of invalidity pages. 2u is the ratio of valid pages to read and write that occur when you move a valid page from one block to another. Age is the time since the page was written to the block. In the Cost-Benefit technique, Age is used to equalize wear, but this is not the cumulative number of erased blocks but the time recorded in the block.

The CAT method is a method that calculates a value for each block and makes the block with the smallest value as the target of the policy, as shown in Equation 2 below. to be.

In Equation 2, u and (1-u) are ratios of valid pages to invalid pages. Age refers to the elapsed time since the block is written, and Age is determined by a predefined Age transform function. Finally, Number Of Cleaning represents the cumulative number of times blocks have been erased. However, when the user write pattern is sequential write or local write, the CAT technique shows better performance because the erase and copy times of the CAT technique are less than those of the Greedy and Cost-Benefit techniques. However, in case of random writes, the erase and copy times are higher and the average throughput is lower than the Greedy method, resulting in lower performance in terms of efficiency.

It is an object of the present invention to provide a file system and a flash memory system including the same which can reduce the mount time.

It is another object of the present invention to provide a file system and a flash memory system including the same which can increase the life of the flash memory.

Exemplary embodiments of the invention include a processing unit; A RAM into which a content-based file system operated by the processing unit is loaded; And a flash memory having blocks managed by the content-based file system. The content-based file system selects at least one block of the blocks when a file write is requested, and determines whether the selected at least one block belongs to one of a pure block, a mixed block, and a free block; The write request file is stored in the at least one block together with block type information indicating the determined block type.

In an exemplary embodiment, when the at least one block is filled with the write requested file, the content-based file system determines the at least one block as the pure block.

In an exemplary embodiment, when some of the at least one block is filled with the write requested file, the content-based file system determines the at least one block as the mixed block.

In an exemplary embodiment, determining whether the selected at least one block belongs to one of pure blocks, mixed blocks, and free blocks divides the size of the write requested file by the size of the block, and corresponds to the share. Are determined as pure blocks and the blocks corresponding to the rest as mixed blocks.

In an exemplary embodiment, the content-based file system reads block type information from a spare area belonging to the first page of each block when mounted, and spare areas of remaining pages of each block according to the read block type information. Determine the scan for.

In an exemplary embodiment, the content-based file system inserts the selected block into a mixed block list after scanning the spare areas of the remaining pages of the selected block when the read block type information indicates a mixed block, When the read block type information indicates one of a free block and a pure block, the open file table is inserted by inserting the selected block into either the free block list or the pure block list without scanning the spare areas of the remaining pages of the selected block. Configure.

Another exemplary embodiment of the present invention provides a block management method of a system including a nonvolatile memory composed of blocks, wherein the block management method includes block type information from a spare area belonging to a first page of a selected block among the blocks. Reading; When the read block type information indicates a mixed block, scanning the spare areas of the remaining pages of the selected block and inserting the selected block into the mixed block list; When the read block type information indicates one of a free block and a pure block, inserting the selected block into either a free block list or a pure block list without scanning the spare areas of the remaining pages of the selected block; ; And repeatedly performing the reading and inserting steps until all the blocks of the nonvolatile memory are selected.

In an exemplary embodiment, the block type information is generated when a file is stored in the nonvolatile memory.

In an exemplary embodiment, when the number of free blocks included in the free block list is lower than a threshold, performing content-based garbage collection based on the pure block list, the mixed block list, and the free block list. Is provided further.

In an exemplary embodiment, the content-based garbage collection selects mixed blocks having a small number of valid pages in the mixed block list, and lists valid pages included in the selected mixed blocks in the free blocks of the free block list. By moving to one of them and inserting the selected mixed blocks as a free block into the free block list.

In an exemplary embodiment, there is further provided a step of performing a block swap to level the wear of the blocks of the nonvolatile memory, wherein the block swap selects the pure block with the smallest wear in the pure block list, and It is achieved by selecting a free block having the largest wear degree in the free block list and swapping the selected pure block and the selected free block when the difference between the wear degree of the selected pure block and the wear degree of the selected free block exceeds a reference value.

In an exemplary embodiment, the block swap is performed at idle time.

In an exemplary embodiment, the block type information is stored in a reserved area belonging to a tag of the spare area.

Another exemplary embodiment of the present invention provides a block management method of a system including a nonvolatile memory composed of blocks, wherein the block management method includes at least one of the blocks of the nonvolatile memory in which a write requested file is to be stored. Determining a type of block; And storing the write requested file in the at least one block together with block type information indicating the determined type.

In an exemplary embodiment, the block type information represents any one of a pure block, a mixed block, and a free block.

In an exemplary embodiment, the at least one block is to the pure block and some of the at least one block to the write requested file when it is determined that the at least one block is filled with the write requested file. When determined to be filled, it is determined by the mixing block.

In an exemplary embodiment, the size of the write requested file is divided by the size of the block, and blocks corresponding to the share are determined as pure blocks and blocks corresponding to the remainder as mixed blocks.

In an exemplary embodiment, the block type information is stored in a tag of a spare area belonging to the first page of the at least one block, and the meta information of the write requested file is stored in the first page of the at least one block. It is stored as a header in the data area to which it belongs.

In an exemplary embodiment, the method may further include: reading, upon mounting, block type information from a spare area belonging to a first page of each block of the nonvolatile memory; The method may further include determining a scan of spare areas of the remaining pages of each block according to the read block type information.

In an exemplary embodiment, when the read block type information indicates a pure block or a free block, scanning of spare areas of the remaining pages of the block determined as the pure block or the free block is omitted; When the read block type information indicates a mixed block, spare areas of the remaining pages of the block determined as the mixed block are scanned.

Content-based file systems according to embodiments of the present invention will provide improved mount speed and wear leveling over existing file systems (eg, YAFFS / JFFS). The content-based file system according to an embodiment of the present invention manages blocks as a content-based list, thereby reducing the mount time by reading only a portion of a spare area when mounting and efficiently managing wear through a block swap technique using a content-based list. Have Therefore, the content-based file system according to the embodiment of the present invention provides a fast mount speed and increases the life of the flash memory through the wear leveling technique.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this specification, a content-based block management technique for improving mount speed will be described, and an erase policy that can improve wear management through efficient content-based garbage collection and content-based block swap will be described.

3 is a block diagram schematically illustrating a flash memory system according to an exemplary embodiment of the present invention.

Referring to FIG. 3, a flash memory system according to an exemplary embodiment of the present invention may be applied to mobile devices, computers, and the like such as a PDA, a PMP, an RFID reader, and the like. However, it will be understood that a flash memory system according to an exemplary embodiment of the present invention is not limited to what is disclosed herein. Flash memory system will mean a computing system including flash memory. A flash memory system according to an exemplary embodiment of the present invention may include a storage unit 140 composed of a processing unit 110, a RAM 120, a memory interface 130, and a flash memory that is a nonvolatile memory. The system of FIG. 3 is shown to include components necessary to describe a file system in accordance with an exemplary embodiment of the present invention. However, it will be understood that the components of the system are not limited to those disclosed herein. It will also be appreciated that the memory constituting the storage medium 140 is not limited to flash memory. For example, exemplary embodiments of the present invention may be applied to PRAM, MRAM, FRAM, NOR flash memory, and the like.

The flash memory constituting the storage medium 140 has the following features. First, it is impossible to overwrite the original address when modifying data. In other words, the memory space corresponding to the write operation must be initialized before the write operation. Second, there is a limited number of erases per block. Accordingly, the system according to an exemplary embodiment of the present invention will include a file system to which the erase policy and the even wear technique are applied to overcome the increase in mount time and the limit of the wear rate and to effectively use the flash memory. Such a file system will be loaded into RAM 120 and operated by processing unit 110.

In an exemplary embodiment, the file system according to an embodiment of the present invention will be YAFFS (hereinafter referred to as content based file system / content based flash file system). However, it will be understood that the file system according to the embodiment of the present invention is not limited to YAFFS.

The content-based file system according to an exemplary embodiment of the present invention determines the block type when a write requested file (required by the application layer) is written to the flash memory. For example, a content-based file system may determine a block as a pure block, mixed block, or free block. A pure block means a block filled with one content file, and a mixed block is a block filled with two or more file contents, or pages filled with one content file and free pages (or files filled with other content). Page). A free block means a block composed of free pages. The block type is determined according to the criteria described above, and information representing the block type (hereinafter referred to as block type information) is stored in a spare area belonging to the first page of the block.

When mounted, the content-based file system according to an exemplary embodiment of the present invention reads the block type information of the spare area belonging to the first page of each block rather than scanning all spare areas, and reads the block type information according to the read block type information. It will determine whether to scan the remaining spare areas. For example, when the read block type information indicates a pure block and when the read block type information indicates a free block, a scan of the remaining spare areas of the block will be omitted. In contrast, when the read block type information indicates a mixed block, the remaining spare areas of the block will be additionally scanned. Therefore, by scanning the spare area of the first page of the block, the information of the file in the block can be grasped and mounted.

A content-based file system according to an exemplary embodiment of the present invention is based on a pure block list, a mixed block list, and a free block list determined according to the block types described above (hereinafter referred to as content-based garbage collection). Will do This will be explained in detail later.

4 is a diagram for describing how a content-based file system determines a block type according to an exemplary embodiment of the present invention.

In the case of a file system such as YAFFS, a header for identifying a file is stored in each block, so that the number of files included in a block can be determined. That is, since one file has one header, the number of headers shows whether the contents of the block consist of one file or several files. The content-based file system according to an exemplary embodiment of the present invention divides blocks into three types, namely, pure blocks, mixed blocks, and free blocks, based on file information (number of files) identified through the number of headers.

The block type will be determined when the file is written to flash memory. For example, it is assumed that the total capacity of the flash memory is 512 Mbytes, one block is 128 Kbytes, and one page is 2 Kbytes. Under this assumption, when an MP3 file having a size of 4.4 Mbytes is used in the flash memory, 35 pure blocks and one mixed block are generated according to Equation 3 below. That is, the size of the file requested to be written may be divided by the size of the block, and blocks corresponding to the share may be determined as pure blocks and blocks corresponding to the remainder as mixed blocks. There will be at least one block corresponding to the quotient, and one block corresponding to the remainder.

In Equation 3, Pn represents a pure block number, Fsize represents a file size, and Bsize represents a block size.

The header is stored in the data area of the first page of pure and mixed blocks. The header stores information about the file (meta data). In addition, a block type field is added to a reserved area of a spare area belonging to a first page of a block for content-based block management. In this block type field, information indicating pure blocks, mixed blocks, or free blocks (that is, block type information) is recorded. Therefore, by scanning the spare area belonging to the first page of the block, the information of the file in the block can be grasped and mounted.

Fig. 5 is a flowchart for schematically illustrating a method of mounting a content-based file system according to an exemplary embodiment of the present invention.

Referring to FIG. 5, first, a content-based file system according to an exemplary embodiment of the present invention scans a spare area of a first page belonging to a spare area of a block (S100). As is well known, when scanning a spare area, the data area belonging to the first page will also be read. The content-based file system determines whether a block indicates a value stored in a block type field belonging to the scanned spare area (S110). If it is determined that the value stored in the block type field represents a pure block, the content-based file system inserts the block into the pure block list without scanning the spare areas of the remaining pages of the block (S120). If it is determined that the value stored in the block type field indicates a mixed block, the content-based file system scans the spare areas of the remaining pages of the block and inserts the block into the mixed block list (S130). If it is determined that the value stored in the block type field indicates a free block, the content-based file system inserts the block into the free block list (S140). The aforementioned steps S100 to S140 will be repeated until all the blocks of the flash memory are selected. Through the aforementioned process, a pure block list, a mixed block list, and a free block list can be constructed.

All you need to do at mount time is to make the data that is stored on the storage medium (eg flash memory) meaningful. Due to the nature of flash memory, certain areas cannot be classified as sections for meaningful data, so each time you mount, you must configure an open file table with headers stored in the data area of the first page of the entire block and file information stored in the spare area. do. Therefore, the mount time is considerable. However, in the case of the content-based file system according to the exemplary embodiment of the present invention, the mount time can be reduced compared to the existing file system by adding a block type field to the spare area belonging to the first page of the block at the time of mounting.

For example, a method of mounting an existing file system (eg, YAFFS) and a method of mounting a content-based file system according to an exemplary embodiment of the present invention are shown in FIGS. 6A and 6B, respectively. As shown in FIG. 6A, since the existing file system cannot grasp the block contents, the file (s) included in the block are checked by scanning the spare area of every page. In contrast, as shown in FIG. 6B, the content-based file system determines whether the corresponding block is a pure block or a mixed block through the block type field of the spare area belonging to the first page of the block during the mounting process. In the case of a pure block, since the data written in the block is the same file, the mount time can be reduced by {(BPN-1) × PBN} by scanning only the first page without having to scan the remaining pages in the block. Here, BPN represents the total number of pages in a block, and PBN represents the number of pure blocks. After the scan, the blocks are composed of their respective lists: pure block list, mixed block list, and free block list. Blocks are content-based.

When a file write or delete operation occurs, the pages containing the file are invalidated. Therefore, in the case of pure blocks, all pages of the pure block are invalidated when the file is deleted because the contents of the same file are included. On the other hand, in the case of mixed blocks, only the pages containing the file are invalidated. Content-based file systems will record files in consideration of wear-leveling in file recording. This will be explained in detail later.

7A to 7C are diagrams for describing a garbage collection method of a content-based file system according to an exemplary embodiment of the present invention.

Garbage collection is performed when free blocks (free space) are lacking, and the process of making a block of invalidated pages into a free block to store new data is made by deleting a block of invalidated pages. Garbage collection, in principle, performs a delete operation on blocks where all pages have been invalidated. However, when the number of free blocks falls below a certain threshold and there is no block in which all pages are invalidated anymore, a block containing some valid data is selected as the target block, and the valid data of the target block is selected as another block. Copying and deleting the invalidated block, that is, the target block, may be used.

According to an exemplary embodiment of the present invention, the garbage collection method may include securing free blocks when the available space is insufficient, and performing a leveling of the saddle through block swap.

First, a process of securing a free block when the available space is insufficient will be described in more detail. Assume that the number of blocks occupied by data is two when data is written. According to this assumption, two free blocks of the free blocks of the free block list are allocated for recording data. When two free blocks are allocated for data recording, as shown in Fig. 7A, the number of free blocks falls below a predetermined threshold (for example, five blocks). The threshold (or referred to as reference value) means the minimum number of free blocks remaining. In an exemplary embodiment, 10% of the total number of blocks (eg, 50) is set as the threshold. If the number of free blocks falls below the threshold (five blocks), there will be a shortage of free blocks. In this case, the content-based file system according to the exemplary embodiment of the present invention checks whether free data can be obtained by collecting various data scattered in the mixed blocks into one mixed block. If it is determined that it is possible to secure a free block, as shown in Fig. 7B, blocks having a small number of valid pages among the mixed blocks are selected, and valid pages of the selected blocks will be moved to the free block. For example, by collecting the valid pages of the mixed blocks 7, 8, 9 into the free block 3, two free blocks can be secured. That is, as shown in FIG. 7C, the free block 3 in which valid pages of the mixed blocks 7, 8, and 9 are stored is allocated to the mixed block, and the mixed blocks 7, 8, and 9 are free blocks. Is assigned to. Therefore, as six free blocks are secured, the block shortage state can be solved.

Existing erase policies select and delete blocks with a large number of invalidated pages when all free blocks are used up. Therefore, the existing erasure policies have the disadvantage of consuming long deletion time. However, the content-based erasure policy according to the exemplary embodiment of the present invention reduces the long deletion waiting time by raising the free block count above the threshold value by merging and deleting the mixed blocks when the free block count remains below a predetermined threshold. The deletion block selection method is used, and as a result, the problem (long deletion time) of the existing deletion block selection method is solved. In other words, according to the existing erasure policies, long deletion time is required because the free space, that is, until the free blocks are almost exhausted and the blocks are cleaned up at once. In contrast, the content-based erase policy does not require a relatively long delete time because the blocks are arranged when the number of free blocks drops below the threshold.

In the case of content-based block management, there may be a difference in wear between pure blocks and mixed blocks. The reason for this is as follows. Pure blocks are often filled with only one large file. When one portion of such a large file is updated, the updated portion of the page is used for empty space in the mixed block. Pages of some of the updated files in a mixed block are written to other mixed blocks when re-updated. For this reason, as shown in FIG. 8 showing the number of wear of the pure block and the mixing block, it can be seen that the number of wear of the mixing block is higher than that of the pure block.

According to content-based garbage collection, wear leveling is achieved through block swapping. Referring to FIG. 9, which shows a block swap process for leveling wear, first, a content-based file system according to an exemplary embodiment of the present invention selects a pure block having the smallest wear from a pure block list. Thereafter, the content-based file system according to the exemplary embodiment of the present invention selects the free block having the highest wear level from the free block list. The content-based file system according to an exemplary embodiment of the present invention determines whether the difference between the wear of the selected pure block and the wear of the selected free block exceeds a certain range (eg, 10). If it is determined that the difference between the wear of the selected pure block and the wear of the selected free block exceeds a certain range (eg, 10) (or a reference value), the file system according to an exemplary embodiment of the present invention may select the selected free block and the selected free block. We will perform a block swap where pure blocks are swapped. In an exemplary embodiment, block swap will be performed at idle time.

The mount time of the content-based block management policy according to an embodiment of the present invention and the life efficiency of the flash memory through wear management in garbage collection may be measured using various actual data. For example, the performance of content-based file systems and existing file systems (eg, YAFFS and JFFS2) in accordance with an embodiment of the present invention will be compared with reference to FIGS. 10-16. To do this, the following two experiments were performed in Linux 2.6.17 environment. For the first experiment, the mount average time of the existing file systems and the content-based file system according to the embodiment of the present invention will be compared. In the second experiment, the lifetime efficiency of the flash memory is measured to compare the erasure policies of the existing file system (eg, YAFFS) and the content-based file system according to the embodiment of the present invention. The same simulation is used.

(1) Wear measurement simulation: consists of measuring the number of block erase commands, measuring the number of multi-block delete commands, measuring the number of blocks requested to be erased, and measuring the number of consecutive blocks requested to be erased.

(2) Postmark: Performs transactions such as writing, reading, changing, and deleting small files. The number of files is 100 and the transactions to be executed are set to 200 and 500, respectively.

10 is a view showing a mount average time of a content-based flash file system and a conventional file system according to an embodiment of the present invention.

The mount average time refers to the time taken to scan spare areas of blocks in flash memory to organize file information on main memory. The parameters used in the experiment are as follows. As experimental data, nine files with various sizes ranging from 64Kbyte to 16Mbyte were used. As shown in FIG. 10, the mount average time of the content-based file system according to an embodiment of the present invention is 82.2% and 42.9%, respectively, when compared to existing file systems (eg, JFFS2 and YAFFS), respectively. Decreases by. In addition, in the content-based file system according to an embodiment of the present invention, it can be seen that the mount average time is reduced by up to 91.7% and 67% when the file size is large rather than when the file size is small.

 11 to 13 are diagrams showing the degree of wear when sequential write, local write, random write.

Sequential write means that data is written by allocating a general free block, and local write means that data written to a block is updated and written to another block. Random write means a case where sequential write and local write occur alternately. As can be seen in FIGS. 11-13 showing the results obtained using Postmark, the erase policy of a content-based file system according to an embodiment of the present invention is compared with Greedy, Cost-Benefit, and CAT erase policies. There is no overhead. In addition, it can be seen from FIG. 11 to FIG. 13 that the erase policy of the content-based file system according to an embodiment of the present invention has no overhead in copying and deleting times compared to Greedy, Cost-Benefit, and CAT erase policies. .

14 and 15 are diagrams showing the results of experiments on the leveling of the wear of the existing file system and the content-based file system according to an embodiment of the present invention.

In the existing file system (eg YAFFS), the number of times of wear varies greatly from block to block, and only some of the blocks used are still in use. However, the content-based file system according to an embodiment of the present invention can be seen that the wear of the blocks is kept uniform. In the postmark simulation, we set the transaction to 200 and 500, respectively, and repeated file deletion and copying.

16 is a view showing the number of times of copying and erasing according to an erase policy and an existing erase policy of a content-based flash file system according to an exemplary embodiment of the present invention. As illustrated in FIG. 16, it can be seen that the erase policy of the content-based file system according to the embodiment of the present invention is reduced by 32% of the number of times of erasing or copying of the content-based file system. According to an erase policy of a content-based file system according to an exemplary embodiment of the present invention, a free block and a pure block are swapped by comparing a minimum value of a free block with a maximum value of a pure block. However, when the swap policy is used in the existing erasure policy, valid data may be used in the block to be swapped if only the maximum and minimum values of wear are compared without classifying the blocks. In this case, the additional new block is set as a temporary block and used during the swap process, which causes additional copy and delete times. Therefore, the erase policy of the content-based file system according to the embodiment of the present invention prevents additional overhead because the swap policy is used by comparing the number of wear of the free block and the pure block.

It will be appreciated that the content-based flash file system according to an embodiment of the present invention is not limited to flash memory but may be applied to file systems of other storage values.

It will be apparent to those skilled in the art that the structure of the present invention can be variously modified or changed without departing from the scope or spirit of the present invention. In view of the foregoing, it is intended that the present invention cover the modifications and variations of this invention provided they fall within the scope of the following claims and equivalents.

1 illustrates a general flash memory structure.

2 illustrates a conventional YAFFS flash memory structure.

3 is a block diagram schematically illustrating a system according to an exemplary embodiment of the present invention.

4 is a diagram for describing how a content-based file system determines a block type according to an exemplary embodiment of the present invention.

Fig. 5 is a flowchart for schematically illustrating a method of mounting a content-based file system according to an exemplary embodiment of the present invention.

6A and 6B illustrate a method of mounting an existing flash file system (eg, YAFFS) and a method of mounting a content-based file system according to an exemplary embodiment of the present invention.

7A to 7C are diagrams for describing a garbage collection method of a content-based file system according to an exemplary embodiment of the present invention.

8 is a view showing the number of wear of the pure block and the mixing block.

9 is a view illustrating a block swap process for leveling wear.

10 is a view showing a mount average time of a content-based flash file system and a conventional file system according to an embodiment of the present invention.

11 to 13 are diagrams showing the degree of wear when sequential write, local write, random write.

14 and 15 are diagrams showing the results of experiments on the leveling of the wear of the existing file system and the content-based file system according to an embodiment of the present invention.

16 is a view showing the number of times of copying and erasing according to an erase policy and an existing erase policy of a content-based flash file system according to an exemplary embodiment of the present invention.

Claims (20)

A processing unit; A RAM into which a content-based file system operated by the processing unit is loaded; And A flash memory having blocks managed by the content-based file system, The content-based file system selects at least one block of the blocks when a file write is requested, and determines whether the selected at least one block belongs to one of a pure block, a mixed block, and a free block; And storing the write requested file in the at least one block with block type information indicating the determined block type. The method of claim 1, And if the at least one block is filled with the write requested file, the content-based file system determines the at least one block as the pure block. The method of claim 2, And if some of the at least one block is filled with the write requested file, the content-based file system determines the at least one block as the mixed block. The method of claim 1, Determination of whether the selected at least one block belongs to one of pure blocks, mixed blocks, and free blocks divides the size of the write requested file by the size of the block, and divides the blocks corresponding to the shares into pure blocks and the rest. A system performed by determining the corresponding block as a mixed block. The method of claim 1, The content-based file system reads block type information from a spare area belonging to the first page of each block when mounted, and determines scanning of spare areas of the remaining pages of each block according to the read block type information. . The method of claim 5, The content-based file system inserts the selected block into the mixed block list after scanning the spare areas of the remaining pages of the selected block when the read block type information indicates the mixed block, Open file table by inserting the selected block into either the free block list or the pure block list without scanning the spare areas of the remaining pages of the selected block when the read block type information indicates one of a free block and a pure block. To configure the system. In a block management method of a system including a nonvolatile memory composed of blocks: Reading block type information from a spare area belonging to a first page of a selected block among the blocks; When the read block type information indicates a mixed block, scanning the spare areas of the remaining pages of the selected block and inserting the selected block into the mixed block list; When the read block type information indicates one of a free block and a pure block, inserting the selected block into either a free block list or a pure block list without scanning the spare areas of the remaining pages of the selected block; ; And Repeatedly performing the reading and inserting steps until all the blocks of the nonvolatile memory are selected. The method of claim 7, wherein And the block type information is generated when a file is stored in the nonvolatile memory. The method of claim 7, wherein Performing content-based garbage collection based on the pure block list, the mixed block list, and the free block list when the number of free blocks included in the free block list is lower than a threshold. How to. The method of claim 9, The content-based garbage collection selects mixed blocks having a small number of valid pages in the mixed block list, moves valid pages included in the selected mixed blocks to one of the free blocks in the free block list, and Achieved by inserting selected mixed blocks into the free block list as free blocks. The method of claim 7, wherein And performing a block swap for leveling wear of the blocks of the nonvolatile memory, wherein the block swap selects a pure block having the smallest wear in the pure block list, and the largest wear in the free block list. And selecting a free block having a predetermined value and swapping the selected pure block and the selected free block when a difference between the wear degree of the selected pure block and the wear degree of the selected free block exceeds a reference value. The method of claim 11, And said block swap is performed at idle time. The method of claim 7, wherein The block type information is stored in the reserved area belonging to the tag of the spare area. In a block management method of a system including a nonvolatile memory composed of blocks: Determining a type of at least one block in which the write requested file is to be stored among the blocks of the nonvolatile memory; And And storing the write request file in the at least one block together with block type information indicating the determined type. The method of claim 14, Wherein the block type information indicates any one of a pure block, a mixed block, and a free block. The method of claim 15, The at least one block is determined as being a pure block when the at least one block is determined to be filled with the write requested file and when it is determined that some of the at least one block is filled with the write requested file. Characterized in that it is determined by a mixing block. The method of claim 15, The size of the write request file is divided by the size of a block, wherein blocks corresponding to the share are determined as pure blocks and blocks corresponding to the remainder as mixed blocks. The method of claim 15, The block type information is stored in a tag of a spare area belonging to the first page of the at least one block, and the meta information of the write request file is stored as a header in a data area belonging to the first page of the at least one block. Characterized in that the method. The method of claim 14, When mounted, reading block type information from a spare area belonging to the first page of each block of the nonvolatile memory; And And determining a scan of spare areas of the remaining pages of each block according to the read block type information. The method of claim 19, When the read block type information indicates a pure block or a free block, scanning of spare areas of the remaining pages of the block determined as the pure block or the free block is omitted; And And when the read block type information indicates a mixed block, spare areas of remaining pages of the block determined as the mixed block are scanned.

Images