JP6269120B2 - 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.

  In addition, the storage system divides a chunk, which is block data of a predetermined capacity, into a plurality of fragment data, and further adds a fragment that becomes redundant data, and stores each of the plurality of fragment data in a plurality of storage devices. ing. Then, by designating the content address later, the fragment data stored at the storage location specified by the content address can be read, and the predetermined data before division can be restored from the plurality of fragment data.

  In this way, since the storage system adds fragment data that becomes redundant data, even if fragment data equal to or less than the number of fragments of the added redundant data is lost, the original chunk can be regenerated.

  In addition, when a storage system includes nodes that are a plurality of storage devices, the storage system has a distributed arrangement function so that loads are not biased or concentrated between the nodes. Here, each node has a data storage area called a component, and fragment data is stored in each component. In the content address storage system, load distribution is performed on a component basis.

  Here, FIG. 1 shows an example of component arrangement in the storage system. First, the storage system has a plurality of “distribution patterns” for distributing a plurality of components each storing a plurality of fragment data generated from one chunk to a plurality of nodes. Here, each “distributed pattern” has twelve components (reference numerals 01 to 12), and five patterns (reference numerals P1 to P5) in which two or three components are arranged in five nodes, respectively. ).

  Then, the storage system calculates a hash value based on the data content for each chunk from which fragment data is generated, and determines a “distribution pattern” according to the hash value. Store fragment data in each component.

  Here, FIG. 2 shows an example from when data is written until it is distributed and stored in each node. First, the storage system divides chunk D determined to be unique, which has not yet been stored by the duplication determination function, into nine pieces of fragment data F1, and adds fragment data F2, which is three pieces of redundant data, These are distributed between nodes. For this block data, “distribution pattern 1” shown in FIG. 1 is selected from the hash value, and each fragment data is stored in the corresponding component of each node.

  As described above, by distributing and storing fragment data among nodes, even when a node failure occurs, the number of fragment data lost at one time is reduced. Even if some fragment data is lost due to a node failure, the original chunk can be regenerated because redundant data is added. For example, as shown in Patent Document 1, data can be restored by rearranging lost data stored in a failed node in a free area of another node.

JP 2010-191558 A

  However, in the above-described storage system, there is a problem that the capacity of the system is compressed by restoring fragment data on the node where the failure has occurred.

As an example, a storage system composed of five nodes is considered, and the capacity of each node is “100” (unit is arbitrary). And when 80% capacity of the system is used, each capacity is as follows.
・ Used capacity: 400 (= 100 x 5 units x 0.8)
-Free capacity: 100 (= 100 x 5 units-used capacity)

When a node failure occurs, fragment data on the failed node is restored and relocated to the remaining nodes. At that time, the free space becomes “0” as follows. In other words, data with a capacity of “80” (= 100 × 0.8) was arranged on the failed node, but these were restored and “20” was stored in the remaining four nodes. The capacity is distributed and arranged.
-Used capacity: 400 (after restoration, equivalent to before node failure)
-Free capacity: 0 (= 100 x 5 units-used capacity)

  As described above, in a storage system in which data is distributed and arranged, there arises a problem that the capacity of the system is exhausted when a node failure occurs.

  Therefore, an object of the present invention is to solve the above-mentioned problem that the capacity of the system is exhausted in the event of a node failure in a storage system in which data is distributed and arranged.

A storage system according to an aspect of the present invention
When storing data in a plurality of storage means, generate a plurality of fragment data consisting of divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data, Distributed storage processing means for distributing and storing each in the storage means;
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. Data restoration means for restoring based on the fragment data and storing in the other storage means,
The data restoration means calculates all data capacities in the storage system after restoration when at least a part of the lost fragment data is restored, and all the calculated data capacities after restoration, Reconstructed data number determining means for determining the number of fragment data to be restored based on the capacity of the storage means in which no failure has occurred;
The configuration is as follows.

In addition, a program which is one embodiment of the present invention is
In the information processing device,
Distributed storage processing means for generating a plurality of fragment data composed of divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data, and storing each of the fragment data in a plurality of storage means,
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. And restoring the data based on the fragment data and storing the data in the other storage means,
Calculate all data capacities in the restored storage system when at least a part of the lost fragment data is restored, and no faults have occurred with the calculated data capacities after restoration. Realizing the data restoration means having restoration data number determination means for determining the number of fragment data to be restored based on the capacity of the storage means;
The configuration is as follows.

A data processing method according to one aspect of the present invention includes:
When storing data in a plurality of storage means, generate a plurality of fragment data consisting of divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data, Each of the storage means is distributed and stored,
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. The data is restored based on the fragment data and stored in the other storage means.
During the data restoration process, all data capacities in the storage system after restoration when at least a part of the lost fragment data is restored are calculated, and all the calculated data capacities after restoration are calculated. And determining the number of fragment data to be restored based on the capacity of the storage means in which no failure has occurred,
The configuration is as follows.

  With the configuration as described above, the present invention can suppress system capacity depletion in the event of a node failure in a storage system in which data is distributed.

It is a figure which shows the distribution pattern of the fragment data with respect to several nodes in the storage system relevant to this invention. It is a figure which shows the mode at the time of the data writing in the storage system relevant to this invention. 1 is a block diagram showing a configuration of an entire system including a storage system in Embodiment 1 of the present invention. FIG. 4 is a block diagram illustrating an overall configuration of the storage system disclosed in FIG. 3. FIG. 5 is a functional block diagram illustrating a configuration of the storage system disclosed in FIG. 4. FIG. 6 is an explanatory diagram for explaining the operation of the storage system disclosed in FIG. 5. FIG. 6 is an explanatory diagram for explaining the operation of the storage system disclosed in FIG. 5. FIG. 6 is an explanatory diagram for explaining the operation of the storage system disclosed in FIG. 5. 6 is a flowchart showing an operation of the storage system disclosed in FIG. 5. 6 is a flowchart showing an operation of the storage system disclosed in FIG. 5. 6 is a flowchart showing an operation of the storage system disclosed in FIG. 5. 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. 3 is a block diagram showing the configuration of the entire system. FIG. 4 is a block diagram showing an outline of the configuration of the storage system, and FIG. 5 is a functional block diagram showing the configuration of the storage system. 6 to 8 are explanatory diagrams for explaining operations of data writing processing and data restoration processing in the storage system. 9 to 11 are flowcharts showing the operation of the storage system.

  Here, 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. 3, the storage system 1 according to the present invention is connected to a backup system 4 that controls backup processing via a network N. Then, the backup system 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.

  As shown in FIG. 4, 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. Note that the number of access nodes 2 and the number of storage nodes 3 are not limited to those shown in FIG. 4, and more nodes 2 and 3 may be connected.

  The storage system 1 in the present embodiment divides and makes the data redundant, stores the data in a plurality of storage devices, and stores the data by a unique content address set according to the content of the stored data. It is a content address storage system for specifying a stored location. This content address storage system will be described in detail later.

  In the following description, it is assumed that the storage system 1 is constructed by the access node 2 and the storage node 3. However, the storage system 1 is not necessarily limited to the access node 2 and the storage node 3, and any configuration is possible. There may be. For example, the storage system in the present invention may be configured by a single computer. Furthermore, the storage system 1 is not limited to being a content address storage system, but may be any storage system having a function of distributing and storing data in a plurality of storage devices.

  FIG. 5 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 a file system service 20, a deduplication function unit 21, and a fragment distribution function unit 22, which are constructed by incorporating a program into the equipped arithmetic device. The node alive monitoring function unit 23 and the data restoration control function unit 24 are provided. Further, the access node 2 has a hash table 25, component capacity information 26, component number information 27, system capacity information 28, and component arrangement information 29 as information stored in the equipped storage device.

  In addition, the storage node 3 includes a fragment storage function unit 31, a data restoration function unit 32, and a disk alive monitoring function unit 33, which are constructed by incorporating a program in the equipped arithmetic device. The storage node 3 has a physical disk 30 as a storage device for storing fragmented data as will be described later. All the storage nodes 3 shown in FIG. 5 have the same configuration.

  Here, as described above, the storage system 1 in the present embodiment is a content address storage system. For this reason, the storage system 1 has a function of storing data in the physical disk 30 of the storage node 3 using the content address. As will be described below, the data is divided and distributed, and the content The storage location is specified by the address, and the data is stored.

  First, data write processing by the storage system 1 will be described with reference to the explanatory diagrams of FIGS. 6 to 8 and the flowchart of FIG.

  First, information stored in the access node 2 will be described. As will be described later, the hash table 25 stores the hash value of the chunk D that is already stored in the physical disk 30 of the storage node 3. The component capacity information 26 stores the capacity per component. The component number information 27 stores the total number of components per distribution pattern. The system capacity information 28 stores information on the storage capacity of each storage node 3. The component arrangement information 29 stores component arrangement information for a plurality of storage nodes 3.

  The write process is started when the file system service 20 of the access node 2 receives the file A, which is the data requested to be written from the host device, as indicated by the arrow Y1 in FIGS. 7 and 8 (step S1). . Then, the deduplication function unit 21 of the access node 2 divides the file A into chunks D (data to be written) that is block data of a predetermined capacity (for example, 64 KB) as shown by an arrow Y2 in FIGS. (Step S2).

  Subsequently, a unique hash value H representing the data content is calculated based on the data content of the divided chunk D (arrow Y3 in FIG. 8) (step S3). For example, the hash value H is calculated from the data content of the chunk D using a preset hash function.

  Subsequently, using the hash value H of the chunk D of the file A, it is checked whether or not the chunk D has already been stored (step S4). At this time, the chunk D that has already been stored is registered in the hash table 25 that is an MFI (Main Fragment Index) file in which the hash value H and the content address CA representing the storage location are stored in association with each other. Therefore, if the hash value H of the chunk D calculated before storage exists in the MFI file, it can be determined that the chunk D having the same content has already been stored (arrow Y4 in FIG. 8). In this case, the content address CA associated with the hash value H in the MFI that matches the hash value H of the chunk D before storage is acquired from the MFI file. Then, this content address CA is returned as the content address CA of the chunk D requested to be written.

  Then, the already stored data referred to by the returned content address CA is used as the chunk D requested to be written. That is, it is assumed that the chunk D requested to be written is stored by designating an area referred to by the returned content address CA as a storage destination of the chunk D requested to be written. As described above, when it is determined that the chunks D related to the write request are duplicated (step S5: Yes), the write completion is notified to the upper level without actually writing.

  On the other hand, when it is determined that the chunk D related to the write request is not duplicated and has not been stored yet (step S5: No), the unique chunk D is written as follows. First, the fragment distribution function unit 22 compresses the chunk D and divides the chunk D into a plurality of pieces of fragment data having a predetermined capacity as indicated by an arrow Y5 in FIG. 8 (step S6). For example, as indicated by reference numerals D1 to D9 in FIG. 7, the chunk 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 chunk D can be restored and added to the divided fragment data F1 (step S7). 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.

  At this time, the fragment distribution function unit 22 sets the number of redundant data fragments based on component number information 27 indicating the total number of components per distribution pattern, which will be described later, and generates a data set composed of a plurality of fragment data. To do. For example, when the number-of-components information is “12”, as described above, a data set composed of 12 fragment data composed of nine divided data F1 and three redundant data F2 is generated. When the number-of-components information is “10”, as described above, a data set composed of ten fragment data composed of nine divided data F1 and one redundant data F2 is generated. The component number information is set to “12” in the initial state, but is a value updated by the data restoration function units 24 and 32 described later.

  Subsequently, the fragment distribution function unit 22 cooperates with the fragment storage function unit 31 of the storage node 3 to transfer each fragment data constituting the data set generated as described above to the physical disks of the plurality of storage nodes 3. 30 are stored in a distributed manner. At this time, the fragment distribution function unit 22 determines the distribution pattern of the fragment data of the chunk based on the hash value of the chunk that is the source of the fragment data. Then, according to the determined distribution pattern, the fragment distribution function unit 22 arranges data storage areas called components on the physical disks 30 of the respective storage nodes 3, and stores fragment data in the respective components (arrows in FIG. 8). (See Y6) (Step S8).

  Here, the distribution pattern is information representing the arrangement of each component in which a plurality of fragment data composed of one chunk D is distributed and stored with respect to a plurality of storage nodes 3. Several types of distribution patterns are set in advance, and are stored in the component arrangement information 29 as shown in FIG. 6, for example. Here, each “distributed pattern” (P1 to P5) has twelve components (reference numerals 01 to 12), and five patterns in which two or three components are arranged in five nodes, respectively. It has become.

  In the example shown in FIG. 7, “distribution pattern 1” (P1) is determined from the hash value of chunk D, each component is arranged in each storage node 3 with this distribution pattern, and each fragment data is stored. That is, in this example, among the five storage nodes, three fragment data are stored in two storage nodes, and two fragment data are stored in three storage nodes.

  At this time, the fragment distribution function unit 22 stores and updates the capacity of each component in the component capacity information 29 based on the capacity of the fragment data written in each component (step S9).

  Subsequently, the fragment distribution function unit 22 generates a content address CA indicating the storage position of the fragment data D1 to D12 stored as described above, that is, the storage position of the chunk D restored by the fragment data D1 to D12. And manage. Specifically, a part (short hash) of the hash value H calculated based on the contents of the stored chunk D (for example, the top 8B (bytes) of the hash value H) and information indicating the logical storage position, In combination, a content address CA is generated. Then, the content address CA is returned to the file system service 20 (arrow Y7 in FIG. 8), and the identification information such as the file name of the data to be backed up is associated with the content address CA and managed by the MFI file described above. As a result, the stored content address CA of the chunk D is stored in the hash table 25. At this time, the distribution pattern of each chunk D is also managed. For example, the distribution pattern when the fragment data of each chunk D is distributed and stored is stored and managed in the component arrangement information 29 and the MFI file.

  Then, the file system service 20 notifies the host device of the completion of writing the data requested to be written (step S10). In this way, the writing process is completed.

  Next, data read processing by the storage system 1 will be described. When a read request is made by designating a specific file to the file system service 20 of the storage system 1 from the host device, first, a short hash and a logical position that are part of the hash value corresponding to the file related to the read request A content address CA consisting of the above information is designated. Then, it is checked whether or not the content address CA is registered in the hash table 25 that is an MFI file. If it is not registered, the requested data is not stored and an error is returned.

  On the other hand, when the content address CA related to the read request is registered, the storage location specified by the content address CA is specified, and each fragment data stored in the specified storage location is read. Read as requested data. At this time, if the data storage file in which each fragment is stored and the storage position of one fragment data in the data storage file are known, the storage position of other fragment data can be specified from the same storage position. .

  Then, the original chunk D is restored from each fragment data read in response to the read request. Further, a plurality of restored chunks D are connected, restored to a group of data such as file A, and returned.

  Next, data restoration processing when a failure occurs in any one of the plurality of storage nodes 3 and data is restored will be described with reference to the flowcharts of FIGS. 9 to 10.

  First, when the node alive monitoring function unit 23 of the access node 2 detects a node failure of the storage node 3 (step S21), the data restoration control function unit 24 of the access node 2 is lost from the component arrangement information 29. It is confirmed whether the fragment data under the component can be restored (step S22). For example, whether or not restoration is possible is confirmed based on whether or not fragment data larger than the number of redundant data is lost. If it is determined that the fragment data cannot be restored (step S23: No), the process is terminated without performing anything. If it is determined that the fragment data can be restored (step S23: Yes), it is determined how far the component is restored (step S24).

  Details of the process for determining how far the component in step S24 is to be restored will be described with reference to the flowchart of FIG. This process is performed by the data restoration function unit 24 of the access node 2, the data restoration function unit 32 and the fragment storage function unit 31 of the storage node 3 in cooperation.

  First, when it is determined that the fragment under the component lost due to the node failure can be restored (step S31), the data restoration function units 24, 32, etc. (data restoration means, restoration data number determination means) Determine whether to restore. At this time, as will be described below, the component to be restored is determined in two stages. In the first stage, processing for determining the components to be restored individually for each distribution pattern is performed, and in the second stage, the total data capacity when a predetermined number of components are restored in all the distribution patterns is preset. It is determined by whether or not the threshold value (allowable capacity value) or less. Note that the threshold value is determined by the system user based on the usage status of the system, and is set to 90% of the total capacity of the remaining storage nodes 3 in which no failure has occurred, for example. However, the threshold value is not limited to the value described above, and may be a value set according to the total capacity or the free capacity of the remaining storage node 3 in which no failure has occurred.

  Specifically, first, the number of components of each chunk D lost in each distribution pattern due to a node failure is confirmed from the component arrangement information 29 and the distribution pattern management information of each chunk D (step S32). If the number of components lost between the distribution patterns is different (step S32: Yes), the restoration of the components of each chunk D corresponding to the distribution pattern with many lost components is determined (step S33). At this time, it is determined to restore the component for each chunk D corresponding to the distributed pattern having a large number of lost components so that the number of components lost in each chunk D of each distributed pattern is the same. For example, when each chunk is stored in the distribution pattern shown in FIG. 6 and a failure occurs in the node 5, two components are lost in the chunk D of the distribution patterns P1, P2, and P3, and the distribution pattern P4 , P5 chunks will lose three components. Therefore, it is determined to restore one component of each chunk D of each of the dispersion patterns P4 and P5. If the number of lost components is uniform in all the distribution patterns (step S32: No), nothing is performed here.

  Next, the data restoration function units 24 and 32 (data restoration means, restoration data number determination means) change the number of components to be restored based on the component arrangement information 29 and the distribution pattern management information of each chunk D. Then, it is confirmed when the restored data capacity when each number of components is restored falls below the threshold (steps S34, S35: No, S36). At this time, the number of restorations is decreased one by one from the minimum number of components lost so that the number of components in the chunk D corresponding to all the distribution patterns is the same, and the number of restorations of components is changed. For example, when a failure occurs in the node 5 in the distributed pattern shown in FIG. 6, the number of restorations is first set to “2”, and then changed to “1” and “0”. The data restoration function units 24, 32, etc. calculate all the restored data capacities based on the component capacity information 26 when the respective restoration number components are restored with all the distributed pattern chunks D. And it is confirmed whether the calculated data capacity is less than the said threshold value.

  When it is confirmed that the data capacity to be restored is less than the threshold when restoring n components (n is an integer other than 0) in all the distributed pattern chunks D (step S36: No), It is determined to restore n components in chunk D of the distribution pattern (step S38). When the number of components to be restored is determined, a value obtained by adding the number to be restored and the number of remaining components is updated as new component number information 27 (step S39).

  The operation after determining the number of components to be restored in all the distribution patterns as described above will be described with reference to step S24 and subsequent steps in FIG. As described above, the number of components to be restored is determined (step S24), and when the component number information 27 is updated (step S25), it is confirmed whether there is a component to be restored (step S26). When there is a component to be restored (step S26: No), it is determined on which storage node 3 the component is restored in consideration of the capacity load of each storage node 3 (step S27). The subordinate fragment is restored (step S28), and the fragment is confirmed in the component (step S29).

  Here, a specific example of data restoration at the time of a node failure in the above-described storage system will be described. First, as shown in FIG. 6, a case is assumed where chunks D of 12 distribution patterns (P1 to P5) of the total number of components are stored in 5 storage nodes. At this time, the capacity of each storage node is “100” (unit is arbitrary), and the capacity of each component is “7”. The threshold value described above is “90%”.

  According to the above conditions, if no failure has occurred in any storage node, the capacity of the entire storage node 3 is “500”, and the total capacity of the components is “420” (= 7 × 60 components). It is. Then, the usage amount of the storage node 3 is “84%”.

  In the above situation, when a failure occurs in “storage node 5”, the number of components lost is “2” for “distribution patterns 1, 2, 3” and “3” for “distribution patterns 4, 5”, respectively. "" Then, since the number of lost components is different in each chunk D of “distribution patterns 1, 2, 3” and “distribution patterns 4, 5”, the “distribution patterns 4, 5” are made to be the same. The “one each” of the components is determined as a restoration target. As a result, the number of components lost when the components are restored becomes the same as “two” in all the distribution patterns.

Next, the number of components to be restored is determined for all distribution patterns. First, the total capacity of components when “two” (n = 2) components are restored in all distribution patterns, that is, the total data capacity after restoration is calculated.
Component total capacity = 420 (= 7 × 12 components × 5 distribution pattern)
The total component capacity is 105% of the total capacity “400” (= 100 × 4) of the four storage nodes after the storage node failure, and thus exceeds the threshold value. The component is restored when n = 2. Absent.

Next, the total capacity of the components when restoring “one” (n = 1) components in all the distribution patterns is calculated.
Component total capacity = 385 (= 7 × 11 components × 5 distribution pattern)
The total component capacity is about 96% of the total capacity “400” of the four storage nodes after the storage node failure, and thus exceeds the threshold value, and the component is not restored when n = 1.

  As a result, in the above-described example, one component is restored in each chunk D corresponding to each of “distribution patterns 4 and 5”. At this time, the total number of components per distributed pattern is “10”, and the component number information 27 is updated. As a result, the fragment distribution function unit 22 generates 12 pieces of fragment data by adding three redundant fragments to nine fragments for one chunk D before the node failure. For one chunk D, one redundant fragment is added to nine fragments to generate ten fragment data.

  In the specific example described above, if the threshold value is 97%, n = 1, that is, the number of restorations is “1” in all the dispersion patterns, and restoration is performed using the above-described “dispersion patterns 4 and 5”. When combined with the component to be restored, the total number of components is restored to “11” in all distribution patterns.

  As described above, when all the components are restored when the capacity becomes “400” after the failure of the storage node, the capacity becomes “420”, and the system capacity may be exhausted during fragment restoration. there were. However, in the storage system according to the present invention, only a fragment of some components is restored, so that the system capacity can be prevented from being exhausted.

  At this time, the free capacity before the failure of the storage node is “100”, and in this system, since a redundant fragment is added, the capacity that can actually be written from the upper level is “75” (= 100 × 9/12). there were. On the other hand, after the storage node failure, the free capacity is “50” (= 400−350), but the number of redundant fragments has changed from “3” to “1”, so that the capacity that can be written from the upper level is available. Becomes “45”. That is, the free space has decreased from “100” to “50” to “50%”, but the actual writable capacity can be reduced by “40%” from “75” to “45”. Thus, efficient use of the storage capacity can be achieved.

  Further, in the above-described example, if no component is restored, data redundancy is reduced. Here, a case is considered in which, for example, a part of the fragment is lost due to, for example, a physical disk failure detected in each storage node 3 detected by the disk alive monitoring function unit 33. In this case, chunks can be reconfigured in “distributed patterns 1, 2, 3”, but in “distributed patterns 4, 5”, three pieces of fragment data are lost for three redundant fragments. It cannot be reconfigured. However, in the present invention, in order to make the number of components the same as “distribution patterns 1, 2, and 3”, one component is restored in each of “distribution patterns 4 and 5”, so a part of fragment data is lost. Even in this case, chunks can be reconfigured even in “distributed patterns 4 and 5”. As a result, the chunk tolerance, that is, the redundancy can be kept constant, and the reliability of the system can be improved.

<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 (see FIG. 12), program, and data processing method according to the present invention will be described below. However, the present invention is not limited to the following configuration.

(Appendix 1)
When storing data in the plurality of storage means 120, a plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data are generated, and the plurality Distributed storage processing means 111 for storing each of the storage means in a distributed manner,
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. Data restoring means 112 for restoring based on the fragment data and storing it in the other storage means,
The data restoration unit 112 calculates all data capacities in the storage system after restoration when at least a part of the lost fragment data is restored, and all the calculated data capacities after restoration A restoration data number determining means 113 for determining the number of fragment data to be restored based on the capacity of the storage means in which no failure has occurred,
Storage system 100.

(Appendix 2)
The storage system according to attachment 1, wherein
The restoration data number determining means determines the number of fragment data to be restored so that the calculated data capacity after restoration is not more than an allowable capacity value based on the capacity of the storage means in which no failure has occurred. To
Storage system.

(Appendix 3)
The storage system according to appendix 1 or 2,
The distributed storage unit distributes the plurality of fragment data to the plurality of storage units according to a distribution pattern representing an arrangement state of the plurality of fragment data generated from one storage target data with respect to each storage unit. Remember,
The restoration data number determining means confirms the number of fragment data lost due to a failure based on the distribution pattern, and determines the number of fragment data to be restored from the lost fragment data. The fragment to be restored is calculated based on all the calculated data capacities in the restored storage system while changing, and the calculated capacity of the restored storage means and the capacity of the storage means in which no failure has occurred. Determine the number of data,
Storage system.

(Appendix 4)
The storage system according to attachment 3, wherein
The distributed storage means stores the fragment data in a distributed manner according to the distribution pattern in which the arrangement state of the fragment data in the plurality of storage means is different for each storage target data,
The restoration data number determination means decides the number of fragment data to be restored so that the number of fragment data after restoration is the same for all the storage target data distributed and stored in each distribution pattern. ,
Storage system.

  According to the above invention, when a failure occurs in any one of the plurality of storage means, the number of fragment data to be restored is determined based on the capacity of the remaining storage means in which no failure has occurred. Restore all or part of the fragment data. Specifically, when there are a plurality of lost fragment data for all the storage target data, the number of fragment data for all the storage target data is the same, and the restored data capacity is the remaining storage capacity. The number of fragment data to be restored is determined so as to be less than the allowable capacity value based on the capacity of the means. Then, by restoring only the determined number of fragment data, it is possible to suppress the depletion of the capacity of the entire system while maintaining the reliability of the system due to the restoration of the data.

(Appendix 5)
The storage system according to appendix 4, wherein
The restoration data number determining means determines the number of lost fragment data for each storage target data distributed and stored in each distribution pattern before determining the number of fragment data to be restored. Deciding to restore the fragment data of the storage target data with a large number of lost fragment data,
Storage system.

(Appendix 6)
The storage system according to appendix 5,
The restoration data number determining means determines the number of lost fragment data for each storage target data distributed and stored in each distribution pattern before determining the number of fragment data to be restored. Deciding to restore the fragment data of the storage target data with a large number of lost fragment data so that the number of the fragment data is the same.
Storage system.

  With the above configuration, when the number of fragment data lost among the storage target data differs due to different distribution patterns, the fragment data for the storage target data with a large number of lost is preferentially restored. As a result, the durability, that is, the redundancy of the data to be stored thereafter can be kept constant, and the reliability of the system can be improved.

(Appendix 7)
The storage system according to any one of appendices 1 to 6,
The distributed storage processing unit generates the number of fragment data for the storage target data after restoration based on the number of fragment data to be restored determined by the restoration data number determination unit from the storage target data. Set as the number of fragment data, generate the number of fragment data from the storage target data, and distribute and store in the plurality of storage means,
Storage system.

  With the above configuration, even after a failure of the storage means, it is possible to ensure redundancy while suppressing the depletion of the capacity, and to efficiently use the storage capacity.

(Appendix 8)
In the information processing device,
Distributed storage processing means for generating a plurality of fragment data composed of divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data, and storing each of the fragment data in a plurality of storage means,
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. And restoring the data based on the fragment data and storing the data in the other storage means,
Calculate all data capacities in the restored storage system when at least a part of the lost fragment data is restored, and no faults have occurred with the calculated data capacities after restoration. Realizing the data restoration means having restoration data number determination means for determining the number of fragment data to be restored based on the capacity of the storage means;
Program for.

(Appendix 8.1)
The program according to attachment 8, wherein
The distributed storage unit distributes the plurality of fragment data to the plurality of storage units according to a distribution pattern representing an arrangement state of the plurality of fragment data generated from one storage target data with respect to each storage unit. Remember,
The restoration data number determining means confirms the number of fragment data lost due to a failure based on the distribution pattern, and determines the number of fragment data to be restored from the lost fragment data. The fragment to be restored is calculated based on all the calculated data capacities in the restored storage system while changing, and the calculated capacity of the restored storage means and the capacity of the storage means in which no failure has occurred. Determine the number of data,
program.

(Appendix 8.2)
A program according to appendix 8.1,
The distributed storage means stores the fragment data in a distributed manner according to the distribution pattern in which the arrangement state of the fragment data in the plurality of storage means is different for each storage target data,
The restoration data number determination means decides the number of fragment data to be restored so that the number of fragment data after restoration is the same for all the storage target data distributed and stored in each distribution pattern. ,
program.

(Appendix 8.3)
A program according to appendix 8.2,
The restoration data number determining means determines the number of lost fragment data for each storage target data distributed and stored in each distribution pattern before determining the number of fragment data to be restored. Deciding to restore the fragment data of the storage target data with a large number of lost fragment data,
program.

(Appendix 8.4)
The program according to attachment 8.3,
The restoration data number determining means determines the number of lost fragment data for each storage target data distributed and stored in each distribution pattern before determining the number of fragment data to be restored. Deciding to restore the fragment data of the storage target data with a large number of lost fragment data so that the number of the fragment data is the same.
program.

(Appendix 9)
When storing data in a plurality of storage means, generate a plurality of fragment data consisting of divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data, Each of the storage means is distributed and stored,
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. The data is restored based on the fragment data and stored in the other storage means.
During the data restoration process, all data capacities in the storage system after restoration when at least a part of the lost fragment data is restored are calculated, and all the calculated data capacities after restoration are calculated. And determining the number of fragment data to be restored based on the capacity of the storage means in which no failure has occurred,
Data processing method.

(Appendix 10)
A data processing method according to attachment 9, wherein
When storing data, each of the plurality of fragment data is distributed to the plurality of storage units according to a distribution pattern representing an arrangement state of the plurality of fragment data generated from one storage target data with respect to each storage unit. Remember,
When determining the number of fragment data to be restored, based on the distribution pattern, the number of fragment data lost due to a failure is confirmed, and the fragment data to be restored is restored. Calculate all data capacities in the restored storage system while changing the number of fragment data, and based on the calculated all data capacities after restoration and the capacity of the storage means in which no failure has occurred Determine the number of fragment data to be restored;
Data processing method.

(Appendix 10.1)
A data processing method according to attachment 10, wherein
When storing data, for each of the storage target data, according to the distribution pattern in which the fragment data arrangement status with respect to the plurality of storage means are different, respectively, the fragment data is respectively distributed and stored,
When determining the number of fragment data to be restored, the fragment data to be restored so that the number of fragment data after restoration is the same for all the storage target data distributed and stored in each of the distribution patterns Determine the number of
Data processing method.

(Appendix 10.2)
A data processing method according to attachment 10.1,
Before determining the number of fragment data to be restored, if the number of fragment data lost is different for each storage target data distributed and stored in each distribution pattern, the lost fragment data Determining to restore the fragment data of the storage target data having a large number;
Data processing method.

(Appendix 10.3)
A data processing method according to attachment 10.2,
Before determining the number of fragment data to be restored, the number of fragment data is the same when the number of lost fragment data is different for each storage target data distributed and stored in each distribution pattern. Determining to restore the fragment data of the storage target data in which the number of lost fragment data is large so that
Data processing method.

  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 20 File system service 21 Deduplication function part 22 Fragment distribution function part 23 Node alive monitoring function part 24 Data restoration function part 25 Hash table 26 Component capacity information 27 Component number information 28 System capacity information 29 Component arrangement information 3 Storage node 30 Physical disk 31 Fragment storage function unit 32 Data restoration function unit 33 Disk alive monitoring function unit 4 Backup system 5 Backup target device 100 Storage system 111 Distributed storage processing unit 112 Data restoration unit 113 Restored data number determination unit 120 Storage unit

Claims (8)

When storing data in a plurality of storage means, generate a plurality of fragment data consisting of divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data, Distributed storage processing means for distributing and storing each in the storage means;
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. Data restoration means for restoring based on the fragment data and storing in the other storage means,
The data restoration means calculates all data capacities in the storage system after restoration when at least a part of the lost fragment data is restored, and all the calculated data capacities after restoration, based on the capacity of the storage means there is no failure, have a restoring data number determining means for determining the number of the fragment data to be restored,
The distributed storage processing unit distributes the plurality of fragment data to the plurality of storage units according to a distribution pattern representing an arrangement state of the plurality of fragment data generated from one storage target data with respect to each storage unit. Remember,
The restoration data number determining means confirms the number of fragment data lost due to a failure based on the distribution pattern, and determines the number of fragment data to be restored from the lost fragment data. The fragment to be restored is calculated based on all the calculated data capacities in the restored storage system while changing, and the calculated capacity of the restored storage means and the capacity of the storage means in which no failure has occurred. Determine the number of data,
Storage system. The storage system according to claim 1,
The restoration data number determining means determines the number of fragment data to be restored so that the calculated data capacity after restoration is not more than an allowable capacity value based on the capacity of the storage means in which no failure has occurred. To
Storage system. The storage system according to claim 1 or 2 ,
The distributed storage processing means stores the fragment data in a distributed manner in accordance with the distribution pattern in which the fragment data arrangement status with respect to the plurality of storage means is different for each storage target data,
The restoration data number determination means decides the number of fragment data to be restored so that the number of fragment data after restoration is the same for all the storage target data distributed and stored in each distribution pattern. ,
Storage system. The storage system according to claim 3 ,
The restoration data number determining means determines the number of lost fragment data for each storage target data distributed and stored in each distribution pattern before determining the number of fragment data to be restored. Deciding to restore the fragment data of the storage target data with a large number of lost fragment data,
Storage system. The storage system according to claim 4 ,
The restoration data number determining means determines the number of lost fragment data for each storage target data distributed and stored in each distribution pattern before determining the number of fragment data to be restored. Deciding to restore the fragment data of the storage target data with a large number of lost fragment data so that the number of the fragment data is the same.
Storage system. The storage system according to any one of claims 1 to 5 ,
The distributed storage processing unit generates the number of fragment data for the storage target data after restoration based on the number of fragment data to be restored determined by the restoration data number determination unit from the storage target data. Set as the number of fragment data, generate the number of fragment data from the storage target data, and distribute and store in the plurality of storage means,
Storage system. In the information processing device,
Distributed storage processing means for generating a plurality of fragment data composed of divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data, and storing each of the fragment data in a plurality of storage means,
If a failure occurs in any of the storage means, the fragment data lost due to the occurrence of the failure is stored in another storage means different from the storage means in which the failure has occurred. And restoring the data based on the fragment data and storing the data in the other storage means,
Calculate all data capacities in the restored storage system when at least a part of the lost fragment data is restored, and no faults have occurred with the calculated data capacities after restoration. Based on the capacity of the storage means, realizing the data restoration means having restoration data number determination means for determining the number of fragment data to be restored ,
The distributed storage processing unit distributes the plurality of fragment data to the plurality of storage units according to a distribution pattern representing an arrangement state of the plurality of fragment data generated from one storage target data with respect to each storage unit. Remember,
The restoration data number determining means confirms the number of fragment data lost due to a failure based on the distribution pattern, and determines the number of fragment data to be restored from the lost fragment data. The fragment to be restored is calculated based on all the calculated data capacities in the restored storage system while changing, and the calculated capacity of the restored storage means and the capacity of the storage means in which no failure has occurred. Determine the number of data,
A program to make things happen . When the information processing apparatus stores data in a plurality of storage means, it generates a plurality of fragment data including divided data obtained by dividing the storage target data into a plurality of pieces and redundant data for restoring the storage target data. Each of the plurality of storage means is distributed and stored,
When the information processing apparatus fails in any of the storage means, the fragment data lost as a result of the failure is stored in another storage means different from the storage means in which the failure has occurred. Is restored based on the other fragment data stored in the data, and is stored in the other storage means,
The information processing apparatus calculates all data capacities in the storage system after restoration when at least a part of the lost fragment data is restored during the data restoration processing, and the calculated restoration Based on all subsequent data capacity and the capacity of the storage means in which no failure has occurred, determine the number of fragment data to be restored ,
When the information processing apparatus stores data, the plurality of fragment data is converted into the plurality of fragment data according to a distribution pattern representing an arrangement state of the plurality of fragment data generated from one storage target data with respect to each storage unit. Each of the storage means is distributed and stored,
When the information processing apparatus determines the number of fragment data to be restored, the information processing apparatus confirms the number of fragment data lost due to a failure based on the distribution pattern, and the lost fragment Calculate all data capacities in the restored storage system while changing the number of fragment data to be restored among the data, and calculate all the data capacities after restoration and the storage means in which no failure has occurred Determine the number of fragment data to be restored based on the capacity of
Data processing method.

Images