TWI835039B - Index node allocation method, data processing device and computer-readable medium - Google Patents

Abstract

An index node (inode) allocation method, a corresponding data processing device and a corresponding computer-readable medium are provided. The method is used in a file system and includes the following steps: allocating a range of a plurality of continuous inodes to a directory in the file system when a directory order mode is enabled, wherein files in the directory are sorted according to hash values of the files; dividing the range of inodes into a plurality of sub-regions; and when a new file is created in the directory, selecting one of the sub-regions according to a hash value of the new file, so as to allocate an inode in the selected sub-region to the new file.

Description

Index node configuration method, data processing device and computer-readable medium

The present invention relates to an index node (inode for short) of a file system, and in particular, to an index node configuration method, a data processing device and a computer-readable medium.

Currently, some file systems, such as the ZFS file system, use index nodes to store metadata such as file creation time, modification time, access mode, size, etc., as well as the storage location of file data in storage devices such as hard disks. In order to quickly look up the metadata of the file, the hash algorithm can be used. For example, the directory can be divided into multiple blocks, and the uniform dispersion characteristics of the hash algorithm can be used to evenly map multiple files in the same directory to each block, and then record the index node number of each file in each block. In this way, when you need to query the metadata of a file, you only need to scan the directory block corresponding to the hash value of the file to obtain the index node number of the file to read the metadata without scanning the entire file. contents.

However, the decentralized nature of the hash algorithm is not conducive to directory traversal. In addition, the general file system allocates consecutive index nodes to be configured to each file one by one according to the chronological order of file creation. However, continuously created files may not be in the same directory, so the same directory The location of the index node of the recorded file is not related, but is distributed. Therefore, when it is necessary to traverse all files and their metadata in a large directory containing many files, index nodes scattered in a large storage space must be read randomly, resulting in slow processing speed.

In addition, the above metadata can also include the extended attributes (EA) and/or access control list (ACL) of the file. Since these metadata have no length limit, it is easy to change the extended attributes and / Or the access control list has a large number, so that the header of its index node cannot be stored, and additional index nodes must be configured to store these metadata. This will make the index nodes more and more scattered, further reducing the directory details. Column processing speed.

In the prior art, the index node is enlarged to store a large amount of metadata, such as extended attributes and/or access control lists. Such metadata can be called embedded in the index node. However, in the metadata When the number is small, storage space will be wasted, so it is not easy to predict the appropriate size of the index node, and because the index node increases, its storage space must increase accordingly, that is, the dispersion range of the index node is expanded, and the index node is dispersed The problem is getting worse, so a technology is needed to at least solve the above problems.

In order to solve the above problems, the present invention provides an index node configuration method. The index node configuration method is used in a file system and includes the following steps: when the directory sequence mode is activated, the directory system configuration in the file system includes consecutive plural indexes. a sequence of nodes, wherein the files in the directory are sorted according to the hash value of the files; the sequence is divided into a plurality of subsequences; and, when a new file is created in the directory, the hash value of the new file is One of the subsequences is selected to configure an index node of the selected one of the subsequences for use by the new file.

The present invention also provides a data processing device, which is installed with a file system to execute the above index node configuration method.

The present invention also provides a computer-readable medium, which is used in a data processing device and stores instructions to execute the above-mentioned index node configuration method.

In the present invention, the index nodes of files in the same directory are continuously arranged in the order of directory listing, so as to speed up the listing speed of large directories and their processing speed.

200,300: Table of Contents

210,310,410,610: sequence of index nodes

file37, file99: file

S100~S124: Method steps

Figure 1 is a flow chart of an index node configuration method according to an embodiment of the present invention.

Figures 2 to 6 are schematic configuration diagrams of index node configuration methods according to different embodiments of the present invention.

The following describes the implementation of the present invention through specific embodiments. Those with ordinary knowledge in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification.

Please refer to Figure 1, which is a flow chart of an index node configuration method according to an embodiment of the present invention. In this embodiment, the method is executed by a file system, and the method can be used for any directory in the file system.

In addition, in order to speed up directory listing, multiple files in the directory can be sorted according to the hash value of the files, where the hash value of the files can be the hash value of the names of the files, or A hash value calculated based on the metadata or other relevant data of these files. The method flow of Figure 1 will be described below.

First, in step S100, it is checked whether the directory meets the startup condition of directory order mode. For example, the startup condition can be that the total number of index nodes used by all files in the directory is greater than a preset threshold. If the directory does not meet the startup condition, the process proceeds to step S110. On the contrary, if the directory meets the startup condition, the process proceeds to step S120.

In step S110, the file system does not activate the directory sequence mode. At this time, the consecutive index nodes to be configured are configured for each file one by one according to the time sequence of file creation in the entire file system.

In step S120, the file system starts the directory sequence mode.

In step S121, when the directory sequence mode is enabled, the file system configures a sequence of k consecutive index nodes in unused index nodes for use by files in the directory. In one embodiment, the sequence can be expressed as (I _x , I _x+k-1 ), where I _x and I _{x +k-1} are the first and last index nodes of the sequence, and x is the The number of the first index node, k is an integer greater than 1, and x+k-1 is the number of the last index node.

In step S122, the sequence is divided into complex subsequences. In one embodiment, each subsequence includes the same number of index nodes. For example, k index nodes of the sequence can be divided into L subsequences, where L is an integer greater than 1, and these subsequences can be expressed as (I _x ,I _x+k/L-1 ), (I _x+k/L ,I _x+2k/L-1 ),...,(I _x+(L-1)k/L ,I _x+k-1 ).

In step S123, when a new file is created in the directory, one of the subsequences is selected according to the hash value of the new file.

In step S124, an index node of one of the selected sub-sequences is configured for use by the new file. For example, the r-th subsequence (I _x+rk/L , I _x+(r+1)k/L-1 ) among the subsequences can be selected as the selected subsequence, where r=F % L , F is the hash value of the name of the new file. In other words, r is the remainder obtained by dividing the hash value of the new file by the number L of the subsequences.

In another embodiment, the number of index nodes included in each subsequence may not be exactly the same or may be completely different. When a new file is created in the directory, the r-th subsequence among the subsequences can still be selected, and the index node in the r-th subsequence can be configured for use by the new file.

For example, when the number of index nodes included in each subsequence of the sequence is not exactly the same or completely different, the subsequences can be expressed as (I _x ,I _x+s1-1 ), (I _x+s1 ,I _x+s2-1 ),...,(I _x+s(L-1) ,I _x+sL-1 ), and record in a table the parameters (s1,s2, ...,sL), but when the file hash value is sufficiently randomly distributed, the index nodes can be configured more efficiently by using the same number of subsequence index nodes as mentioned above.

In addition, after the file system activates the directory sequence mode in step S120, the previously scattered index nodes of each file in the directory can be collectively moved to the sub-sequence corresponding to the file based on the above-mentioned correspondence between the files and the sub-sequence. sequence, in a configuration that conforms to the directory sequence mode. Therefore, if the algorithm used by the file system to calculate the hash value is random enough, it can ensure that each subsequence is used evenly and fully.

Figure 2 is a schematic configuration diagram of an index node configuration method according to an embodiment of the present invention. As shown in Figure 2, the directory 200 includes a plurality of files and a plurality of blocks.

In one embodiment, in the table of block 0 of the directory 200, the files are sorted according to the mantissa (i.e., the last four bits) of the hash value of the file name. Except for block 0, each file is Each block records the index node number assigned to each file, and the table in block 0 records the index number of each file. The correspondence between the mantissa and the block number recording its index node number. For example, the file system configures the sequence 210 including 128 consecutive index nodes numbered from 1025 to 1152 to the directory 200 for use by the files of the directory 200, and divides the sequence 210 into 16 subsequences, each subsequence includes 8 consecutive index nodes. For example, subsequence 0 of sequence 210 includes 8 consecutive index nodes numbered from 1025 to 1032, and subsequence 1 of sequence 210 includes 8 consecutive index nodes numbered from 1033 to 1040.

In this embodiment, the length of the hash value mantissa of the file is 4 bits, and 2 ⁴ is equal to 16, which is the number L of the above-mentioned subsequences. Therefore, the mantissa of the hash value of each file is equal to the hash value of the file divided by the subsequence. The remainder r obtained from quantity L. In other words, the index nodes of files with the same hash value mantissa are located in the same subsequence.

Take file file99 as an example. The mantissa of the hash value of file file99 is 0001, which corresponds to subsequence 1 and corresponds to block 5 in the table of block 0. Therefore, the file system can index an index in subsequence 1 of sequence 210. The node is configured for use by file file99. For example, the file system configures index node 1034 for use by file file99, and records the index node number 1034 of file file99 in block 5. After that, when the metadata of file file99 needs to be accessed, the index node number of file file99 recorded in block 5 can be obtained from the table of block 0, and the index node number of file file99 can be obtained from block 5 as 1034, and then in the index Node 1034 accesses the metadata of file file99.

In addition, when the index nodes previously allocated to any subsequence of the first sequence of a directory have all been allocated to files in the directory, the file system can configure the unused index nodes to also include consecutive k A second sequence of index nodes for use by files in this directory. In one embodiment, the second sequence can be expressed as (I _y , I _y+k-1 ), where I _y and I _y+k-1 are the first and last index nodes of the second sequence. , y is the number of the first index node of the second sequence, y+k-1 is the number of the last index node of the second sequence. The file system will divide the second sequence into plural subsequences.

In one embodiment, each subsequence includes the same number of index nodes. For example, the k index nodes of the second sequence can also be divided into L subsequences, and the subsequences can be expressed as (I _y , I _y+k/L-1 ), (I _y+k/L , I _y+2k/L-1 ),...,(I _y+(L-1)k/L ,I _y+k-1 ).

In addition, when the number of index nodes required by any file in the directory exceeds what the selected subsequence corresponding to the file in the first sequence can provide (for example, the file has just been created and requires one index node, and The index nodes of the selected subsequence have all been allocated for use by other files), and the file system selects one of the subsequences of the second sequence based on the hash value of the file (such as the hash value of the file name) to Configure an index node in the selected subsequence of the second sequence for use by the file. For example, the r-th subsequence (I _y+rk/L , I _y+(r+1)k/L-1 ) among the subsequences can also be selected as the selected subsequence, where r is the file The remainder obtained by dividing the hash value of the second sequence by the number L of subsequences of the second sequence.

In another embodiment, the number of index nodes included in each subsequence of the second sequence may not be exactly the same or may be completely different, wherein the subsequences may be expressed as (I _y , I _y+s1-1 ) ,(I _y+s1 ,I _y+s2-1 ),...,(I _y+s(L-1) ,I _y+sL-1 ), and record in a table the factors that determine the length of each subsequence Parameters (s1, s2,...,sL), but when the file hash value is sufficiently randomly distributed, the index nodes can be configured more efficiently by using the same number of subsequence index nodes as mentioned above.

Thereafter, when a new file is created in the directory and the new file has an index node in the selected subsequence in the first sequence that can be allocated to the new file, the file system configures the index node of the selected subsequence, for use with this new file. On the contrary, if all the index nodes of the selected subsequence have been allocated to other files, the file system allocates the index nodes of the selected subsequence corresponding to the file in the second sequence for use by the new file. In addition, when the index node of any subsequence in the second sequence has When all are configured for use by files in this directory, the file system can configure another sequence for use by files in this directory, and so on.

Figure 3 is a schematic configuration diagram of an index node configuration method according to an embodiment of the present invention. As shown in Figure 3, the file system configures the sequence 210 including 128 consecutive index nodes numbered 1025 to 1152 to the directory 300 for use by the archives of the directory 300, and divides the sequence 210 into 16 subsequences, each subsequence Both include 8 consecutive index nodes.

After that, the index nodes of at least one subsequence of the sequence 210 have all been configured for use by files in the directory 300. Therefore, the file system further configures the sequence 310 including 128 consecutive index nodes numbered from 1153 to 1280 to the directory 300 for use. The file of the directory 300 is used, and the sequence 310 is divided into 16 sub-sequences, each sub-sequence includes 8 consecutive index nodes.

Take the file file37 as an example. File file37 requires an index node. However, the index nodes of subsequence 1 corresponding to file file37 in sequence 210 have all been allocated to other files. Therefore, the file system configures the index nodes of subsequence 1 corresponding to file file37 in sequence 310 for file file37. use. For example, the file system configures index node 1161 for use by file file37, and records the index node number 1161 of file file37 in block 5.

In one embodiment, the file in the directory may have too much metadata to be stored in the index node of the file. Therefore, the file system can configure a third index node including wk consecutive index nodes in the unused index node. Three sequences for storing metadata of files. In one embodiment, the third sequence can be expressed as (I _z , I _z+wk-1 ), where I _z and I _z+wk-1 are the first and last index nodes of the third sequence. , w is a positive integer, z is the number of the first index node of the third sequence, z+wk-1 is the number of the last index node of the third sequence. The file system will divide the third sequence into plural sub-sequences.

In one embodiment, each subsequence includes the same number of index nodes. For example, the wk index nodes of the third sequence can also be divided into L subsequences, and the subsequences can be expressed as (I _z ,I _z+wk/L-1 ), (I _z+wk/L , I _z+2wk/L-1 ),...,(I _z+(L-1)wk/L ,I _z+wk-1 ).

In addition, when the metadata of a file in the directory is too large to be stored in the original index node of the file, the file system selects the third sequence based on the hash value of the file (such as the hash value of the file name). One of the subsequences is used to configure w index nodes in the selected subsequence of the third sequence for storing metadata of the file, where the metadata may include extended attributes of the file and /or access control list. For example, the r-th subsequence (I _z+rwk/L , I _{z+(r+1)wk/L-1} ) among the subsequences can also be selected as the selected subsequence, where r is the file The remainder obtained by dividing the hash value by the number L of subsequences of the third sequence.

In another embodiment, the number of index nodes included in each subsequence of the third sequence may not be exactly the same or may be completely different, wherein the subsequences may be expressed as (I _z , I _z+s1-1 ) ,(I _z+s1 ,I _z+s2-1 ),...,(I _z+s(L-1) ,I _z+sL-1 ), and record in a table the factors that determine the length of each subsequence Parameters (s1, s2,...,sL), but when the file hash value is sufficiently randomly distributed, the index nodes can be configured more efficiently by using the same number of subsequence index nodes as mentioned above.

4 and 5 are schematic configuration diagrams of an index node configuration method according to an embodiment of the present invention. As shown in Figures 4 and 5, the file system first configures the sequence 210 including 128 consecutive index nodes numbered from 1025 to 1152 to the directory 200 for use by the files in the directory 200, and then configures the sequence 210 including the 128 consecutive index nodes numbered from 1153 to 1408. A sequence 410 of 256 consecutive index nodes (w equals 2) is allocated to directory 200 for storing metadata of its files. Sequence 210 is divided into 16 subsequences, each subsequence includes 8 consecutive index nodes. The sequence 410 is also divided into 16 subsequences, and each subsequence includes 16 consecutive index nodes. For example, subsequence 1 includes 16 consecutive index nodes numbered from 1169 to 1184.

Take file file99 as an example. The mantissa of the hash value of file file99 is 0001, which corresponds to subsequence 1 and corresponds to block 5 in the table of block 0. Therefore, the file system configures the index node 1034 of subsequence 1 of sequence 210. Used for file file99, and the index node number 1034 of file file99 is recorded in block 5. Then, because the metadata of file file99 is too much to be stored in index node 1034, the file system allocates two index nodes 1169 and 1170 of subsequence 1 of sequence 410 to file file99 to store the extended attributes and attributes of file file99. /or access control list, and the extended attributes and/or access control list of file file99 recorded in index node 1034 are stored in index nodes 1169 and 1170.

In addition, when it is necessary to access the extended attributes and/or access control list of file file99, the index node number of block 5 recording file file99 can be obtained from the table in block 0, and the index node number of file file99 can be obtained from block 5 The number is 1034, and then it is learned from index node 1034 that the extended attributes and/or access control list of file file99 are stored in index nodes 1169 and 1170, and then the extended attributes and/or access list of file99 can be accessed at index nodes 1169 and 1170. Control list.

In addition to the above-mentioned third sequence, the file system can also configure a fourth sequence including w'k consecutive index nodes in unused index nodes for storing file metadata. In one embodiment, the fourth sequence can be expressed as (I _z' , I _z'+w'k-1 ), where I _z' and I _z'+w'k-1 are the The first and last index nodes, w' is a positive integer, z' is the number of the first index node of the fourth sequence, z'+w'k-1 is the number of the last index node of the fourth sequence number. The division and arrangement of the fourth sequence may be compared to the third sequence, just as the division and arrangement of the second sequence are to be compared to the first sequence.

In addition, when there is too much metadata for a certain file and cannot be all stored in the w index nodes configured for the file in the third sequence, the file system can configure the selected subsequence corresponding to the file in the fourth sequence. w' index nodes in the file's metadata to store more than w index nodes in the file's metadata. Part of the capacity of the boot node. If the fourth sequence is not enough to store all the metadata, the file system can configure another sequence to store additional metadata, and so on.

Figure 6 is a schematic configuration diagram of an index node configuration method according to an embodiment of the present invention. As shown in Figure 6, the file system configures the sequence 210 including 128 consecutive index nodes numbered 1025 to 1152 to the directory 200 for use by the archives of the directory 200, and divides the sequence 210 into 16 subsequences, each subsequence Both include 8 consecutive index nodes. Next, the file system configures the sequence 410 including 256 consecutive index nodes numbered 1153 to 1408 to the directory 200 to store metadata such as extended attributes and/or access control lists of the files in the directory 200, and the sequence 410 Divided into 16 subsequences, each subsequence includes 16 consecutive index nodes. Then, the index nodes of at least one subsequence of the sequence 210 have all been configured for use by files in the directory 200. Therefore, the file system further configures the sequence 610 including 128 consecutive index nodes numbered 1409 to 1536 to the directory 200 for use. The archive of directory 200 is used, and the sequence 610 is divided into 16 subsequences, each subsequence including 8 consecutive index nodes.

In this embodiment, the length of the hash value mantissa of the file is 4 bits, and 2 ⁴ is equal to 16, which is the number L of the above-mentioned subsequences. Therefore, the mantissa of the hash value of each file is equal to the hash value of the file divided by the subsequence. The remainder r obtained from quantity L. In other words, files with the same hash value mantissa all correspond to the same directory block and the same subsequence number. For example, files with a hash value ending in 0000 all correspond to directory block 10 and subsequence 0 of sequences 210, 410, and 610. Files with a hash value ending in 0001 all correspond to directory block 5 and subsequence 1 of sequences 210, 410, and 610. And so on.

As mentioned above, the files in the directory 200 are sorted according to their hash value mantissas, and files with the same hash value mantissas all correspond to the same subsequence number. Therefore, when the file system performs traversal of the directory 200, only In each sequence, pre-fetch its subsequences sequentially, for example First pre-read the sub-sequence 0 of sequences 210, 410 and 610, then pre-read the sub-sequence 1 of sequences 210, 410 and 610, and so on. The index node of each file can be quickly obtained in directory order without randomization. Read a large number of scattered index nodes.

The present invention also provides a data processing device, such as a smart phone, a computer, a server, or any other electronic device with data processing and storage functions. The data processing device is installed with a file system to execute the above index node configuration method.

The invention also provides a computer-readable medium, such as a memory, a floppy disk, a hard disk or an optical disk. The computer-readable medium is used in the data processing device and stores instructions to execute the above-mentioned index node configuration method.

To sum up, the files in the directory of the file system of the present invention are mapped to a block in the directory according to the mantissa of its hash value. Therefore, when looking up the metadata of a single file, only the file needs to be scanned. For the corresponding directory block, the index node number can be obtained without scanning the entire directory. In addition, when performing directory listing, you only need to pre-read the sub-sequences in each sequence in order, and you can quickly obtain the index nodes of each file in the directory order, without the need to randomly read a large number of scattered index nodes. Therefore, the present invention can quickly execute the query of a single file and the detailed listing of the entire directory, and increase the processing speed.

The above embodiments are only illustrative to illustrate the principles and effects of the present invention, but are not intended to limit the present invention. Anyone with ordinary knowledge in this technical field can modify and change the above embodiments without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be as listed in the patent application scope described below.

S100~S124: Method steps

Claims (10)

An index node configuration method for a file system, including the following steps: when the directory sequence mode is activated, the directory system in the file system is configured to include a first sequence of consecutive plurality of index nodes, wherein the plurality of files in the directory Sort according to the hash value of the files; divide the first sequence into plural first sub-sequences; when a new file is created in the directory, select one of the first sub-sequences according to the hash value of the new file Or, to configure the first index node of one of the selected first sub-sequences for use by the new file; when all the index nodes of any one of the first sub-sequences have been configured for use When, the directory is configured to include a second sequence of consecutive plural index nodes; the second sequence is divided into a plurality of second sub-sequences; and when the number of index nodes required for any file in the directory exceeds the number of index nodes for the file in the When one of the selected first sub-sequences corresponding to the first sequence can be provided, one of the second sub-sequences of the second sequence is selected based on the hash value of the file, so as to Configuring a second index node of one of the selected second sub-sequences of the second sequence for use by files in the directory. The index node configuration method described in request item 1 further includes the following steps: when the total number of index nodes used by the files in the directory is greater than the default value, activate the directory sequential mode to configure the directory with the first sequence. The index node configuration method as described in request item 1, wherein the hash value of the files is the hash value of the names of the files. The index node configuration method as described in claim 1, wherein each of the first sub-sequences includes the same number of index nodes. The index node configuration method as described in claim 1, wherein each of the first sub-sequences includes a different number of index nodes. The index node configuration method as described in request item 1, wherein the step of selecting one of the first subsequences based on the hash value of the new file includes: dividing the hash value of the new file by the The remainder obtained from the number of first subsequences is used to select one of the first subsequences to configure the first index node of the selected one of the first subsequences. An index node configuration method for a file system, including the following steps: when the directory sequence mode is activated, the directory system in the file system is configured to include a first sequence of consecutive plurality of index nodes, wherein the plurality of files in the directory Sort according to the hash value of the files; divide the first sequence into plural first sub-sequences; when a new file is created in the directory, select one of the first sub-sequences according to the hash value of the new file Or, to configure the first index node of one of the selected first sub-sequences for use by the new file; the directory is configured to include a third sequence of consecutive plural index nodes; the third sequence Dividing into a plurality of third subsequences; and selecting one of the third subsequences of the third sequence based on the hash value of the new file to configure the selected third subsequences of the third sequence The third index node of one of them is used to store the metadata of the new file. The index node configuration method as described in request 7, wherein the metadata of the new file includes extended attributes and/or access control lists of the new file. A data processing device installed with a file system to execute the index node configuration method described in any one of claims 1-8. A computer-readable medium, used in a data processing device, stores instructions to execute the index node configuration method described in any one of claims 1-8.

Images