TW201640378A - File management and access system and method thereof - Google Patents

Abstract

A file management and access system and a method thereof are disclosed. The invention utilizes a distributed memory cache as a storage medium, wherein a client end uses a Filesystem in Userspace (FUSE) to interface with a local file system, so as to provide the client end with a transparent file operation mechanism. By means of utilizing batch transfer and multiple buffer techniques, the invention effectively transmits massive file access requests generated by the FUSE to a remote machine. After remotely dissembling and assembling file data, the invention saves the files as key values into a memory cache cluster composed of a plurality of hosts via distributed memory storage software, so as to increase synchronous access capability. Meanwhile, the invention utilizes a distributed lock to avoid metadata damage problem due to massive synchronous access.

Description

File management and access system and method

The present invention relates to a file management and access system and method, and more particularly to a file management and access system and method using a distributed memory cache as a storage medium.

At present, the data processing process (such as: accounting processing) contains a large number of multi-stage file processing and conversion, and each stage will initiate multiple parallel processing of the same processing program to improve processing efficiency. This process requires a storage medium to exchange files between stages that need to be able to maintain stable efficiency with a large amount of concurrent access. At this stage, the Network File System (NFS) is used as a storage medium method, and there are obvious performance impairments when multiple processors are simultaneously accessed.

Replacing the file system with a single-machine memory as a medium for exchanging files can improve efficiency, but has the following limitations: 1. It requires a host with a large amount of memory, which is expensive; 2. Speed is limited by the slowest stage, not suitable Computationally intensive applications; 3. Not easy to expand.

In view of the problem that the speed of the network and the memory are not equal, in the above application, it is necessary to read and write a large amount of the remote memory in a short time, so there is a consideration of different network speeds and memory access speeds. In US Patent Publication No. US 20140019680, a heterogeneous file system that can be a unified interface of a client space file system is disclosed, but a large amount of remote access through a network is not considered, so in an environment with a low network speed, Because of the above problems, the performance bottleneck is caused.

In addition, for the problem of metadata management and updating in a decentralized environment, U.S. Patent Publication No. US 20140040286 discloses a design method for a distributed file system that combines functions. However, this patent publication only focuses on the conversion and normalization of archival materials, and does not solve the problem of efficiency of synchronous access. In the case of massive access by multiple clients, this patent publication does not propose a mechanism to resolve synchronization conflicts or protection.

In view of the above-mentioned problems of the prior art, the object of the present invention is to provide a file management and access system and method for improving synchronous access capability and avoiding metadata corruption caused by a large number of simultaneous accesses.

The file management and access system of the present invention comprises a client space file system interface component, a client transport mediation component and a server-side transport mediation component. The client space file system interface component converts a request command issued by the client space file system to become the first file processing unit; the client The transport mediation component includes a first multiple buffer, and after receiving the first file processing unit, the first file processing unit is respectively configured to configure one of the first multiple buffers, and one part of the first multiple buffer When the data is full, the data in the first multiple buffer is transmitted; the server-side transport intermediary component includes a second multiple buffer, and the second file processing unit is generated according to the first file processing unit, and the second file is processed. The unit respectively configures one part of the second multiple buffer, and when one part of the second multiple buffer is filled, the data in the second multiple buffer is transmitted.

The file management and access method of the present invention comprises the steps of: converting a request command issued by a client space file system into a first file processing unit; and configuring a first file processing unit to respectively configure one of the first multiple buffers; Transmitting the data in the first multiple buffer when one of the multiple buffers is full; generating the second file processing unit according to the first file processing unit, and correspondingly configuring the second file processing unit One of the multiple buffers; and transmitting the data in the second multiple buffer when one of the second multiple buffers is full.

Wherein, one of the first multiple buffer and one of the second multiple buffers are available for reading while the data is being written, another portion of the first multiple buffer and another portion of the second multiple buffer.

The first multiple buffer and the second multiple buffer comprise a plurality of buffer sub-regions, wherein two of the buffer regions of the first multiple buffer and the second multiple buffer are alternately used, that is, the First multiple buffer and the second One part of the multiple buffer and the other part, and the remaining buffer area is used to temporarily store the data.

Wherein, when the two buffer sub-areas alternately used by the first multiple buffer and the second multiple buffer are erased, if data is still not transmitted, the data is transferred to the remaining buffer regions for temporary storage. .

The client transport intermediate component and the server transport intermediate component further include a transport component, which detects whether the remaining buffer regions are filled, and when the remaining buffer regions are filled. The transmission element transmits the data temporarily stored in the remaining buffer sub-areas.

As described above, the file management and access system and method according to the present invention may have one or more of the following advantages:

1. Compared with NFS, it is difficult to maintain performance during a large number of simultaneous accesses. The technology disclosed by the present invention has horizontal expansion capability, which can increase the number of clients as the number of clients increases, and provides a larger and more distributed memory cache to alleviate a large number of synchronizations. Access requirements.

2. The technology disclosed in the present invention can avoid the network transmission as a bottleneck and play the efficient reading and writing ability of the memory.

3. The present invention utilizes a buffer to buffer the processing speed of different components, and then uses parallel processing to hide the speed difference as much as possible.

1100‧‧‧Client

1110‧‧‧Application

1120‧‧‧File Management and Access System

1121‧‧‧Customer Space File System Interface

1122‧‧‧Client Intermediary Software

1130‧‧‧Client Operating System

1131‧‧‧Virtual File System

1132‧‧‧Customer Space File System

1133‧‧‧Unix file system

1200‧‧‧ server side

1210‧‧‧File Management and Access System

1211‧‧‧Server-side mediation software

1220‧‧‧Storage memory software

1230‧‧‧Distributed memory cache

2100‧‧‧Files

2101‧‧‧ Metadata

Blocks 1~5 from 2012~2106‧‧

2200‧‧‧Key value storage space

2201~2204‧‧‧ key value

2300‧‧‧Lock

3110, 3210‧‧ ‧ file processing unit

3111, 3211‧‧‧ multiple buffering

3120, 3220‧‧‧ transmission components

4001~4006, 5001~5005‧‧‧ steps

Figure 1 is a schematic diagram of the system architecture of the file management and access system of the present invention.

Figure 2 is a schematic diagram of the key value storage mode of the file management and access method of the present invention.

FIG. 3 is a schematic diagram of an embodiment of a client intermediary software and a server-side intermediary software of the file management and access system of the present invention.

Figure 4 is a flow chart of an embodiment of writing a file for the file management and access method of the present invention.

Figure 5 is a flow chart of an embodiment of a file archive of the file management and access system of the present invention.

The file management and access of the present invention comprises the steps of: translating a request for an archive into a request for one or more data blocks; establishing a multi-buffer on each of the client and the server (eg: first multi-buffering) The file processing unit of the area and the second multiple buffer (for example, the first file processing unit and the second file processing unit); if the client needs to transmit data, the data to be transmitted is written into the client file processing unit. Buffer, the server-side file processing unit detects the change of the client's multiple buffers, and pulls the data to the server-side file processing unit; if it is a write request, it needs to obtain all the upper-layer folder locks and update before writing. Metadata of all upper folders, and release the lock of the upper folder; if it is a write request, it will first collect the data block to be written to a certain size and then write; If it is a read request, the confirmation file is not being written; if it is a read request, the data to be read is read from the storage memory software and then stored in the server side multiple buffer, and the client file processing unit detects the client. The multi-buffer change changes the data to the server-side file processing unit; if it is a write request, obtains the lock of the file to be read and written, updates the metadata of the file, and releases the lock of the file to be read and written; As a read request, the block data is assembled into a file.

Please refer to FIG. 1 , which illustrates the interaction between the file management and access system 1120&1210 (Distributed Virtual Memory File System, DVMFS) and the operating system and the decentralized memory cache.

When the application 1110 of the client 1100 initiates a request to read the file, the request is transmitted to the virtual file system 1131 (VFS) in the core space (Kernel Space) 1130 of the client operating system. The virtual file system recognizes that the file of this request is located in the directory where the file management and access system is mounted, so the request is transmitted to the client system file system 1132 (Filesystem in Userspace, FUSE for short) instead of the Unix file system. ) 1133. The client space file system 1132 can accept file requests for the virtual file system, thus allowing application 1110 transparent operational files to be provided. The client space file system 1132 transmits the request to the client space file system interface 1121 of the file management and access system 1120 of the client (also referred to as the client space file system interface component of the present invention), and the client space file system interface 1121 translates. Customer space file The request of the file system 1132 forwards the request to the client mediation software 1122 (Middleware, also referred to as the client transport mediation component of the present invention). The client intermediary software 1122 sends a request for reading the file to the server-side mediation software 1211 of the file management and access system 1210 of the server 1200 (also referred to as the server-side transport intermediary component of the present invention), and the server-side intermediary The software 1211 checks the lock of the file (Lock) to ensure that there is no read/write conflict, and calculates the key value section of the file, and obtains the required file from the distributed memory cache 1230 through the memory storage software 1220 (Memory Storage Software). The fragment is transmitted to the client intermediary software 1122 in the file management and access system 1120 of the client 1100 via the network batch when the file fragment bit code is aggregated to a certain size or after the assembly is completed. The multi-buffer mechanism is activated at the same time of transmission, and the server-side mediation software 1211 can continue to obtain the required file segment bit code from the storage memory software 1220 during transmission without interrupting the processing due to the transmission. The client mediation software 1122 also initiates a multiple buffering mechanism upon receiving, on the one hand, the data is passed back to the client space file system interface 1121, and on the other hand, the next data segment prepared by the server-side mediation software 1211 is also waiting. After the client space file system interface 1121 receives all the requested data, it transmits it to the client space file system 1132, transfers it to the virtual file system 1131, and transmits it back to the application 1110.

When the application 1110 of the client 1100 initiates a request to write a file, the request and the data to be written are transmitted to the virtual file system virtual file system in the Kernel Space 1130 of the client operating system, 1131, the virtual file system. The file identifying this request is located in the mounted file management and The system's directory is accessed, so this request is passed to the customer space file system 1132 instead of the Unix File System 1133. The client space file system 1132 can accept file requests for the virtual file system, thus allowing application 1110 transparent operational files to be provided. The client space file system 1132 transmits the request to the client space file system interface 1121 of the file management and access system client 1120. The client space file system interface 1121 translates the request from the client space file system 1132 and hands the request to the client mediation software. 1122. The client mediation software 1122 aggregates the written data received by the client space file system interface 1121. When the data is aggregated to a certain size or after the aggregation is completed, the file management and access system 1210 is transmitted to the server terminal 1200 through the network batch. Server-side mediation software 1211. The client mediation software 1122 simultaneously starts the multiple buffering mechanism, and continues to collect the data from the client space file system interface 1121 during the transmission, and the server-side mediation software 1211 locks and updates the metadata of the upper layer folder (Metadata) to ensure no reading. After writing the conflict and the consistency of the meta-data, the key value segment corresponding to the file segment is calculated, and the data is written into the decentralized memory cache 1230 through the storage memory software 1220, and the server-side mediation software 1211 also starts multi-buffering. The mechanism processes the write request of the next data segment of the client mediation software 1122 at the same time. When the file is written, the metadata of the file is locked and updated.

Hereinafter, the process of the server-side mediation software 1211 calculating the key value section of the file and storing the file-related data in the key memory (Key-Value) by the storage memory software 1220 into/reading the distributed memory cache 1230 is described. .

There are two types of files that can be written to the file management and access system, one is a common file, and the other is a file folder. Each file corresponds to three kinds of data in the file management and access system: Metadata, Data, and Lock: Metadata: Basic data related to the file, such as the permissions and size of the file. Establish time, last modified time. The metadata of the folder file contains the file name under this folder to support the Tree Traverse.

Data: The content of the file, with the file block as the access unit, and one file consists of one or more file blocks.

Lock: Protects access to metadata, ensuring that metadata is not corrupted due to large amounts of simultaneous access.

The above data is stored in the decentralized memory cache 1230 by the server-side mediation software 1211. The key values are as follows: meta-data: each meta-data corresponds to a decentralized memory cache key combination, and the key is the path of the file. Or the value calculated by the path, the value is the metadata content.

Data: Multiple file blocks are aggregated into one large data block and then stored in a decentralized memory cache. Therefore, a single distributed memory cache key combination can contain multiple file blocks. This batch write practice reduces the number of accesses to the decentralized memory cache. The key is "the path of the file, the range or list of saved files" or the value calculated by it.

Lock: Each lock corresponds to a key-value combination of a decentralized memory cache. Key is Based on the value calculated by the path of the file, the value content can be customized (for example, 1 is locked and 0 is unlocked).

When writing and deleting files, locks are used to ensure no read/write conflicts and metadata consistency. The process of writing data is as follows: obtaining the lock of the upper folder of the file; updating the metadata of the upper folder; releasing the upper folder Lock; write data; after all the data of the file is written, obtain the lock of the file; write the metadata of the file; and release the lock of the file.

The process of deleting the data is as follows: obtaining the lock of the upper folder of the file; deleting the record of the file in the folder; updating the metadata of the upper folder; releasing the lock of the upper folder; obtaining the lock of the file; obtaining and deleting the file Metadata at the far end; release the lock of the file; calculate the key value corresponding to all the data according to the metadata of the file just taken out; and delete all the data.

Please refer to FIG. 3, which illustrates a preferred embodiment of the client-side mediation software 1122 and the server-side mediation software 1211 of the file management and access system.

The client mediation software 1122 includes one or more file processing units 3110 and a single threaded transmission component 3120. Each file processing unit 3110 processes only one file request from the client space file system interface 1121. If there are more than two file requests at the same time, it will be processed by multiple file processing units 3110. Each file processing unit 3110 includes a multi-buffer 3111, and the multi-buffer 3111 contains at least two fixed-size buffers. When one of the buffers is writing data, the transport component 3120 can simultaneously send other filled buffers. Information to reduce buffer delays. The transmission component 3120 transmits data only when the buffer is full or the file processing unit 3110 requests the transmission, so as to improve the network transmission efficiency. The transport component 3120 only contains a separate thread processing, on the one hand to reduce the use of the thread to improve stability, and on the other hand, because of the use of the buffer technology, the network bandwidth can be maximized, in this case additional threads Will not increase too much performance.

The server-side mediation software 1211 includes one or more file processing units 3210 and a single-threaded transmission component 3220. Each file processing unit 3210 processes only one data return request from the transmission component 3120. If there are more than two data return requests at the same time, it is processed by multiple file processing units 3210. Each file processing unit 3210 contains a multiple buffer 3211. The multi-buffer 3211 includes at least two fixed-size buffers. When one of the buffers is writing data, the transmission component 3220 can simultaneously send other data that has been filled in the buffer to reduce the buffering delay. The transmission component 3220 transmits data only when the buffer is full or the file processing unit 3210 requests transmission, usually at the end of data writing, to improve network transmission efficiency. The transport component 2330 only includes a separate thread processing, on the one hand to reduce the use of the thread to improve stability, and on the other hand, because of the use of the buffer technology, the network bandwidth can be maximized, in this case additional threads There won't be much performance gains. In each data backhaul request, the transport component 3220 may need to request one or more key values from the storage memory software 1220.

Multiple buffers 3111 & 3211 are generally used in two buffers, but more than two buffers can be used. In a preferred embodiment, the big_write option of the client space file system 1132 is enabled, and the client space file system 1132 sends a read request to the client mediation software 1122 in a multi-thread mode through the client space file system interface 1121 to speed up reading. Take efficiency. However, multiple threads can cause non-linear use of multiple buffers, that is, the buffer contents are not used sequentially, so that when the buffer tail content is consumed, there is no guarantee that the buffer header content has been consumed. Therefore, the third buffer can be used as a temporary storage, and the data to be erased is temporarily stored in the third buffer to avoid erasing the unconsumed data. If this method is used, if the data is not found in the preset two buffers when reading data from multiple buffers, the third buffer must be searched. Similarly, you can add a fourth, fifth, or even more Multiple buffers are staged to match the parallel effect of the client space file system big_write option.

A preferred embodiment uses Memcached as the storage memory software 1220. Memcached is a well-known open source memory storage software that supports multi-language interface key-value access. It is the first choice for many network system cache functions. Compared with other memory caches, Redis is also effective.

The client space file system interface 1121 is responsible for the translation of the read and write requests. A preferred implementation is a request to translate the data read and write request {path, offset, data size} into {path, {data block number}}, translation formula For: data block number = offset / data block size, if the offset plus the required data size exceeds the size of one data block, you need to request multiple data blocks at the same time.

A preferred key value storage method is: Metadata: The lock format is [Path]

Data: The lock format is [block number start value: block number end value: path], such as 1:250:/data/doc.txt to store the 1~250 file blocks of the /data/doc.txt file. .

Lock: The key format is [L: Path].

Please refer to FIG. 2, which is an example of the preferred key value storage mode. The file 2100 of the file 2100 with a path of abc.txt corresponds to the key value storage space 2200. The key is abc.txt2201; If the block 1 to block 42102-2105 of the file is stored in the key value storage space 2200, the key is 1:4:abc.txt2202; if the file block 5 2106 is stored in the key value storage space 2200, the key is 5 :abc.txt2203; and the lock of the file 2300 stored in the key value storage space 2200 is L:abc.txt2204.

The following describes the processing method of the file reading and writing in the client intermediary software 1122 and the server side intermediary software 1211.

An example of writing a file (using the cp command to copy a file from the client's native file system to the file management and access system), see Figure 4: The client application 1110 sends a cp/src/abc.txt/ The dest/folder/abc.txt command goes to kernel1130. Kernel1130 determines that the file system mounted in the target directory /dest or /dest/folder is the file management and access system, so the file write request is directed to the client space file system interface 1121. . The format of the write request is: {path, offset, size of the data to be written}. If the offset is 20 KB, the size of the data to be written is 1 KB, indicating that the data to be written by the request is the 20th to 21KB of the file.

Step 4001: The client space file system interface 1121 translates the data write request into a request of {path, {data block number}}, and the translation formula is data block number=offset/data block size, if offset plus If the required data size exceeds the size of one data block, multiple data blocks are required at the same time, and The client intermediary software 1122 is required to obtain these data blocks.

Step 4002: After receiving the request, the client mediation software 1122 creates an archive processing unit 3110 and configures a multiple buffer 3111 to request the client space file system interface 1121 to write the data into the multiple buffer 3111. The write request is transmitted to the server-side mediation software 1211, and the server-side mediation software 1211 is required to establish an archive processing unit 3210, and a multi-buffer 3211 is configured to wait to receive the data to be written.

Step 4003: The server-side transmission component 3220 will write the request {path, {data block number when the buffer in the multiple buffer 3211 is full or the file processing unit 3210 requests to transmit, usually at the end of the data writing. }, the buffer data} is transferred to the storage memory software 1220.

Step 4004: If the write request is the first buffer data received by the path, that is, the first write of the file, the following request is first sent to the storage memory software 1220: Get "L" :/dest" If the result is 1 indicating that another program is updating this directory, wait for a while and try again; if the result is 0, proceed to the next step; Put "L:/dest", 1, to obtain the lock of this directory; Put "/dest", Get "L:/dest".append "folder", modify the metadata of the /dest directory; Put "L:/dest", 0, release the lock of this directory; Get "L:/dest/ Folder" If the result is 1 indicating that other programs are updating this directory, wait for a while and try again; if the result is 0, proceed to the next step; Put "L:/dest/folder", 1, get the lock of this directory; Put "/dest/folder", Get "L:/dest/folder".append "abc.txt", modify the /dest/folder directory Metadata; and Put "L:/dest/folder", 0, releases the lock for this directory.

Step 4005: The server-side transmission component 3220 aggregates the buffer data according to the block sequence. When the upper limit of the size of the 1 MB Memcached single data is aggregated or assembled to the last block of the file, the write data is sent to the storage memory software 1220. Requirements. If the block is the 1-250th block of the file, the write request is: Write "1:250:/dest/folder/abc.txt", the binary data of the 1-250th block.

Step 4006: After the server-side transmission component 3220 writes all the contents of the file, the following request is sequentially sent to the storage memory software 1220: Get "L:/dest/folder/abc.txt", if the result is 1, it indicates that there are other The program is updating the metadata of this file, waiting for a while and then retrying; if the result is 0, proceed to the next step; Put "L:/dest/folder/abc.txt", 1, get the lock of this file; and Put "/dest/folder/abc.txt", Metadata, modify the meta-data of the /dest directory. The metadata includes the permissions, size, creation time, and last modification time of the file.

An example of reading a file (using a cp command to copy a file from the file management and access system to the client's native file), see Figure 5 for instructions: The client application 1110 sends a command of cp/src/abc.txt/dest/abc.txt to kernel1130, which is copied from the file management and access system/src/abc.txt into the client file system/ Dest/abc.txt, the kernel determines that the file system mounted in the source directory/src is the file management and access system, so the file write request is directed to the client space file system interface 1121. The format of the read request is: {path, offset, size of the data to be read}.

Step 5001: The client space file system interface (1121) translates the data write request into a {path, {data block number}} request, and the translation formula is data block number=offset/data block size, if offset In addition, if the required data size exceeds the size of one data block, multiple data blocks are required at the same time.

Step 5002: After receiving the request, the client mediation software 1122 creates an archive processing unit 3110 and configures a multiple buffer 3111 to prepare to receive data from the server-side mediation software 1211. The read request is transmitted to the server-side mediation software 1211, and the server-side mediation software 1211 is required to establish an archive processing unit 3210, and a multi-buffer 3211 is configured to wait for the server-side transport component 3220 to write data.

Step 5003: The server-side transmission component 3220 issues a request to read the metadata to the storage memory software 1220: Get "L:/src/abc.txt", confirming that the file is not being written. And sequentially read the request of each data block, and write to the multi-buffer 3211 of the server-side transmission component. For example, if the multi-buffer is empty at the beginning, the size of the Buffer is read: Get "1:100/src/abc .txt", if there is 1:100 in the multi-buffer, read the next data block: Get "101:200/src/abc.txt".

Step 5004: When any of the buffers in the Double Buffer 3211 is full or all the data blocks have been read, the buffer data is transferred to the multiple buffer 3111 of the client file processing unit 3110.

Step 5005: The client space file system 1132 obtains the data in the multiple buffer 3111 through the client space file system interface 1121. After all the data is obtained, a file is generated, and the file file is written by the virtual file system 1131 to /dest/abc.txt. .

In summary, the file management and access system and method of the present invention is more difficult to maintain performance in a large number of simultaneous accesses compared to NFS. The present invention has the capability of horizontal expansion, which can increase the number of clients and increase the number of clients. Larger and more fragmented memory caches to alleviate a large number of simultaneous access needs. Moreover, through the invention, the network transmission can be prevented from becoming a bottleneck, and the efficient reading and writing ability of the memory should be exerted. .

The above is intended to be illustrative only and not limiting. Any equivalent modifications or alterations to the spirit and scope of the invention are intended to be included in the scope of the appended claims.

5001~5005‧‧‧Steps

Claims (10)

A file management and access system comprises: a client space file system interface component, converting a plurality of request commands issued by a client space file system into a plurality of first file processing units; and a client transport intermediary component comprising one a first multiple buffer, and after receiving the first file processing units, respectively configuring each of the first file processing units to configure one of the first multiple buffers, and one of the first multiple buffers Transmitting a portion of the first multiple buffer when the portion is full; and transmitting a mediation component to the server, including a second multiple buffer, and generating the first file processing unit according to the first file processing unit a plurality of second file processing units, and each of the second file processing units respectively corresponding to one part of the second multiple buffer, and transmitting the first part when the second multiple buffer is full Information on one part of the two multiple buffers. According to the file management and access system of claim 1, wherein one of the first multiple buffer and one of the second multiple buffers are written in the data, the first multiple buffer is another A portion and another portion of the second multiple buffer can provide read. According to the file management and access system of claim 2, wherein the first multiple buffer and the second multiple buffer comprise a plurality of buffer sub-areas, wherein two of the buffer sub-areas are alternately used That is the first multiple buffer The area and one part of the second multiple buffer and the other part, and the remaining buffer area is used for temporary storage of data. According to the file management and access system of claim 3, wherein the two buffer sub-areas that are used alternately are transferred to the remaining buffer sub-areas if the data is not transmitted when being erased. . According to the file management and access system of claim 3, further comprising a transmission component, detecting whether the remaining buffer sub-areas are filled, and when the remaining buffer sub-areas are filled The transmission element transmits the data temporarily stored in the remaining buffer sub-areas. A file management and access method includes: converting a plurality of request commands issued by a client space file system into a plurality of first file processing units; respectively configuring each of the first file processing units to configure a first multiple buffer a portion; when a portion of the first multiple buffer is filled, transmitting data of a portion of the first multiple buffer; generating a plurality of second files according to the first file processing unit Processing unit, and each of the second file processing units is configured to correspond to a portion of a second multiple buffer; and when one of the second multiple buffers is full, transmitting the second multiple Information on one part of the buffer. According to the file management and access method of the sixth application patent scope, The method includes the following steps: reading a portion of the first multiple buffer and the second multiple buffer when one of the first multiple buffer portion and one of the second multiple buffer portions are writing data The other part. According to the file management and access method of claim 7, wherein the first multiple buffer and the second multiple buffer comprise a plurality of buffer sub-regions, the method further comprises the steps of: alternately using the buffers Two of the sub-areas; and the temporary data in the remaining buffer sub-area. According to the file management and access method of claim 8 of the patent application scope, the method further comprises the following steps: if the alternately used two buffer sub-areas still have data not being transmitted when being erased, the data is temporarily stored to the remaining Buffer subzone. According to the file management and access method of claim 3, the method further comprises the steps of: detecting whether the remaining buffer sub-areas are filled; and when the remaining buffer sub-areas are filled, Transmitting data temporarily stored in the remaining buffer sub-areas.