JP6343952B2 - Storage system - Google Patents

Description

  The present invention relates to a storage system, and more particularly to a storage system that distributes data and stores it in a plurality of storage devices.

  In recent years, with the development and spread of computers, various types of information have been converted into digital data. As a device for storing such digital data, there are storage devices such as a magnetic tape and a magnetic disk. Since the data to be stored increases day by day and becomes enormous, a large-capacity storage system is required. In addition, reliability is required while reducing the cost of the storage device. In addition to this, it is necessary that data can be easily retrieved later. As a result, there is a demand for a storage system that can automatically increase storage capacity and performance, eliminate duplicate storage, reduce storage costs, and have high redundancy.

  In response to such a situation, in recent years, a content address storage system has been developed as shown in Patent Document 1. In this content address storage system, data is distributed and stored in a plurality of storage devices, and the storage location where the data is stored is specified by a unique content address specified according to the content of the data.

  As described above, since the content address is generated so as to be unique according to the content of the data, if it is duplicated data, the data of the same content can be acquired by referring to the data at the same storage position. . Therefore, it is not necessary to store the duplicate data separately, and duplicate recording can be eliminated and the data capacity can be reduced. That is, the content address storage system has a deduplication function in which new data is stored only when data of the same content is not stored.

  Further, in the storage system, a predetermined amount of block data is divided into a plurality of fragment data, and a fragment that becomes redundant data is further added, and the plurality of fragment data are respectively stored in a plurality of storage devices. Then, by designating the content address later, the fragment data stored in the storage location specified by the content address can be read, and the block data before division can be restored from the plurality of fragment data.

  As described above, since the storage system adds fragment data that becomes redundant data, even when fragment data equal to or less than the number of fragments of the added redundant data is lost, the original block data can be regenerated. .

  Furthermore, when storing fragment data in multiple storage devices, the storage system has a distributed arrangement function so that a large amount of fragment data is not lost at one time and the load is not concentrated on a predetermined storage device. ing. Here, each storage device has a data storage container, and fragment data is stored in each data storage container to distribute the load.

  Here, in the storage system as described above, it is determined whether or not the same data is already stored in the storage device at the time of writing. It's not easy. Therefore, in order to improve the efficiency of data duplication determination, a method is used in which fingerprints (hash values) calculated based on data contents are compared rather than comparing actual data. Since the hash value has a data size smaller than that of the actual data, it is possible to speed up duplication determination. Similarly, at the time of reading data, it is possible to speed up the search by comparing the content address of the target data with the already stored hash value.

  On the other hand, since the hash value described above is calculated in units of block data obtained by dividing (chunking) the write data into a certain size, the number increases as the data amount increases. In many storage systems, writing and reading performance is regarded as important, and thus a method for efficiently searching for the same hash value is required to speed up duplication determination.

  Therefore, the storage system uses a short hash that is a part of the hash value of the block data as a key for duplication determination, and a hash table that is a table that summarizes information for accessing data corresponding to the hash value. Have. When a hash table is used, search is performed based on a short hash, so that search performance can be improved as compared with a case where full hash values are compared. In addition to the determination of duplication, a hash table is used when searching for the corresponding block data requested to be read not only when data is read.

JP2013-182476

  Here, since the storage system is a system used exclusively for data storage, first, the reliability of stored data (data block and its full hash value) becomes important. For this reason, the stored data is stored in a file system having a journal function with an emphasis on reliability. Therefore, after a failure occurs, the failed data can be recovered from the journal information.

  On the other hand, the hash table is less important data because it is intended for high-speed data retrieval as described above and can be generated again from stored data. It has a configuration that emphasizes. That is, unlike the above-described stored data, it does not have a journal function. Therefore, if an abnormal restart such as a power failure occurs, the data stored in the hash table cannot be consistent with the data stored in the hash table. Therefore, the process of building the hash table again must be performed.

  However, since the hash table is information that is first accessed when searching for stored data at the time of reading, there arises a problem that access to the stored data becomes impossible when the data is damaged or during reconstruction. This restructuring process has become longer due to the recent enlargement of data, and there is a problem that the time during which stored data cannot be accessed is prolonged. As a result, there arises a problem that the performance of the storage system is degraded.

  Therefore, an object of the present invention is to solve the above-described problem that the performance of the storage system is degraded.

A storage system according to an aspect of the present invention
A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Distributed storage processing means;
Data for searching a plurality of fragment data constituting the search request data stored in the plurality of storage means by searching the summary table and the index data based on the summary data corresponding to the search request data A search means,
The distributed storage processing means stores the index data including write status information indicating a status at the time of writing the fragment data referred to by the index data,
The data search means stores the fragment data further referenced in the index data referenced by the unavailable summary table when at least a part of the summary table is unavailable. Searching the index data corresponding to a storage means different from the storage means, and writing the fragment data constituting the search target data based on the summary data of the search target data and the writing of the fragment data The other fragment constituting the search target data by specifying the status information, searching the index data corresponding to the specific storage means based on the specified write status information and the summary data of the search target data Identify the data,
The configuration is as follows.

The program which is the other form of this invention is:
In the storage system control unit,
A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Distributed storage processing means;
Data for searching a plurality of fragment data constituting the search request data stored in the plurality of storage means by searching the summary table and the index data based on the summary data corresponding to the search request data Search means;
And realize
The distributed storage processing means stores the index data including write status information indicating a status at the time of writing the fragment data referred to by the index data,
The data search means stores the fragment data further referenced in the index data referenced by the unavailable summary table when at least a part of the summary table is unavailable. Searching the index data corresponding to a storage means different from the storage means, and writing the fragment data constituting the search target data based on the summary data of the search target data and the writing of the fragment data The other fragment constituting the search target data by specifying the status information, searching the index data corresponding to the specific storage means based on the specified write status information and the summary data of the search target data Identify the data,
Make it happen,
The configuration is as follows.

A data processing method according to another aspect of the present invention includes:
A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Perform distributed storage processing
Data for searching a plurality of fragment data constituting the search request data stored in the plurality of storage means by searching the summary table and the index data based on the summary data corresponding to the search request data A data processing method for performing search processing,
During the distributed storage process, the index data is stored including write status information indicating a status at the time of writing the fragment data referred to by the index data,
A specification for storing the fragment data further referred to in the index data referred to by the unavailable summary table when at least a part of the summary table is unavailable during the data search process The index data corresponding to the storage means different from the storage means is searched, and a part of the fragment data and the fragment data constituting the search target data based on the summary data of the search target data The write status information is specified, and the index data corresponding to the specific storage means is searched based on the specified write status information and the summary data of the search target data. Identifying the fragment data;
The configuration is as follows.

  By configuring as described above, the present invention can suppress a decrease in performance of the storage system.

1 is a block diagram showing a configuration of an entire system including a storage system in Embodiment 1 of the present invention. FIG. 2 is a block diagram illustrating an overall configuration of the storage system disclosed in FIG. 1. FIG. 3 is a functional block diagram illustrating a configuration of a storage system disclosed in FIG. 2. FIG. 4 is an explanatory diagram for explaining a state when data is written to the storage system disclosed in FIG. 3; FIG. 4 is an explanatory diagram for explaining a state when data is written to the storage system disclosed in FIG. 3; FIG. 4 is an explanatory diagram for explaining a state of data written in the storage system disclosed in FIG. 3. FIG. 4 is an explanatory diagram for explaining a state of data written in the storage system disclosed in FIG. 3. FIG. 4 is an explanatory diagram for explaining a state when searching for data written in the storage system disclosed in FIG. 3; 4 is a flowchart showing the operation of the storage system disclosed in FIG. 3. 4 is a flowchart showing the operation of the storage system disclosed in FIG. 3. 4 is a flowchart showing the operation of the storage system disclosed in FIG. 3. It is a figure which shows the structure of the storage system in attachment 1 of this invention.

<Embodiment 1>
A first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a block diagram showing the configuration of the entire system. FIG. 2 is a block diagram showing an outline of the configuration of the storage system, and FIG. 3 is a functional block diagram showing the configuration of the storage system. 4 to 8 are explanatory diagrams for explaining data written in the storage system and a state at the time of retrieval. 9 to 11 are flowcharts showing the operation of the storage system.

[Constitution]
This embodiment shows a specific example of a storage system or the like described in an appendix to be described later. In the following, a case where the storage system is configured by connecting a plurality of server computers will be described. However, the storage system according to the present invention is not limited to being configured by a plurality of computers, and may be configured by a single computer.

  As shown in FIG. 1, the storage system 1 in the present invention is connected to a backup server 4 that controls backup processing via a network N. Then, the backup server 4 acquires backup target data (storage target data) stored in the backup target device 5 connected via the network N and requests the storage system 1 to store it. Thereby, the storage system 1 stores the backup target data requested to be stored for backup.

  Here, the storage system 1 according to the present embodiment divides and redundantly stores the data, distributes and stores the data in a plurality of storage devices, and stores the data according to the content address set according to the content of the stored data. This is a content address storage system for specifying the storage location. This content address storage system will be described in detail later.

  As shown in FIG. 2, the storage system 1 in this embodiment has a configuration in which a plurality of server computers are connected. Specifically, the storage system 1 includes an access node 2 that is a server computer that controls storage and reproduction operations in the storage system 1 itself, and a storage node 3 that is a server computer including a storage device that stores data. Yes.

  The access node 2 transmits / receives data to / from the backup server 4 and provides a file system service to the backup server 4. The storage node 3 actually stores the backup target data in the disk. In FIG. 2, two access nodes 2 and four storage nodes 3 are shown, but the number of access nodes 2 and the number of storage nodes 3 are not limited to those shown in FIG.

  FIG. 3 shows the configuration of the storage system 1 in this embodiment. As shown in this figure, the access node 2 constituting the storage system 1 has an I / O processing unit 21, a data transmission / reception unit 22, and a duplication determination unit 23 that are constructed by incorporating a program in the equipped arithmetic device. It is equipped with. In addition, the storage node 3 includes a fragment processing unit 31, a fragment search unit 32, a time stamp acquisition unit 33, and a node state monitoring unit 34, which are constructed by incorporating a program into the equipped arithmetic device. Then, the storage node 3 stores the hash table 36, the index file 37, and the storage file 38 in the storage device 35 equipped.

  Each unit described above may be provided in either the access node 2 or the storage node 3. That is, each unit described above may be configured in a control device equipped in the storage system. Hereinafter, each configuration will be described in detail.

  The I / O processing unit 21 of the access node 2 has a function of transmitting and receiving backup data with the backup server 4 via the internal network.

  The data transmission / reception unit 22 (distributed storage processing means) of the access node 2 uses the content address to write data requested to be written from the backup server 4 in cooperation with the duplication determination unit 23 and each unit 31 of the storage node 3. In addition, the storage node 3 has a function of storing data while performing deduplication.

  Specifically, as shown in FIG. 4, the data transmitting / receiving unit 22 first divides the file A requested to be written from the backup server 4 into block data D (data to be written) having a predetermined capacity (for example, 64 KB). Subsequently, based on the data content of the divided block data D, a hash value H that is summary data based on the data content is calculated. For example, the hash value H is calculated from the data content of the block data D using a preset hash function. In the present embodiment, the hash value H is cited as the summary data, but the summary data may be calculated by any method based on the contents of the block data.

  Here, the duplication determination unit 23 checks whether the block data D is already stored by using the hash value H of the block data D of the file A. The overlap determination process will be described later.

  When the data transmission / reception unit 22 determines that the block data D is already stored in the storage node 3 as a result of the duplication determination by the duplication determination unit 23, the storage position of the block data D that is already stored is determined. Is used as the content address of the block data requested to be written. That is, in the case of duplication, the storage position of the block data D that is already stored is referred to without actually writing, and the writing is completed.

  On the other hand, if the data transmission / reception unit 22 determines that the block data D related to the write request is not duplicated and is not yet stored as a result of the duplication judgment by the duplication judgment unit 23, Write block data D. The data transmitting / receiving unit 22 divides the block data D into a plurality of fragment data having a predetermined capacity. For example, as indicated by reference numerals D1 to D9 in FIG. 4, the block data D is divided into nine fragment data (divided data F1). Further, even if some of the divided fragment data is missing, redundant data is generated so that the original block data D can be restored and added to the divided fragment data F1. For example, three pieces of fragment data (redundant data F2) are added as indicated by reference numerals D10 to D12 in FIG. As a result, a data set composed of 12 fragment data composed of 9 divided data F1 and 3 redundant data F2 is generated. The number of fragment data that is redundant data can be changed by setting.

  Then, the data transmitting / receiving unit 22 transmits the data set composed of the plurality of generated fragment data so as to be distributed and stored in the storage nodes 3.

  The fragment processing unit 31 (distributed storage processing means) of the storage node 3 stores the fragment data sent from the access node 2 on the disk. At this time, the fragment processing unit 31 stores the data in a plurality of data storage containers provided in the virtual nodes V01, V02,... Formed across the plurality of storage nodes 3.

  Here, the virtual nodes V01, V02... Will be described with reference to FIG. In the example of FIG. 5, the virtual node V01 is formed by twelve data storage containers (reference numerals 01 to 12) serving as a data storage area formed in each storage node 3 across a plurality of storage nodes 3. ing. Here, it is assumed that four virtual nodes V01, V02, V10, and V11 are formed.

  Then, which virtual node stores the 12 fragment data generated from the block data D is determined by confirming a predetermined number of bits from the head of the hash value of the block data D. In the example of FIG. 5, since there are four virtual nodes, the storage location is determined based on the value of the first 2 bits of the hash value of the block data. For example, when the hash value is “00001111...”, It is determined to store the first two bits “00” in the virtual node “V00”. At this time, as shown in FIG. 4, one piece of fragment data is stored in each of the 12 data storage containers (01 to 12).

  The fragment processing unit 31 manages the data as follows when distributing and storing the fragment data as described above. First, as shown in FIG. 6, the data storage containers V00 and V01 have a storage file 38 for storing fragment data, and an index file 37 for storing metadata such as a full hash value related to the fragment data. The fragment processing unit 31 additionally stores the sent fragment data in the storage file 38. The storage file 38 is created in ascending order from 1 as the file ID, and when a certain number of fragment data is stored, the value of the file ID is increased and written to the next storage file. Specifically, as shown in FIG. 7C, fragment data itself is stored in the storage file 38 at each offset position representing a predetermined storage position in the storage file.

  Further, when the fragment processing unit 31 finishes writing the fragment data in the storage file 38, the fragment processing unit 31 subsequently stores the index data, which is metadata for each fragment data, as an entry in the index file. FIG. 7B shows an example of an index file. In this way, the index file 37 stores index data that is metadata corresponding to fragment data stored in the storage file 38 for each offset that is the storage position of the index file. Each index data is written with the full hash value of the block data D that is the source of the fragment data stored in the storage file 38, the storage file ID indicating the pointer (reference information) to the fragment data, its offset, and the fragment data. Includes a time stamp (writing time information) based on the recorded time.

  The index file 37 is stored in the storage node 3 for each storage node 3 in which fragment data referred to by each index data included in the index file 37 is stored. That is, when fragment data is stored in a certain storage node 3, the index file 37 stores the index data corresponding to the fragment data in the index file 37 formed in the same storage node 3. At this time, the index data is added to the storage area of the index file 37 in order.

  Further, the fragment processing unit 31 associates an entry for each fragment data stored in the storage file 38 with the hash table 36 (summary table) and associates the information referring to the index data corresponding to the fragment data with the short hash value. Store. FIG. 7A shows an example of the hash table 36. Each entry of the hash table 36 includes an index file that represents a short hash value of the block data D that is the source of the fragment data stored in the storage file 38 and a pointer (reference information) to index data that refers to the fragment data. ID and its offset are stored. The short hash (partial summary data) is a part of the full hash value (for example, 20 bytes) of the block data D that is the source of the fragment data (for example, 8 bytes from the beginning).

  The hash table 36 is stored in the storage node 3 for each storage node 3 in which the index file referred to in the hash table 36, that is, fragment data corresponding to each index data is stored. That is, when fragment data is stored in a certain storage node 3, an entry of the reference information and the short hash value of the index file including the index data that refers to the fragment data is created in the same storage node 3. Store in table 36. At this time, each entry is added to the storage area of the hash table 36 in order.

  Then, the fragment processing unit 31 returns the hash value of the block data D to the data transmitting / receiving unit 22 as the storage position of the block data D. Then, the data transmitting / receiving unit 22 manages the identification information of the block data D in association with the hash value. Thereby, as will be described later, when there is a read request for the block data D, the storage position can be retrieved from the hash value.

  Here, an instruction to write the plurality of fragment data generated from the block data D to the fragment processing unit 31 from the data transmitting / receiving unit 22 is performed at the same timing. For this reason, writing in each fragment processing unit 31 of each storage node 3 is performed at almost the same timing. Then, the writing of the plurality of fragment data generated from one block data D to each storage node 3 is performed at substantially the same time. As a result, the time stamp value included in each index data referring to the fragment data is also the same or approximate value. In addition, since each index data is stored in the index file 37 in the storage area of the index file 37 in order, the index data is stored in the order of writing, that is, in the order of the time stamp. It becomes.

  In addition, the “full hash value” included in each index data stored in the index file 37 is a hash value of the block data D before being divided, which is a source of fragment data referred to by each index data. . Therefore, the full hash value included in each index data of the 12 fragment data is the same in each index file 37 of all storage nodes 3 storing each fragment data. By using this property, as will be described later, when an index file of any one storage node 3 storing fragment data generated from one block data D can be accessed, the time stamp in which the fragment data is written Can be obtained. The remaining fragment data can be easily searched based on the time stamp.

  By storing the block data D as described above, the data transmitting / receiving unit 22 and the fragment processing unit 31 (distributed storage processing unit, data search unit) determine whether or not there is a data read request from the backup server 4 Sometimes, it is possible to access the hash value or fragment data that is the content address of the block data D stored as shown in FIG.

  First, the hash value of the block data D to be searched is confirmed, the hash table 36 stored in the local node 3 is searched from the short hash value, and the target index file ID and offset are obtained (arrow) Y1). Subsequently, the index file 37 is accessed from the acquired information to search for index data including the same full hash value as the full hash value of the data to be searched (see arrow Y2). Thereby, it is possible to perform duplication determination such as whether or not the same full hash value as the block data D to be searched exists. Furthermore, the storage file 38 can be accessed from the index data found by the search, and the fragment data to be searched can be read (see arrow Y3). Then, by combining the fragment data read from each storage node 3, the block data D to be searched is generated, and by returning the block data D, the data can be read.

  If the hash table 36 of the storage node 3 that is the storage node 3 cannot be used, the storage node 3 is stored as follows by the fragment search unit 32 and the time stamp acquisition unit 33 (data search unit). The fragment data constituting the block data D is accessed.

  The fragment search unit 32 uses the time stamp acquisition unit 33 to acquire the time stamp of the fragment data constituting a part of the block data D to be searched that is stored in another storage node 3 that is not itself. The fragment search unit 32 then writes the fragment data written at a close time from the time stamp in the index file 37 stored in the storage node 3 (specific storage node) that is the self based on the acquired time stamp. Get a list of

  In response to the request from the fragment search unit 32, the time stamp acquisition unit 33 remotely accesses the hash table 36 of another storage node 3 that stores the fragment data of the block data D to be searched. Then, referring to the hash table 36 of another storage node 3, the index data in the index file 37 stored in another storage node 3 is specified based on the hash value of the block data D to be searched. . Thereby, the time stamp in which the fragment data corresponding to the specified index data is written is acquired. Then, the acquired time stamp is returned to the fragment search unit 32.

  The node state monitoring unit 34 has a function of acquiring a list of storage nodes 3 that are operating normally. The normal storage node 3 is a node that has a function as a storage service and can access data such as the hash table 36.

  The functions of the fragment search unit 32, time stamp acquisition unit 33, and node state monitoring unit 34 described above will be described in detail in the following description of the operation.

[Operation]
Next, operations of the storage system having the above-described configuration, in particular, data write processing, data read processing, and data read processing during hash table reconstruction will be described with reference mainly to the flowcharts of FIGS. .

(Write process)
First, the data writing process will be described with reference to FIG. When the access node 2 receives a write request from the backup server 4 (step S1), the stream data to be written is divided into block data D of a certain size (step S2). For this block data D, a hash value is calculated and acquired using a hash function. Duplication determination is performed with the obtained hash value as an input (step S3). Here, the overlap determination process will be described with reference to FIG.

  Based on the hash value of the block data D calculated as described above, the hash table 36 of an arbitrary storage node 3 is accessed. Then, a short hash value of the hash value is searched from the hash table 36, and it is confirmed whether there is an entry (index data) having the same short hash value (see arrow Y1). At this time, if no entry is found, data having the same content as the block data D to be stored is not written to the storage node 3 and becomes “non-overlapping”. If an entry is found, the index file ID and offset corresponding to the short hash value are acquired. Since the hash table 36 compares only with the short hash value, a plurality of entries may be found.

  Subsequently, the index file 37 is accessed from the entry information of the found hash table 36, and the full hash value in the index file 37 is compared with the calculated full hash value of the block data D (arrow Y2). Here, when an entry having the same full hash value is found, it can be determined that data having the same content as the block data D to be written has already been written in the storage node 3, resulting in “duplication”. On the other hand, if the same entry is not found, it becomes “non-overlapping”.

  In the duplication determination described above, when “duplication” is obtained (step S3: Yes), writing completion is returned (step S8). In the case of “non-overlapping” (step S3: No), the block data D is divided into fragment data and distributed to each storage node 3 (step S4). Thereafter, the processing surrounded by the broken line A is performed in each storage node 3, and fragment data is stored.

  The storage node 3 stores the received fragment data in each data storage container as shown in FIG. There are twelve data storage containers in each virtual node (such as V00 in FIG. 5) (reference numeral 01-12). Here, the first two bits of the received hash value are confirmed, and it is determined to which virtual node data is allocated. For example, when the hash value is “1010101111...”, The virtual node V10 stores the 12 pieces of fragment data. Each data storage container stores one piece of fragment data.

  As shown in FIG. 6, the fragment data is stored in the storage file 38 of each data storage container as shown in FIG. 6 (step S5). The storage file 38 is created in ascending order from 1 as the file ID, and when a certain number of fragment data is accumulated, it is written to the next file.

  After the fragment data is stored in the storage file 38, the index data of the fragment data is added to the index file 37 as each entry. At this time, as index data, reference information (storage file ID, offset) that is a pointer to fragment data, a full hash value of block data D that is the source of fragment data, and a time stamp at which fragment data is written are shown in FIG. Store as (B) (step S6).

  Finally, an entry that refers to the index data corresponding to the written fragment data is written in the hash table 36. Specifically, as each entry, the short hash value of the block data D that is the source of the fragment data and the pointer (index file ID, offset) to the index data that refers to the fragment data are as shown in FIG. (Step S7). Then, the stored block data D and the hash value are managed in association with each other, and the writing is completed (step S8).

(Reading process)
Next, the data reading process will be described with reference to FIG. When the access node 2 receives a read request from the backup server 4 (step S11), the hash value of the block data D to be read is confirmed from the managed information (step S12). Then, the access node 2 distributes the hash value of the block data D to be read to all the storage nodes 3 (step S13).

  Subsequently, each storage node 3 performs a process surrounded by a broken line B to read fragment data. First, the hash table 36 as shown in FIG. 7C is searched using the short hash value of the received hash value (see arrow Y1 in FIG. 8), and the target index file ID and offset are obtained. (Step S14). Then, the storage node 3 accesses the index file 37 as shown in FIG. 7B from the acquired information and searches for index data that is an entry that is the same as the full hash value of the target data ( Step S15) (see arrow Y2 in FIG. 8). A storage file 38 as shown in FIG. 7A is accessed from the index data that is an entry of the index file 37 found by the search (see arrow Y3 in FIG. 8), and the target fragment data is read (step S16). .

  Thereafter, the storage node 3 or the access node 2 combines the fragment data read from each storage node 3, generates block data D to be read (step 17), and returns the block data D (step 17). S18).

(Read processing during hash table reconstruction)
Next, with reference to FIG. 11, data reading processing during hash table reconstruction will be described. When the access node 2 receives a read request from the backup server 4 (step S21), the hash value of the block data D to be read is confirmed from the managed information (step S22). Then, the access node 2 distributes the hash value of the block data D to be read to all the storage nodes 3 (step S23).

  Subsequently, each storage node 3 performs a process surrounded by a broken line C to read fragment data. Here, the storage node 3 checks whether or not the hash table 36 stored in its own storage node 3 is normal (step S24), and if the hash table 36 is normal (step S24: No). A normal reading process as shown in FIG. 10 is executed (step S25). On the other hand, when the hash table 36 stored in the storage node 3 that is the storage node 3 is being reconstructed (step S24: Yes), the storage node 3 reads the fragment data to be read by the following method. In the following description, the storage node 3 whose hash table 36 is being reconstructed is referred to as a specific storage node 3, and the other storage nodes 3 are referred to as other storage nodes 3.

  First, the specific storage node 3 acquires a list of normal storage nodes 3 using the node state monitoring unit 34 (step S26). A normal node is a node that operates as a storage service and can access data such as the hash table 36. A storage node 3 in which fragment data generated from the same block data D is stored is searched from the list of normal nodes.

  Specifically, a search for another storage node 3 in which the target fragment data is stored by a specific storage node 3 can use a fragment data distribution rule at the time of data storage. Here, the virtual node in which the fragment data is stored is determined by using the first two bits of the hash value of the block data D, and each virtual node has the position information of the data storage container in which the fragment data is stored. Yes. For this reason, the specific storage node 3 specifies another storage node 3 having the fragment data from the hash value of the block data D composed of the target fragment data.

  Subsequently, the specific storage node 3 accesses the hash table 36 stored in another storage node 3 having the target fragment data using the short hash value of the block data D as a key. Then, the specific storage node 3 is the metadata of the target fragment data from the hash table 36 of another storage node 3, and information for accessing the index data stored in the other storage node 3 ( Index file ID and offset) are acquired (step S27). Then, the specific storage node 3 specifies the index data of the target fragment data from the index data acquired from another storage node 3 using the full hash value of the block data D as a key. As a result, the specific storage node 3 obtains a time stamp that is stored in another storage node 3 and represents the time when the fragment data to be read is written (step S28).

  Subsequently, the specific storage node 3 refers to the index file 37 stored in the specific storage node 3 that is the self, based on the time stamp acquired from another storage node 3, and determines the value of the time stamp. The index data that is an entry close to is searched (step S29). At this time, the index data searched from the index file 37 is index data having a time stamp in the range of about 1 minute before and after the time stamp acquired from another storage node 3. Here, in the case of storage without a rule, it is inefficient to search all entries based on the time stamp. However, in the case of the storage system in this embodiment, the index file has a write-once data structure as a matter of fact. ing. For this reason, since the index data in the index file are arranged in the order of the time stamps, it is easy to search for index data having similar time stamp values.

  Subsequently, the specific storage node 3 sequentially compares the full hash values of the entries in the index file searched from the time stamp, and searches for index data having the same hash value (step S30). Here, if the target entry is not found, it is returned to the upper level as a read error. If the target entry is found, fragment data is read from the storage file 38 referenced by the index data that is the entry (step S31). Thereafter, the specific storage node 3 or access node 2 combines the fragment data read from each storage node 3, generates block data D to be read (step S32), and returns the block data D (Step S33).

  As described above, in the storage system according to the present invention, even when the hash table provided for high-speed search becomes a failure and recovery processing is in progress, it is possible to suppress the decrease in reading performance and continue the operation. Can do. The reason will be described below.

  First, in the storage system of the present invention, the storage file for storing fragment data and the index file for storing fragment data metadata have substantially a write-once data structure in all storage nodes. For this reason, the data in each file is arranged in the order of the time stamp. Therefore, if the stored data information cannot be obtained from the hash table when the hash table is being reconstructed, the fragment data is stored in which file as long as the time (time stamp) at which the fragment data is written is known. It can be easily guessed.

  In the present invention, since each fragment data generated from the target block data is immediately distributed and stored in each node, it is highly likely that these data will be written in a close time zone, and all fragment data Also has a similar time stamp. Therefore, by acquiring a time stamp in which other fragment data is written, it is possible to estimate the time in which the fragment data of the own node is written.

  During reconstruction of the hash table, the hash table is not used, and only the time stamps in which other fragment data are written are used to narrow down from the own node. At this time, as described above, since the metadata and the like are arranged in the order of the time stamp, the data can be narrowed down with extremely high accuracy, and the storage location of the fragment data can be searched at high speed.

  Here, the case where the hash table 36 of another storage node 3 is first searched and the time stamp of the fragment data is acquired has been exemplified above, but the present invention is not limited to acquiring the time stamp by such a method. For example, the index data stored in another storage node 3 is searched from the full hash value of the block data D to be read, the index data including the full hash value is acquired, and the time stamp is specified. Also good.

  Further, although the index data described above includes a time stamp indicating the time when the fragment data to be referenced is stored, other data may be stored instead of the time stamp. For example, timing information including a value indicating that fragment data generated from the same block data D is written at the same timing may be stored in the index data instead of the time stamp. At this time, the timing information is information that can be distinguished from other timings. Note that the information to be replaced with the time stamp may be information (write status information) that can distinguish the status at the time of writing fragment data generated from the same block data D from the status at the time of writing other block data. Any information may be used.

  Moreover, although the case where the index file 37 and the hash table 36 are stored in the storage node in which the fragment data to be referenced is stored has been described above, the present invention is not limited to such a configuration. For example, the index file 37 and the hash table 36 may be stored in any storage device. However, each index file and hash table 36 may be stored separately for each storage node in which the fragment data to be referenced is stored.

<Appendix>
Part or all of the above-described embodiment can be described as in the following supplementary notes. The outline of the configuration of the storage system 100 (see FIG. 12), the program, and the data processing method in the present invention will be described below. However, the present invention is not limited to the following configuration.

(Appendix 1)
A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in the plurality of storage units 120;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Distributed storage processing means 111;
Data for searching a plurality of fragment data constituting the search request data stored in the plurality of storage means by searching the summary table and the index data based on the summary data corresponding to the search request data Search means 112,
The distributed storage processing unit 111 stores the index data including write status information indicating a status at the time of writing the fragment data referred to by the index data,
The data search means 112 stores the fragment data further referenced in the index data referenced by the unavailable summary table when at least a part of the summary table is unavailable. Searching the index data corresponding to a storage means different from the storage means, and a part of the fragment data constituting the search target data based on the summary data of the search target data and the fragment data The write status information is specified, and the index data corresponding to the specific storage means is searched based on the specified write status information and the summary data of the search target data, and the other search data constituting the search target data Identify fragment data,
Storage system 100.

  According to the storage system of the present invention, first, when writing data, a plurality of fragment data is generated from the storage target data, and these are distributed and stored in a plurality of storage means. At this time, the index data in which the information for referring to the fragment data, the writing status information of the fragment data, and the summary data of the storage target data are associated with each other are stored for each storage unit. Furthermore, a summary table in which information that refers to the index data and partial summary data that is a part of the summary data of the storage target data is associated is also stored.

  In addition, the storage system searches whether the target data (search target data) is already stored when reading the data. At this time, first, index data is specified by searching a summary table using partial summary data that is a part of summary data of search target data. Then, the specified index data is searched using the summary data of the search target data to check whether the search target data exists. Thereby, the read processing is performed by reading out the fragment data specifying the storage position and regenerating the block data.

  When a part or all of the summary table is unavailable, the storage system searches for data to be read as follows. First, from the index data corresponding to another storage means different from the specific storage means corresponding to the unavailable summary table, the index data associated with the one that matches the summary data of the search target data is searched, A part of fragment data of the search target data and its writing status information are specified. Then, using this specified write status information, the index data associated with the summary data of the search target data is searched from the index data corresponding to the specific storage means, and the remaining fragment data Is identified. As described above, by using the writing status information of a part of fragment data of the search target data, the remaining fragment data can be easily and quickly identified. As a result, even when at least a part of the summary table cannot be used, data can be searched easily and quickly, and the performance of the storage system can be prevented from deteriorating.

(Appendix 2)
The storage system according to attachment 1, wherein
The distributed storage processing unit stores the summary table separately for each storage unit serving as a storage destination of the fragment data further referred to by the index data referred to by the summary table,
The data search means searches the available summary table corresponding to the other storage means based on partial summary data which is a part of the summary data of the search target data, and the other storage means The index data corresponding to the search target data is searched, the fragment data constituting the search target data and the write status information of the fragment data are specified, and the specified write status information and the summary data of the search target data are specified. The index data corresponding to the specific storage means is searched based on and the other fragment data constituting the search target data is specified.
Storage system.

(Appendix 3)
The storage system according to appendix 2,
The distributed storage processing unit stores the index data in the storage unit serving as a storage destination of the fragment data referred to by the index data, and the summary table is referred to by the index table. Storing it in the storage means as a storage destination of the fragment data further referred to by data;
The data search means, when the summary table stored in the specific storage means is not available, based on partial summary data that is a part of the summary data of the search target data, Searching the available summary table stored in the means, searching the index data stored in the other storage means, and a part of the fragment data and the fragment constituting the search target data The write status information of data is specified, and based on the specified write status information and the summary data of search target data, the index data stored in the specific storage means in which the summary table is unavailable Searching to identify other fragment data constituting the search object data;
Storage system.

  According to the storage system having the above configuration, when a part of the summary table is unavailable, first, an index associated with the one that matches the summary data of the search target data using only the available summary table. The data is searched to identify a part of the fragment data of the search target data and its write status information. Then, using the specified writing status information, the remaining fragment data is specified also from the index data related to the unusable summary table. As a result, even when a part of the summary table cannot be used, data can be easily and quickly searched, and degradation of the performance of the storage system can be suppressed.

(Appendix 4)
The storage system according to any one of appendices 1 to 3,
The distributed storage processing unit stores the plurality of fragment data generated from one storage target data in a distributed manner in the plurality of storage units at the same timing, and the index data includes the index data. Write timing information representing a value based on the write timing of the fragment data referred to is included and stored as the write status information,
The data search means includes the write timing information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target data Based on the summary data, the index data stored in the specific storage means is searched, and the other fragment data constituting the search target data is specified.
Storage system.

(Appendix 5)
The storage system according to any one of appendices 1 to 4,
The distributed storage processing means stores, in the index data, write time information representing a value based on a write time of the fragment data referred to by the index data, including the write status information,
The data search means includes the write time information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target data Based on the summary data, the index data stored in the specific storage means is searched, and the other fragment data constituting the search target data is specified.
Storage system.

(Appendix 6)
The storage system according to appendix 5,
The data search means searches the index data corresponding to the specific storage means, from the index data associated with the write time information within a predetermined time before and after the specified write time information, Specifying the other fragment data constituting the search target data;
Storage system.

(Appendix 7)
The storage system according to any one of appendices 4 to 6,
The distributed storage processing means additionally writes in order to the storage area in each storage means when the plurality of fragment data is distributed and stored in the plurality of storage means, respectively.
Storage system.

  According to the storage system configured as described above, first, during distributed storage, information indicating the timing and time of writing fragment data is stored in the index data. Then, the storage system first searches the index data corresponding to the available summary table, and specifies part of the fragment data of the search target data and the write timing and time information that is the write status information. To do. The remaining fragment data is also specified from the index data associated with the unusable summary table from the specified writing timing and time information. At this time, it is possible to identify fragment data associated with information having similar writing timing and time. If the fragment data is stored in the storage area in each storage means in the order of writing, it becomes easier to search for the fragment data based on the write timing and time. As a result, even if a part or all of the summary table cannot be used, data can be easily and quickly searched, and the performance of the storage system can be prevented from deteriorating.

(Appendix 8)
In the storage system control unit,
A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Distributed storage processing means;
Data for searching a plurality of fragment data constituting the search request data stored in the plurality of storage means by searching the summary table and the index data based on the summary data corresponding to the search request data Search means;
And realize
The distributed storage processing means stores the index data including write status information indicating a status at the time of writing the fragment data referred to by the index data,
The data search means stores the fragment data further referenced in the index data referenced by the unavailable summary table when at least a part of the summary table is unavailable. Searching the index data corresponding to a storage means different from the storage means, and writing the fragment data constituting the search target data based on the summary data of the search target data and the writing of the fragment data The other fragment constituting the search target data by specifying the status information, searching the index data corresponding to the specific storage means based on the specified write status information and the summary data of the search target data Identify the data,
A program to make things happen.

(Appendix 8.1)
The program according to attachment 8, wherein
The distributed storage processing unit stores the summary table separately for each storage unit serving as a storage destination of the fragment data further referred to by the index data referred to by the summary table,
The data search means searches the available summary table corresponding to the other storage means based on partial summary data which is a part of the summary data of the search target data, and the other storage means The index data corresponding to the search target data is searched, the fragment data constituting the search target data and the write status information of the fragment data are specified, and the specified write status information and the summary data of the search target data are specified. The index data corresponding to the specific storage means is searched based on and the other fragment data constituting the search target data is specified.
program.

(Appendix 8.2)
A program according to appendix 8.1,
The distributed storage processing unit stores the index data in the storage unit serving as a storage destination of the fragment data referred to by the index data, and the summary table is referred to by the index table. Storing it in the storage means as a storage destination of the fragment data further referred to by data;
The data search means, when the summary table stored in the specific storage means is not available, based on partial summary data that is a part of the summary data of the search target data, Searching the available summary table stored in the means, searching the index data stored in the other storage means, and a part of the fragment data and the fragment constituting the search target data The write status information of data is specified, and based on the specified write status information and the summary data of search target data, the index data stored in the specific storage means in which the summary table is unavailable Searching to identify other fragment data constituting the search object data;
program.

(Appendix 8.3)
A program according to any one of appendices 8 to 8.2,
The distributed storage processing unit stores the plurality of fragment data generated from one storage target data in a distributed manner in the plurality of storage units at the same timing, and the index data includes the index data. Write timing information representing a value based on the write timing of the fragment data referred to is included and stored as the write status information,
The data search means includes the write timing information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target data Based on the summary data, the index data stored in the specific storage means is searched, and the other fragment data constituting the search target data is specified.
program.

(Appendix 8.4)
A program according to any one of appendices 8 to 8.3,
The distributed storage processing means stores, in the index data, write time information representing a value based on a write time of the fragment data referred to by the index data, including the write status information,
The data search means includes the write time information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target data Based on the summary data, the index data stored in the specific storage means is searched, and the other fragment data constituting the search target data is specified.
program.

(Appendix 9)
A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Perform distributed storage processing
Data for searching a plurality of fragment data constituting the search request data stored in the plurality of storage means by searching the summary table and the index data based on the summary data corresponding to the search request data A data processing method for performing search processing,
During the distributed storage process, the index data is stored including write status information indicating a status at the time of writing the fragment data referred to by the index data,
A specification for storing the fragment data further referred to in the index data referred to by the unavailable summary table when at least a part of the summary table is unavailable during the data search process The index data corresponding to the storage means different from the storage means is searched, and a part of the fragment data and the fragment data constituting the search target data based on the summary data of the search target data The write status information is specified, and the index data corresponding to the specific storage means is searched based on the specified write status information and the summary data of the search target data. Identifying the fragment data;
Data processing method.

(Appendix 10)
A data processing method according to attachment 9, wherein
During the distributed storage process, the summary table is stored separately for each storage unit that is a storage destination of the fragment data that is further referred to by the index data referenced by the summary table,
During the data search process, based on partial summary data that is a part of the summary data of the search target data, the available summary table corresponding to the other storage means is searched for, The index data corresponding to the storage unit is searched, the fragment data constituting the search target data and the write status information of the fragment data are specified, and the specified write status information and the search target data Search the index data corresponding to the specific storage means based on the summary data, and specify the other fragment data constituting the search target data,
Data processing method.

(Appendix 10.1)
A data processing method according to attachment 10, wherein
In the distributed storage processing, the index data is stored in the storage unit that is a storage destination of the fragment data referenced by the index data, and the summary table is referenced by the summary table. Storing it in the storage means as a storage destination of the fragment data further referred to by index data;
When the summary table stored in the specific storage means is unavailable during the data search process, the different summary data is based on partial summary data that is part of the summary data of the search target data. The available summary table stored in the storage means is searched, the index data stored in the other storage means is searched, a part of the fragment data constituting the search target data, and The index stored in the specific storage means that identifies the write status information of the fragment data, and based on the specified write status information and the summary data of the search target data, the summary table is unavailable Search the data to identify the other fragment data constituting the search target data.
Data processing method.

(Appendix 10.2)
A data processing method according to any one of appendices 9 to 10.1,
In the distributed storage process, the plurality of fragment data generated from one storage target data is distributed and stored in the plurality of storage units at the same timing, and the index data is stored in the index data. Write timing information representing a value based on the write timing of the fragment data referred to is included and stored as the write status information,
In the data search process, the write timing information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target Searching the index data stored in the specific storage means based on the summary data of the data, and specifying the other fragment data constituting the search target data,
Data processing method.

(Appendix 10.3)
A data processing method according to any one of appendices 9 to 10.2,
During the distributed storage process, the index data includes write time information representing a value based on the write time of the fragment data referred to by the index data as the write status information.
In the data search process, the write time information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target Searching the index data stored in the specific storage means based on the summary data of the data, and specifying the other fragment data constituting the search target data,
Data processing method.

(Appendix 10.4)
A data processing method according to attachment 10.3,
The data search means searches the index data corresponding to the specific storage means, from the index data associated with the write time information within a predetermined time before and after the specified write time information, Specifying the other fragment data constituting the search target data;
Data processing method.

(Appendix 10.5)
A data processing method according to any one of appendices 10.2 to 10.4,
In the distributed storage process, when the plurality of fragment data are respectively distributed and stored in the plurality of storage units, they are added to the storage areas in the storage units in order.
Storage system.

  Note that the above-described program is stored in a storage device or recorded on a computer-readable recording medium. For example, the recording medium is a portable medium such as a flexible disk, an optical disk, a magneto-optical disk, and a semiconductor memory.

  Although the present invention has been described with reference to the above-described embodiment and the like, the present invention is not limited to the above-described embodiment. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Storage system 2 Access node 21 I / O processing part 22 Data transmission / reception part 23 Duplication determination part 3 Storage node 31 Fragment processing part 32 Fragment search part 33 Time stamp acquisition part 34 Node state monitoring part 35 Disk 36 Hash table 37 Index file 38 Storage file 4 Backup system 5 Backup target device 100 Storage system 111 Distributed storage processing means 112 Data search means 120 Storage means

Claims (10)

A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Distributed storage processing means;
Data that searches the summary table and the index data based on the summary data corresponding to the search target data, and searches a plurality of fragment data constituting the search target data stored in the plurality of storage means A search means,
The distributed storage processing means stores the index data including write status information indicating a status at the time of writing the fragment data referred to by the index data,
The data search means stores the fragment data further referenced in the index data referenced by the unavailable summary table when at least a part of the summary table is unavailable. Searching the index data corresponding to a storage means different from the storage means, and writing the fragment data constituting the search target data based on the summary data of the search target data and the writing of the fragment data The other fragment constituting the search target data by specifying the status information, searching the index data corresponding to the specific storage means based on the specified write status information and the summary data of the search target data Identify the data,
Storage system. The storage system according to claim 1,
The distributed storage processing unit stores the summary table separately for each storage unit serving as a storage destination of the fragment data further referred to by the index data referred to by the summary table,
The data search means searches the available summary table corresponding to the other storage means based on partial summary data which is a part of the summary data of the search target data, and the other storage means The index data corresponding to the search target data is searched, the fragment data constituting the search target data and the write status information of the fragment data are specified, and the specified write status information and the summary data of the search target data are specified. The index data corresponding to the specific storage means is searched based on and the other fragment data constituting the search target data is specified.
Storage system. The storage system according to claim 2,
The distributed storage processing unit stores the index data in the storage unit serving as a storage destination of the fragment data referred to by the index data, and the summary table is referred to by the index table. Storing it in the storage means as a storage destination of the fragment data further referred to by data;
The data search means, when the summary table stored in the specific storage means is not available, based on partial summary data that is a part of the summary data of the search target data, Searching the available summary table stored in the means, searching the index data stored in the other storage means, and a part of the fragment data and the fragment constituting the search target data The write status information of data is specified, and based on the specified write status information and the summary data of search target data, the index data stored in the specific storage means in which the summary table is unavailable Searching to identify other fragment data constituting the search object data;
Storage system. The storage system according to any one of claims 1 to 3,
The distributed storage processing unit stores the plurality of fragment data generated from one storage target data in a distributed manner in the plurality of storage units at the same timing, and the index data includes the index data. Write timing information representing a value based on the write timing of the fragment data referred to is included and stored as the write status information,
The data search means includes the write timing information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target data Based on the summary data, the index data stored in the specific storage means is searched, and the other fragment data constituting the search target data is specified.
Storage system. The storage system according to any one of claims 1 to 4,
The distributed storage processing means stores, in the index data, write time information representing a value based on a write time of the fragment data referred to by the index data, including the write status information,
The data search means includes the write time information which is the write status information of a part of fragment data constituting the search target data specified from the index data corresponding to the other storage means, and the search target data Based on the summary data, the index data stored in the specific storage means is searched, and the other fragment data constituting the search target data is specified.
Storage system. The storage system according to claim 5,
The data search means searches the index data corresponding to the specific storage means, from the index data associated with the write time information within a predetermined time before and after the specified write time information, Specifying the other fragment data constituting the search target data;
Storage system. The storage system according to any one of claims 4 to 6,
The distributed storage processing means additionally writes in order to the storage area in each storage means when the plurality of fragment data is distributed and stored in the plurality of storage means, respectively.
Storage system. In the storage system control unit,
A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Distributed storage processing means;
Data that searches the summary table and the index data based on the summary data corresponding to the search target data, and searches a plurality of fragment data constituting the search target data stored in the plurality of storage means Search means;
And realize
The distributed storage processing means stores the index data including write status information indicating a status at the time of writing the fragment data referred to by the index data,
The data search means stores the fragment data further referenced in the index data referenced by the unavailable summary table when at least a part of the summary table is unavailable. Searching the index data corresponding to a storage means different from the storage means, and writing the fragment data constituting the search target data based on the summary data of the search target data and the writing of the fragment data The other fragment constituting the search target data by specifying the status information, searching the index data corresponding to the specific storage means based on the specified write status information and the summary data of the search target data Identify the data,
A program to make things happen. A plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces, and the plurality of fragment data are respectively distributed and stored in a plurality of storage units;
The index data that associates the information that refers to the fragment data with the summary data that is calculated based on the data content of the storage target data that is configured with the fragment data is referred to in the index data. Storing each of the storage means as fragment data storage destinations separately,
Further, a summary table that associates information that refers to the index data and partial summary data that includes a part of the summary data included in the index data is stored.
Perform distributed storage processing
Data that searches the summary table and the index data based on the summary data corresponding to the search target data, and searches a plurality of fragment data constituting the search target data stored in the plurality of storage means A data processing method for performing search processing,
During the distributed storage process, the index data is stored including write status information indicating a status at the time of writing the fragment data referred to by the index data,
A specification for storing the fragment data further referred to in the index data referred to by the unavailable summary table when at least a part of the summary table is unavailable during the data search process The index data corresponding to the storage means different from the storage means is searched, and a part of the fragment data and the fragment data constituting the search target data based on the summary data of the search target data The write status information is specified, and the index data corresponding to the specific storage means is searched based on the specified write status information and the summary data of the search target data. Identifying the fragment data;
Data processing method. A data processing method according to claim 9, wherein
During the distributed storage process, the summary table is stored separately for each storage unit that is a storage destination of the fragment data that is further referred to by the index data referenced by the summary table,
During the data search process, based on partial summary data that is a part of the summary data of the search target data, the available summary table corresponding to the other storage means is searched for, The index data corresponding to the storage unit is searched, the fragment data constituting the search target data and the write status information of the fragment data are specified, and the specified write status information and the search target data Search the index data corresponding to the specific storage means based on the summary data, and specify the other fragment data constituting the search target data,
Data processing method.

Images