JP6961045B2 - System and its control method and program - Google Patents

Description

The present invention relates to a system, a control method thereof, and a program, and is suitable for application to, for example, an information processing system including a plurality of storage nodes on which one or a plurality of SDSs (Software Defined Storage) are implemented. In the following, SDS refers to a storage device constructed by mounting software having a storage function on a general-purpose server device.

Conventionally, the existing control software (storage control software) used for the controller of the storage device is configured as a dedicated product corresponding to the hardware on which the storage control software is mounted. Therefore, there is a problem that it is difficult to construct a scale-out configuration due to the difference in architecture just by installing this storage control software on a bare metal server as it is.

This is because the existing storage control software is configured only for the processing that is completed in the own storage device. Therefore, it is not possible to use the existing storage control software in a scaled-out configuration. could not. Although Patent Document 1 below discloses that two storage devices having an HA configuration cooperate to perform offload data copying, this is also limited to the cooperation of the two storage devices. rice field.

Further, since the processing of the shared memory executed by the existing storage control software depends on the hardware, there is also a problem that the same processing cannot be executed by a general-purpose server device.

On the other hand, in recent years, the development of SDS constructed by implementing storage control software on a general-purpose server device has been promoted. Demand for SDS is on the rise because it does not require dedicated hardware and is highly expandable. As a technique related to such SDS, for example, Patent Document 2 below discloses a technique for transferring I / O (Input / Output) between servers having an SDS configuration.

U.S. Patent Application Publication No. 2017/0017433 U.S. Patent Application Publication No. 2016/0173598

By the way, in recent years, the amount of data accumulated in companies and government offices has been steadily increasing. Considering this situation, it is desirable that the storage device has a configuration that can be easily scaled out.

In addition, if the system can be constructed so that the higher-level device can easily access the desired data without being aware of the storage device to which the I / O request is issued even after the scale-out, the setting of the higher-level device after the scale-out can be performed. It is considered that it becomes unnecessary and the expandability of the system can be improved.

The present invention has been made in consideration of the above points, and an object of the present invention is to propose a system capable of improving extensibility, a control method thereof, and a program.

In order to solve such a problem, in the present invention, the system includes a cluster composed of a plurality of storage nodes, and each storage node includes a storage device, a logical storage area corresponding to the storage device, and the like. When receiving a logical volume creation instruction based on a request from the management device, the storage node can be identified in the cluster associated with a host volume for reading and writing data provided to a higher-level device. A virtual global pool volume is created in the local node, and the logical storage area in the local node can be associated with the global pool volume created in the local node so that the global pool volume can be identified in the local node. the virtual internal volume created in this node, is registered as the corresponding relationship Gama mappings information and the global pool volumes created, and the node to which the global pool volume was created, the mapping information is said cluster When each storage node receives an I / O request from the higher-level device, the storage node is designated as an I / O destination in the I / O request based on the mapping information. The storage node of the I / O destination is specified from the information of the storage node corresponding to the global volume associated with the host volume, and if the specified storage node is the own node, the I / O is specified. The volume number of the host volume of the I / O destination specified in the / O request is rewritten to the volume number of the corresponding internal volume, I / O processing is executed for the internal volume, and the specified storage node In the case of another storage node, the I / O request is distributed to the storage node.

Further, in the present invention, in a system including a cluster composed of a plurality of storage nodes, the method is executed by the storage nodes, and each storage node is a storage device and a logical corresponding to the storage device. When a logical volume creation instruction based on a request from a management device is received, the storage node is associated with a host volume for reading and writing data provided to a higher-level device. The first step of creating a virtual global pool volume that can be identified in the own node is associated with the logical storage area in the own node and the global pool volume created in the own node. Te, a second step of creating a virtual internal volume that can be identified in the local node in the local node, and the global pool volumes created, a corresponding relationship between the node that created the global pool volume registered as mapping information, the mapping information provided and a third step of updating synchronously among the storage nodes in the cluster, each storage node, receives the I / O request from said host device Then, based on the mapping information, the storage of the I / O destination is based on the information of the storage node corresponding to the global volume associated with the host volume designated as the I / O destination in the I / O request. A node is specified, and if the specified storage node is its own node, the volume number of the host volume of the I / O destination specified in the I / O request is rewritten to the volume number of the corresponding internal volume. I / O processing is executed for the internal volume, and when the specified storage node is another storage node, the I / O request is distributed to the storage node.

Further, in the present invention, in a system including a cluster composed of a plurality of storage nodes, the program is executed by the storage nodes, and each storage node is a storage device and a logical corresponding to the storage device. When a logical volume creation instruction based on a request from a management device is received, the storage node is associated with a host volume for reading and writing data provided to a higher-level device. The first step of creating a virtual global pool volume that can be identified in the own node is associated with the logical storage area in the own node and the global pool volume created in the own node. Then, the second step of creating a virtual internal volume that can be identified in the own node in the own node, and the correspondence between the created global pool volume and the node that created the global pool volume are as follows . When a third step of registering as pping information and synchronously updating the mapping information between the storage nodes in the cluster is provided, and each storage node receives an I / O request from the higher-level device. , Based on the mapping information, from the information of the storage node corresponding to the global volume associated with the host volume designated as the I / O destination in the I / O request, the storage node of the I / O destination. If the specified storage node is its own node, the volume number of the host volume of the I / O destination specified in the I / O request is rewritten to the volume number of the corresponding internal volume, and the said. I / O processing is executed for the internal volume, and when the specified storage node is another storage node, the I / O request is distributed to the storage node.

According to the system of the present invention, its control method, and the program, the host device can read and write desired data regardless of whether the storage node is scaled out or not without being aware of the storage node to which the I / O request is issued. It can be carried out.

According to the present invention, it is possible to realize a system capable of improving expandability, a control method thereof, and a program. Issues, configurations and effects other than those described above will be clarified by the following description of the embodiments for carrying out the invention.

It is a block diagram which shows the whole structure of the information processing system by this embodiment. It is a block diagram which shows the hardware configuration of a storage node. It is a block diagram which shows the software configuration of a storage node. It is a block diagram which provides the explanation of the correspondence relation of the storage area in a storage node. It is a figure which shows the configuration example of the storage control part pair table. Front-end table It is a figure which shows the configuration example of the volume table. It is a figure which shows the configuration example of the mapping table. It is a figure which shows the configuration example of the front-end target table. It is a figure which shows the configuration example of the storage control part information management table. It is a figure which shows the configuration example of a global pool volume table. It is a conceptual diagram provided to explain the correspondence relationship of information between each table. It is a sequence diagram which shows the flow of processing at the time of making an internal volume. It is a sequence diagram which shows the flow of a series of processing executed in the cluster when a failure occurs in the storage node in a cluster. It is a sequence diagram which shows the flow of a series of processing executed in a cluster when a storage node is added to the cluster. It is a sequence diagram which shows the flow of the write processing executed in this information processing system. It is a sequence diagram which shows the flow of the read processing executed in this information processing system.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. The following description and drawings are examples for explaining the present invention, and are appropriately omitted and simplified for the purpose of clarifying the description. Also, not all combinations of features described in the embodiments are essential to the means of solving the invention. The present invention is not limited to embodiments, and any application that fits the ideas of the present invention is included in the technical scope of the present invention. Those skilled in the art can make various additions and changes within the scope of the present invention. The present invention can also be implemented in various other forms. Unless otherwise specified, each component may be plural or singular.

In the following explanation, various information may be explained by expressions such as "table", "table", "list", and "queue", but various information may be expressed by data structures other than these. good. The "XX table", "XX list", etc. may be referred to as "XX information" to indicate that they do not depend on the data structure. When explaining the contents of each information, expressions such as "identification information", "identifier", "name", "ID", and "number" are used, but these can be replaced with each other.

Further, in the following description, the reference code or the common number in the reference code is used when the same type of element is not distinguished, and the reference code of the element is used when the same type of element is described separately. Alternatively, the ID assigned to the element may be used instead of the reference code.

Further, in the following description, a process performed by executing a program may be described, but the program is executed by at least one or more processors (for example, a CPU) to appropriately store a predetermined process. Since it is performed while using resources (for example, memory) and / or interface device (for example, communication port), the main body of processing may be a processor. Similarly, the subject of processing for executing a program may be a controller having a processor, a device, a system, a computer, a node, a storage system, a storage device, a server, a management computer, a client, or a host. The subject of processing performed by executing a program (for example, a processor) may include a hardware circuit that performs part or all of the processing. For example, the subject of processing performed by executing a program may include a hardware circuit that executes encryption and decryption, or compression and decompression. The processor operates as a functional unit that realizes a predetermined function by operating according to a program. A device and system including a processor is a device and system including these functional parts.

The program may be installed from the program source into a device such as a calculator. The program source may be, for example, a program distribution server or a storage medium readable by a computer. When the program source is a program distribution server, the program distribution server may include a processor (for example, a CPU) and a storage resource, and the storage resource may further store the distribution program and the program to be distributed. Then, when the processor of the program distribution server executes the distribution program, the processor of the program distribution server may distribute the program to be distributed to other computers. Further, in the following description, two or more programs may be realized as one program, or one program may be realized as two or more programs.

(1) Configuration of Information Processing System According to the Present Embodiment FIG. 1 is a diagram showing a configuration of an information processing system 1 according to the present embodiment. The information processing system 1 includes a plurality of compute nodes 2, a plurality of storage nodes 3, and a management node 4.

Between the compute node 2, the storage node 3, and the management node 4, for example, via a storage service network 5 composed of Fiber Channel (FC), Ethernet (registered trademark), InfiniBand, wireless LAN (Local Area Network), and the like. Is connected. Further, each storage node 3 is connected via a back-end network 6 composed of Ethernet (registered trademark), InfiniBand, wireless LAN, or the like.

However, the storage service network 5 and the back-end network 6 may be configured by the same network, and each compute node 2 and each storage node 3 are connected to a management network other than the storage service network 5 and the back-end network 6. It may have been done.

The compute node 2 is a general-purpose computer device that functions as a host (upper device) with respect to the storage node 3. The compute node 2 may be a virtual computer device such as a virtual machine. The compute node 2 makes a read request or a write request to the storage node 3 via the storage service network 5 in response to a user operation or a request from an implemented application program (hereinafter, as appropriate, these are collectively input / O (Input)). / Output) Send a request).

The storage node 3 is a general-purpose physical server device that provides a storage area for reading and writing data to the compute node 2. As shown in FIG. 2, the storage node 3 includes a CPU (Central Processing Unit) 10, a memory 11, a plurality of storage devices 12, and first and second communication devices 13 and 14, and includes the CPU 10 and the storage node 3. The storage device 12 and the first and second communication devices 13 and 14 are connected to each other via the internal network 15. Each storage node 3 includes one or more CPU 10, a memory 11, a storage device 12, and first and second communication devices 13 and 14, respectively.

The CPU 10 is a processor that controls the operation of the entire storage node 3. Further, the memory 11 is composed of a volatile semiconductor memory such as SRAM (Static RAM (Random Access Memory)) and DRAM (Dynamic RAM), and is used as a work memory of the CPU 10 to temporarily hold various programs and necessary data. It is used for. When at least one or more CPUs 10 execute the program stored in the memory 11, various processes as a whole of the storage node 3 as described later are executed.

The storage device 12 includes an NVMe (Non-Volatile Memory) drive, a SAS (Serial Attached SCSI (Small Computer System Interface)) drive, a SATA (Serial ATA (Advanced Technology Attachment)), an SSD (Solid State Drive), or an SCM (Storage Class). It is composed of a large-capacity non-volatile storage device such as Memory), and provides a storage area for reading and writing data to the compute node 2.

The first communication device 13 is an interface for the storage node 3 to communicate with the compute node 2 and the management node 4 via the storage service network 5, and is composed of, for example, an FC card or a wireless LAN card. The first communication device 13 controls the protocol at the time of communication with the compute node 2 and the management node.

The second communication device 14 is an interface for the storage node 3 to communicate with another storage node 3 via the back-end network 6, and is composed of, for example, a NIC (Network Interface Card) or a wireless LAN card. .. The second communication device 14 controls the protocol at the time of communication with the other storage node 3.

In the case of the present embodiment, as shown in FIG. 1, each storage node 3 is grouped together with the other one or a plurality of storage nodes 3 into a group called a cluster 7, and is managed in units of the cluster 7. In the example of FIG. 1, the case where only one cluster 7 is set is illustrated, but a plurality of clusters 7 may be set in the information processing system 1. Cluster 7 may be referred to as a distributed storage system.

The management node 4 is a computer device used by the administrator of the information processing system 1 (hereinafter, referred to as a system administrator) to perform configuration management and various settings of the information processing system 1. The management node 4 gives necessary instructions to the compute node 2 and the storage node 3 according to the operation of the system administrator.

FIG. 3 shows the logical configuration of the storage node 3 according to the present embodiment. As shown in FIG. 3, each storage node 3 includes a data plane 24 having a front end unit 20, one or a plurality of storage control units 21, a capacity control unit 22, and a back end unit 23, and a cluster control unit 25 and a node. It is configured to include a control plane 27 having a control unit 26. The data plane 24 is a functional unit that executes I / O processing related to reading and writing data, and the control plane 27 is a functional unit that controls the entire cluster 7 (FIG.) and its own node.

The front end unit 20 is software that functions as a front end for I / O processing related to scale-out in the storage node 3. For example, when an I / O request is given from the compute node 2, the front-end unit 20 sends the I / O request to the storage control unit 21 in its own node that should execute the I / O request, or the storage control unit 21. The storage control unit 21 that should execute the I / O request is distributed to another storage node 3 in which the storage control unit 21 is arranged.

The storage control unit 21 is software that functions as a controller for SDS (Software Defined Storage). The storage control unit 21 may be called a storage control software or a storage control program. The storage control unit 21 receives the I / O request from the compute node 2 given from the front end unit 20, and issues an I / O command corresponding to the received I / O request to the back end unit 23.

Here, in the case of the present embodiment, each storage control unit 21 mounted on the storage node 3 is set as a pair forming a redundant configuration together with another storage control unit 21 arranged on another storage node 3. NS. In the following, this pair will be referred to as a storage control unit pair 28.

Note that FIG. 3 shows a case where one storage control unit pair 28 is configured by the two storage control units 21, and also in the following, the storage control unit pair 28 is configured by the two storage control units 21. However, one redundant configuration may be formed by three or more storage control units 21.

In the storage control unit pair 28, one storage control unit 21 is set to a state in which it can accept an I / O request from the compute node 2 (a state of the active system, hereinafter referred to as an active mode). The other storage control unit 21 is set to a state in which the I / O request from the compute node 2 is not accepted (a state of a standby system, hereinafter referred to as a passive mode).

Then, in the storage control unit pair 28, a failure occurs in the storage control unit (hereinafter, appropriately referred to as an active storage control unit) 21 set in the active mode and the storage node 3 in which the active storage control unit 21 is arranged. In such a case, the state of the storage control unit (hereinafter, appropriately referred to as a passive storage control unit) 21 which has been set to the passive mode up to that point is switched to the active mode. As a result, when the active storage control unit 21 cannot operate, the passive storage control unit 21 constituting the same storage control unit pair 28 takes over the I / O processing executed by the active storage control unit 21. be able to.

The capacity control unit 22 provides a physical storage device 12 (FIG. 2) in the own node or another node with respect to the storage control unit pair 28 configured by the active storage control unit 21 arranged in the own node. Software that has the function of allocating various storage areas. The capacity control unit 22 may be called a capacity control software or a capacity control program.

The backend unit 23 is software that functions as a backend for I / O processing in the storage node 3. For example, the back-end unit 23 has a storage area allocated by the capacity control unit 22 to the storage control unit pair 28 configured by the active storage control unit 21 according to the above-mentioned I / O command given from the active storage control unit 21. Read and write data to.

On the other hand, the cluster control unit 25 is software having a function of executing control processing related to the entire cluster 7 and control processing related to scale-out of the cluster 7. The cluster control unit 25 may be called a cluster control software or a cluster control program. In the information processing system 1, one of the cluster control units 25 mounted on each storage node 3 in the cluster 7 is set to the master, and the cluster control unit 25 set to the master (hereinafter, hereinafter, Only 25 (which is called a master cluster control unit) executes various control processes while maintaining the consistency of the entire cluster 7.

For example, the master cluster control unit 25 sets a storage control unit pair 28 in the cluster 7 in response to a request from the management node 4, and sets the storage control unit pair 28 in the storage control unit pair table 30 which will be described later with reference to FIG. Register and manage.

Further, the cluster control unit 25 other than the master cluster control unit 25 has an operation mode of either a hot standby (Hot stand-by) mode or a worm standby (Warm stand-by) mode in preparation for a failure of the master cluster control unit 25. Is set to.

In the hot standby mode, when a failure occurs in the master cluster control unit 25 or the storage node 3 on which the master cluster control unit 25 is mounted, the processing previously executed by the master cluster control unit 25 can be immediately taken over. This is the operation mode that maintains the activated standby state.

The cluster control unit 25 in the hot standby mode manages all the processes managed by the master cluster control unit 25 such as the storage control unit pair table 30 (FIG. 5) described above so that the processing executed by the master cluster control unit 25 can be immediately taken over. It holds the same management information as the management information.

When the management information held by the master cluster control unit 25 is updated, the difference before and after the update is used as difference data for cluster control in all hot standby modes from the master cluster control unit 25 via the back-end network 6. Based on the difference data given to the unit 25, the management information held by the cluster control unit 25 is updated by the cluster control unit 25 in the same manner as the management information held by the master cluster control unit 25.

In this way, the cluster control unit 25 in the hot standby mode always holds the same management information as the master cluster control unit 25, so that a failure occurs in the master cluster control unit 25 and the like, and the cluster that was in the hot standby mode until then. Even when the control unit 25 is switched to the "master", the control processing previously executed by the original master cluster control unit 25 is taken over by the cluster control unit (master cluster control unit) 25 that has been switched to the "master". It becomes possible.

The worm standby mode is an operation mode in which the startup is stopped and is in a standby state. When the number of cluster control units set in the hot standby mode becomes equal to or less than a preset threshold value, the state of any of the cluster control units 25 set in the worm standby mode is switched to the hot standby mode.

In addition, the master cluster control unit 25 operates three or more cluster control units 25 in order to prevent a state in which two or more master cluster control units 25 exist, and a majority vote is made from among these operated cluster control units 25. Is selected by. Then, the remaining cluster control unit 25 that has been operated is set to the hot standby mode.

The node control unit 26 is a program having a function of executing various control processes completed within the own node in response to a request from the master cluster control unit 25. The node control unit 26 may be called a node control software or a node control program. In practice, the master cluster control unit 25 requests the node control unit 26 of each storage node 3 to execute the processing completed in each storage node 3 so that the load is not concentrated on itself. Then, when such a request is given, the node control unit 26 executes control processing for the data plane 24 of its own node in accordance with this request.

As described above, in the information processing system 1, commands and requests from the master cluster control unit 25 to the front end unit 20 and the storage control unit 21 in each storage node 3 are given via the node control unit 26 of the storage node 3. .. However, in the following, for ease of understanding, commands and requests from the master cluster control unit 25 to the front end unit 20 and the storage control unit 21 in each storage node 3 are sent from the master cluster control unit 25 to the front end. It will be described as being directly given to the unit 20 and the storage control unit 21.

FIG. 4 shows the correspondence between the storage areas in the information processing system 1 of the present embodiment. As shown in FIG. 4, in the present embodiment, the physical storage area provided by the storage device 12 in each storage node 3 by the capacity control unit 22 is a storage area of a predetermined size (hereinafter, this). Is called a physical chunk) It is divided into PCs and managed.

Further, the capacity control unit 22 sets a plurality of physical chunk PCs provided by the storage devices 12 in different storage nodes 3 into a logical storage area having the same size as the physical chunk PC (hereinafter, this is referred to as a logical chunk). It is associated with LC, and this logical chunk LC is associated with a pool volume PLVOL having the same size as the logical chunk LC via an external volume ExtVOL.

Further, the capacity control unit 22 creates a pool PL by collecting a plurality of pool volume PLVOLs associated with the logical chunk LC in this way, and the active storage control unit 21 in which the created pool PL is arranged on the own node belongs to. It is assigned to the storage control unit pair 28 (FIG. 3). In this way, the capacity control unit 22 allocates a pool PL, which is a dedicated storage area, to each storage control unit pair 28.

Further, on these pool PLs, one or a plurality of virtual logical volumes, the internal volume IVOL, are created according to the instruction from the system administrator via the management node 4. These internal volume IVOLs are defined in the storage node 3 in which the active storage control unit 21 of the two storage control units 21 constituting the storage control unit pair 28 to which the pool PL is associated is arranged.

Further, each internal volume IVOL is a virtual defined in the active storage control unit 21 that constitutes the corresponding storage control unit pair 28 (the storage control unit pair 28 associated with the internal volume IVOL via the pool PL). Active storage control unit 21 and the virtual port VPT defined in the passive storage control unit 21 constituting the storage control unit pair 28, respectively, via a typical port (hereinafter referred to as a virtual port) VPT. It is associated with the global pool volume GVOL, which is a virtual logical volume defined in each storage node 3 in which the passive storage control unit 21 is arranged, on a one-to-one basis.

Then, each global pool volume GVOL in the cluster 7 is collectively managed as one pool (hereinafter, referred to as a global pool) GPL that crosses all the storage nodes 3 in the cluster 7.

Further, each global pool volume GVOL is each storage in which the active storage control unit 21 and the passive storage control unit 21 that constitute the storage control unit pair 28 associated with the global pool volume GVOL via the internal volume IVOL are arranged. It is associated with the host volume HVOL, which is a virtual logical volume defined in each node 3, in a one-to-N manner. This host volume HVOL is a volume provided to the compute node as a storage area for reading and writing data.

Further, each host volume HVOL is located in a storage node 3 in which the active storage control unit 21 of the storage control unit pair 28 corresponding to the internal volume IVOL to which the host volume HVOL is associated via the global pool volume GVOL is arranged. It is associated one-to-one with the physical port PPT of the storage node 3 via the defined target TG.

The global pool volume GVOL is created by assigning a unique identification number in the cluster 7 in response to a creation instruction from the system administrator via the management node 4. At this time, together with the global pool volume GVOL, an internal volume IVOL associated with the global pool volume GVOL on a one-to-one basis is created in association with the designated storage control unit pair 28.

Further, when the system administrator creates a target TG in the storage node 3 via the management node 4, the global pool volume GVOL corresponds to the target TG by registering the global pool volume GVOL in the target TG. Associated with the attached host volume HVOL.

Incidentally, the intention of the global pool volume GVOL is to not duplicate the identification number of the internal volume IVOL associated with the host volume HVOL in the cluster 7. In fact, in the case of the present embodiment, the IVOL number of the internal volume IVOL associated with each storage control unit pair 28 is unique in the storage node 3, but not unique in the cluster 7. Therefore, in the present embodiment, the correspondence between the host volume HVOL and the internal volume IVOL is established by interposing the global pool volume GVOL having a unique GVOL number in the cluster 7 between the internal volume IVOL and the host volume HVOL. It is designed to be uniquely identifiable.

On the other hand, on the compute node 2 side, a plurality of storage nodes 3 constituting the cluster 7 are seen as one storage device, and the host volume HVOL is recognized as a storage area provided by the storage device to its own node.

However, the host volume HVOL that can be recognized by each compute node 2 is only the host volume HVOL that is set in the compute node 2 by the system administrator via the management node 4. Therefore, the compute node 2 recognizes only the host volume HVOL set by the management node 4 as the storage area provided by the storage device.

Further, each compute node 2 manages a path from the physical port (not shown) of its own node to the target TG defined in any storage node 3 in the cluster 7. When the host computer 2 reads / writes data to a desired host volume HVOL, the host computer 2 targets the read / write destination host volume HVOL and the read / write destination storage area in the host volume HVOL. An I / O request specifying the data length of the data is transmitted to the target TG recognized by the own node.

At this time, in each storage node 3, the front-end unit 20 (FIG. 3) has a front-end table volume table 31, which will be described later in FIG. 6, a mapping table 32, which will be described later in FIG. 7, and a global pool volume, which will be described later in FIG. Table 35 is used to manage the correspondence between the host volume HVOL, the global pool volume GVOL, and the internal volume IVOL, and the storage control unit 21 manages the correspondence from the internal volume IVOL to the logical chunk LC using a management table (not shown). The capacity control unit 22 (FIG. 3) manages the correspondence between the logical chunk LC and the physical chunk PC by using a management table (not shown).

Then, when the front end unit 20 receives the I / O request from the compute node 2, the front end unit 20 refers to the mapping table 32 and the global pool volume table 35 and sets the host volume HVOL specified in the I / O request. The storage node 3 in which the active storage control unit 21 associated with the global pool volume GVOL and the internal volume IVOL is arranged is specified.

Then, when the specified storage node 3 is its own node, the front end unit 20 sets the volume number of the read / write destination host volume HVOL included in the I / O request to the host volume HVOL and the global pool volume. After rewriting the volume number of the internal volume IVOL associated via the GVOL, the I / O request is transferred to the active storage control unit 21 in the own node.

Further, when the front-end unit 20 determines that the specified storage node 3 is another storage node 3, the I / O request to the storage node 3 should be executed by the other storage node 3. Transfers the I / O request to the specified storage node 3 via the back-end network 6. Thus, the front end portion 20 of the storage node 3 to which the I / O request has been transferred sets the volume number of the read / write destination host volume HVOL included in the I / O request to the host volume HVOL and the global pool volume. After rewriting the volume number of the internal volume IVOL associated via the GVOL, the I / O request is transferred to the corresponding active storage control unit 21 in the own node.

Then, the active storage control unit 21 that has received this I / O request sequentially responds to the internal volume IVOL specified in the I / O request, the storage area in the internal volume IVOL, the pool PL, and the external volume ExtVOL. The attached logical chunk LC and the storage area in the logical chunk LC are specified by using the management table (not shown above).

Thus, in the storage node 3, when the I / O request is a write request, the back-end unit 23 stores data in the corresponding storage area of all the physical chunk PCs associated with the logical chunk LC. Are each lit. When the I / O request is a read request, the back-end unit 23 reads data from one of the physical chunk PCs associated with the logical chunk LC, which is It is transferred to the compute node 2 that is the source of the read request.

In the case of the present embodiment, as described above, the data from the compute node 2 is a logical chunk corresponding to the host volume HVOL specified in the I / O request, the global pool volume GVOL, the internal volume IVOL, and the like in sequence. It is redundantly stored in a plurality of physical chunk PCs associated with the LC. Therefore, the number of physical chunk PCs associated with each logical chunk LC by the capacity control unit 22 is determined by the setting content of the redundancy method in the information processing system 1.

For example, in the case of setting to duplicate and store data, two physical chunk PCs are associated with one logical chunk LC, and the setting is made to multiplex and store data more than triplet. For example, three or more required number of physical chunk PCs are associated with one logical chunk LC.

In this case, if the physical chunk PC associated with the logical chunk LC is selected from the physical chunk PCs in the storage node 3 different from the storage node 3 in which the active storage control unit 21 is arranged, the active storage control thereof. When the capacity control unit 22 that has received the I / O command from the unit 21 reads / writes data to the physical chunk PC, communication with the storage node 3 that provides the physical chunk PC is required, and the communication is required. Therefore, the response performance of the entire system deteriorates. Therefore, when associating a plurality of physical chunk PCs with the logical chunk LC, one of the physical chunk PCs is provided by the storage device 12 in the storage node 3 in which the active storage control unit 21 is arranged. It is desirable to select from these from the viewpoint of the response performance of the entire system.

Further, considering that the passive storage control unit 21 is switched to the active mode when a failure occurs in the storage node 3 in which the active storage control unit 21 is arranged in the storage control unit pair 28, for the same reason as described above. , One of the physical chunk PCs associated with the logical chunk LC should be selected from the physical chunk PCs provided by the storage device 12 in the storage node 3 in which the passive storage control unit 21 is arranged. It is also desirable from the viewpoint of performance.

Therefore, in the information processing system 1, the capacity control unit 22 of each storage node 3 assigns a logical chunk LC to the storage control unit pair 28, and when associating a plurality of physical chunk PCs with the logical chunk LC, the storage control unit A physical chunk PC provided by the storage device 12 in the storage node 3 in which the active storage control unit 21 of the pair 28 is arranged, and a storage in the storage node 3 in which the passive storage control unit 21 of the storage control unit pair 28 is arranged. The physical chunk PC provided by the device 12 is preferentially associated with the logical chunk LC.

(2) Configuration of various tables FIGS. 5 to 10 are for managing the configuration of each storage control unit pair 28 and the correspondence between the host volume HVOL, the global pool volume GVOL, and the internal volume IVOL as described above. A configuration example of various tables included in a database (not shown) held in the memory 11 of each storage node 2 is shown.

Among these tables, the storage control unit pair table 30 shown in FIG. 5, the front-end table volume table 31 shown in FIG. 6, the mapping table 32 shown in FIG. 7, the front-end target table 33 shown in FIG. 8, and FIG. The indicated global pool volume 35 is managed by the front end unit 20. Further, the storage control unit configuration information table 34 shown in FIG. 9 is managed by the storage control unit 21.

Further, the storage control unit pair table 30, the front-end table volume table 31, the mapping table 32, the front-end target table 33, and the global pool volume table 35 are updated when any of the storage nodes 3 in the cluster 7 is updated. It is similarly updated in all the other storage nodes 3 in synchronization, so that the contents are always kept the same among the storage nodes 3.

The storage control unit pair table 30 is a table used by the master cluster control unit 25 (FIG. 4) to manage the storage control unit pair 28 (FIG. 3) set in the cluster 7, and is as shown in FIG. In addition, a pair number column 30A, an active column 30B, and a passive column 30C are provided. In the storage control unit pair table 30, one row corresponds to one storage control unit pair 28, and all the storage control unit pairs 28 in the cluster 7 are registered in the storage control unit pair table 30.

Then, in the pair number column 30A, an identification number (pair number) unique to the storage control unit pair 28 assigned to the corresponding storage control unit pair 28 is stored.

Further, the active column 30B is divided into a node number column 30BA and a storage control unit column 30BB. Then, in the node number column 30BA, the identification number (node number) of the storage node 3 in which the active storage control unit 21 of the two storage control units 21 constituting the corresponding storage control unit pair 28 is arranged is stored. , The identification number (storage control unit number) of the active storage control unit 21 is stored in the storage control unit number column 30BB.

Similarly, the passive column 30C is divided into a node number column 30CA and a storage control unit number column 30CB. Then, in the node number column 30CA, the node number of the storage node 3 in which the passive storage control unit 21 of the two storage control units 21 constituting the corresponding storage control unit pair 28 is arranged is stored, and the storage control unit is stored. The storage control unit number of the passive storage control unit 21 is stored in the number column 30CB.

Therefore, in the case of the example of FIG. 5, the storage control unit pair 28 to which the pair number "1" is assigned is the active storage control unit number "0" arranged in the storage node 3 having the node number "0". It is shown that the storage control unit 21 is composed of a storage control unit 21 and a passive storage control unit 21 having a storage control unit number “1” arranged in the storage node 3 having a node number “1”.

On the other hand, the front-end table volume table 31 is a table used for managing the correspondence between the host volume HVOL and the internal volume IVOL. Front-end table As shown in FIG. 6, the volume table 31 has a host volume ID column 31A, a UUID column 31B, a first node specific information column 31C, a second node specific information column 31D, and an active node specific information column 31E. , Storage control unit number field 31F, virtual port number field 31G, and internal volume number field 31H. In this front-end table volume table 31, one row corresponds to one host volume HVOL, and all host volume HVOLs in the cluster 7 are registered.

Then, in the volume number column 31A, a volume number unique to the host volume HVOL assigned to the corresponding host volume HVOL is stored. This volume number is an identification number (volume number) of the host volume HVOL recognized by the compute node 2.

Further, the UUID (Universally Unique Identifier) unique to the host volume HVOL assigned to the corresponding host volume HVOL in the cluster 7 is stored in the UUID column 31B, and the corresponding host is stored in the internal volume number column 31H. The volume number of the internal volume IVOL associated with the volume HVOL is stored.

Further, in the first node identification information column 31C, information for identifying the storage node 3 in which the active storage control unit 21 of the storage control unit pair 28 associated with the corresponding internal volume IVOL is arranged (hereinafter, this). Is called storage node specific information), and the storage node specific information of the storage node 3 in which the passive storage control unit 21 of the storage control unit pair 28 is arranged is stored in the second node specific information field 31D. NS. In the present embodiment, the IP (Internet Protocol) address on the storage service network 5 of the corresponding storage node 3 is applied as the storage node specific information.

Further, the storage node identification information of the storage node 3 in which the active storage control unit 21 is arranged is stored in the active side node identification information column 31E. Further, the storage control unit number of the active storage control unit 21 is stored in the storage control unit number column 31F, and the virtual port number column 31G corresponds to the virtual port VPT defined in the active storage control unit 21. The identification number (port number) of the virtual port VPT to which the host volume HVOL and the corresponding internal volume IVOL are connected is stored.

Therefore, in the case of the example of FIG. 6, the active storage control unit 21 having the storage control unit number "1" arranged in the storage node 3 of the node specific information "sn1" and the storage node of the node specific information "sn2" An internal volume IVOL with a volume number of "1" is defined on the pool PL assigned to the storage control unit pair 28 composed of the passive storage control unit 21 arranged in 3, and this internal volume IVOL is its active. It is shown that the storage control unit 21 is associated with the host volume HVOL having the volume number "1" via the virtual port VPT having the port number "Cl1-a". FIG. 6 also shows that the UUID of the host volume HVOL in cluster 7 is "Xxxxxxxxxxx".

On the other hand, the mapping table 32 is a table used for managing the correspondence between the global pool volume GVOL and the internal volume IVOL. As shown in FIG. 7, the mapping table 32 includes a global pool volume number column 32A, a UUID column 32B, an internal volume number column 32C, a node number column 32D, and a storage control unit number column 32E. In the mapping table 32, one row corresponds to one global pool volume GVOL, and all global pool volume GVOLs in the cluster 7 are registered.

The volume number of the corresponding global pool volume GVOL is stored in the global pool volume number column 32A, and the UUID in the cluster 7 of the global pool volume GVOL is stored in the UUID column 32B.

Further, in the internal volume number column 32C, the volume number of the internal volume IVOL assigned to the internal volume IVOL associated with the global pool volume GVOL is stored. Further, the node number column 32D stores the node number of the storage node 3 in which the internal volume IVOL exists (usually, the storage node 3 in which the corresponding active storage control unit 21 is arranged).

Further, the storage control unit number column 32E constitutes a storage control unit pair 28 (a storage control unit pair 28 in which the internal volume IVOL is defined on the allocated pool PL) associated with the internal volume IVOL. Of the two storage control units 21, the storage control unit number of the active storage control unit 21 is stored.

Therefore, in the case of the example of FIG. 7, the global pool volume GVOL having the volume number "1" is given the UUID "Xxxxxxxxxxxx" in the cluster 7, and the active storage control unit 21 having the storage control unit number "1" is assigned. It is shown that it is associated with the internal volume IVOL of the volume number "1" defined in the storage node 3 of the arranged node number "2".

The front-end target table 33 is used to manage the correspondence between the target TG (FIG. 4) set 1: 1 for each physical port PPT (FIG. 4) of the storage node 3 and the host volume HVOL. This is the table used. As shown in FIG. 8, the front-end target table 33 includes a target number column 33A, a target name column 33B, a UUID column 33C, a target IP column 33D, a host volume number list column 33E, and an initiator name column 33F. .. In this front-end target table 33, one row corresponds to one target TG, and all target TGs defined in the cluster 7 are registered.

Then, the target number column 33A stores the identification number (target number) unique to the target TG assigned to the corresponding target TG, and the target name column 33B stores the name (target) assigned to the corresponding target TG. First name) is stored. This target name is a name given by the user or the management node 4.

Further, the UUID column 33C stores the UUID of the target TG in the cluster 7 assigned to the corresponding target TG. Further, in the target IP column 33D, the IP address on the storage service network 5 of the physical port PPT in which the target TG is set is stored.

Further, the volume numbers of all the host volumes HVOL associated with the corresponding target TG are stored in the host volume number list column 33E, and the initiator name column 33F is the compute node 2 capable of logging in to the corresponding target TG. The name (initiator name) is stored.

Therefore, in the case of the example of FIG. 8, the target TG with the target name "AAAA" given the target number "1" is given the UUID "Xxxxxxxxxxxx" in the cluster 7, and the target TG is set. The IP address on the storage service network 5 of the physical port PPT is "xx.xx.xx.xx", and the target TGs are assigned volume numbers "1", "2", ... "N", respectively. It is shown that a plurality of host volumes HVOL are associated with each other, and the compute node 2 having an initiator name of "lqn.xxxx.xxx" is set to be able to log in to the target TG.

The storage control unit configuration information table 34 is a table used for managing the correspondence between the virtual port VPT and the internal volume IVOL, and as shown in FIG. 9, the virtual port number column 34A and the internal volume number column 34B are displayed. Be prepared. The storage control unit configuration information table 34 is created for each storage control unit 21 and is managed by the corresponding storage control unit 21.

The virtual port number field 34A stores the port number of the virtual port VPT defined in the corresponding storage control unit 21, and the internal volume number field 34B stores the volume of the internal volume IVOL connected to the virtual port VPT. The number is stored.

Therefore, in the case of the example of FIG. 9, it is shown that the internal volume IVOL of the volume number “1” is connected to the virtual port VPT “Cl1-a” of the corresponding storage control unit 21.

The global pool volume table 35 is a table used for managing the global pool volume GVOL defined in the cluster 7, and as shown in FIG. 10, the global pool volume number column 35A, the target number column 35B, and the host It is configured to include a volume number field 35C. In this global pool volume table 35, one row corresponds to one global pool volume GVOL, and all global pool volume GVOLs defined in the cluster 7 are registered.

The global pool volume number column 35A stores an identification number (volume number) unique to the global pool volume GVOL assigned to the corresponding global pool volume GVOL, and the host volume number column 35C stores the corresponding global pool. The volume numbers of all host volumes HVOL associated with the volume GVOL are stored. Further, in the target number column 35B, the target number of the target TG associated with the corresponding host volume HVOL is stored.

Therefore, in the case of FIG. 10, the global target volume GVOL assigned the volume number "1" is associated with the host volume HVOL assigned at least the volume number "1", and the host volume HVOL is associated with the host volume HVOL. It is shown that the target TG to which the target number "1" is assigned is associated with the target TG to which the target number "2" is assigned (see FIG. 4).

FIG. 11 shows the correspondence between the front-end table volume table 31, the mapping table 32, the front-end target table 33, the storage control unit configuration information table 34, and the global pool volume 35.

As shown in FIG. 11, the virtual port numbers and volume numbers stored in the virtual port number column 31G (FIG. 6) and the internal volume number column 31H (FIG. 6) of the front-end table volume table 31, respectively, are virtual. Virtual port number column 34A (FIG. 9) and internal volume number column 34B (FIG. 9) of the storage control unit configuration information management table 34 managed by the storage control unit 21 having a virtual port VPT (FIG. 4) to which a port number is assigned. Corresponds to the virtual port number and volume number stored in each.

Therefore, between the front-end table volume table 31 and the storage control unit configuration information management table 34, the combination of the port number of the virtual port VPT and the volume number of the internal volume IVOL associated with the virtual port VPT is used as a key. It is possible to recognize the correspondence between rows.

The volume number stored in the global pool volume number column 32A (FIG. 7) of the mapping table 32 corresponds to the volume number stored in the global pool volume number column 35A of the global pool volume table 35. Therefore, between the mapping table 32 and the global pool volume table 35, the correspondence relationship between the rows can be recognized by using the volume number of the global pool volume GVOL as a key.

Further, the volume number stored in the internal volume number column 32C (FIG. 7) of the mapping table 32 is the storage control unit configuration information managed by the active storage control unit 21 that provides the internal volume IVOL (FIG. 4) of the internal volume number. It corresponds to the volume number stored in the internal volume number column 34B (FIG. 9) of the management table 34. Therefore, between the mapping table 33 and the storage control unit configuration information management table 34, the correspondence relationship between the rows can be recognized by using the volume number of the internal volume IVOL as a key.

Further, the target number stored in the target number column 35B (FIG. 10) of the global pool volume table 35 corresponds to the target number stored in the target number column 33A of the front-end target table 33 (FIG. 8). Therefore, between the global pool volume table 35 and the front-end target table 33, the correspondence between the rows can be recognized by using the target number as a key.

(3) Flow of creating an internal volume Next, regarding the flow of a series of processes for creating the above-mentioned internal volume IVOL in the cluster 7 in FIG. 4, associating it with the host volume HVOL, and making various settings related to the internal volume IVOL. This will be described with reference to FIG.

In the information processing system 1, when the system administrator creates a new host volume HVOL to be provided to the compute node 2, the system administrator operates the management node 4 to associate the global pool with the host volume HVOL to be created at that time. A volume GVOL creation request (hereinafter, a global pool volume creation request) is transmitted to the storage node 3 in which the master cluster control unit 25 is arranged (S1).

Upon receiving this global pool volume creation request, the master cluster control unit 25 first analyzes the received command (global pool volume creation request) (S2), and recognizes that this command is a global pool volume creation request. The optimum storage control unit pair 28 (FIG. 3) for associating the global pool volume GVOL to be created at that time is selected (S3). For example, the master cluster control unit 25 selects the storage control unit pair 28 in which the active storage control unit 21 is arranged on the storage node 3 having the lowest load in this step S3.

Further, the master cluster control unit 25 creates a global pool volume GVOL for the storage node 3 in which the active storage control unit 21 is arranged among the storage control units 21 constituting the storage control unit pair 28 selected in step S3. An instruction to the effect (hereinafter referred to as a global pool volume creation instruction) and an instruction to create an internal volume IVOL (hereinafter referred to as an internal volume creation instruction) are given via the back-end network 6. (S4A, S4B).

Thus, the front end unit 20 of the storage node 3 that has received the global pool volume creation instruction creates the global pool volume GVOL in its own node (S5A). Further, the active storage control unit 21 of the storage node 3 is assigned to the pool PL (FIG. 4) in the own node to the storage control unit pair 28 configured by the local storage control unit 21 according to the internal volume creation instruction (FIG. 4). And the internal volume IVOL are created in association with the global pool volume GVOL created in step S5A (S5B).

Since the global pool volume GVOL and the internal volume IVOL are virtual logical volumes that have no substance, the information of the global pool volume GVOL and the internal volume IVOL is used in the global pool volume table 35 (FIG. 10). This is performed by registering in the corresponding storage control unit configuration information table 34 (FIG. 9).

In practice, the front-end unit 20 adds a new row to the global pool volume table 35 (FIG. 10) that it manages, and the global pool volume in the global pool volume number column 35A (FIG. 10) of that row. Stores the volume number assigned to GVOL.

Then, the front-end unit 20 subsequently sets the volume number of the global pool volume GVOL created in this way and the volume number of the internal volume IVOL associated with the global pool volume GVOL to the back-end network 6. Notify the master cluster control unit 25 via (S6A).

Further, the active storage control unit 21 adds a new row to the storage control unit configuration information table 34 (FIG. 9) managed by itself, and creates a new row in the internal volume number column 34B (FIG. 9) of that row at that time. The volume number of the internal volume IVOL to be to be stored is stored, and the port number of the virtual port VPT (FIG. 4) of the own storage control unit 21 associated with the internal volume IVOL is entered in the virtual port number column 34A (FIG. 9) of that line. Store.

Then, the active storage control unit 21 then backends a completion notification including the volume number of the internal volume IVOL created in this way and the port number of the virtual port VPT associated with the internal volume IVOL. Notify the master cluster control unit 25 via the network 6 (S6B).

On the other hand, the master cluster control unit 25 that has received these completion notifications sets the global pool volume GVOL and its internal volume IVOL in the front-end table volume table 31 (FIG. 6), mapping table 32 (FIG. 7), and mapping table 32 (FIG. 7) in the own node. Instruct the front end unit 20 in the own node to register in the global pool volume table 35 (FIG. 10). Thus, the front-end unit 20 that receives this instruction displays the front-end table volume table 31, the mapping table 32, and the global pool volume table 35 (FIG. 10) stored in the memory 11 (FIG. 2) of the own node. Register the global pool volume GVOL and its internal volume IVOL (S7).

Specifically, the front-end unit 20 adds a new row to the front-end table volume table 31 stored in the memory 11 (FIG. 2) of the own node, and the first node specific information column 31C of that row. FIG. 6 stores the storage node identification information of the storage node 3 in which one of the storage control units 21 constituting the storage control unit pair 28 selected in step S3 is arranged. Further, the front end unit 20 is a storage in which the other storage control unit 21 constituting the storage control unit pair 28 selected in step S3 is arranged in the second node specific information field 31D (FIG. 6) of the row. Storage of node 3 Stores node specific information.

Further, in the front end unit 20, the active storage control unit 21 of the two storage control units 21 constituting the storage control unit pair 28 is arranged in the active side node identification information column 31E (FIG. 6) of the row. The storage node specific information of the storage node 3 is stored, and the storage control unit number of the active storage control unit 21 is stored in the storage control unit number column 31F (FIG. 6) of the row.

Further, the front end unit 20 stores the port number of the virtual port VPT included in the completion notification transmitted from the corresponding storage control unit 21 received in step S7 in the virtual port number column 31G (FIG. 6) of the line. At the same time, the volume number of the internal volume included in the completion notification is stored in the internal volume number column 31H (FIG. 6) of the line.

Similarly, the front end unit 20 adds a new row to the mapping table 32 (FIG. 7) and fills the global pool volume number column 32A (FIG. 7) and UUID column 32B (FIG. 7) of that row, respectively. The volume number and UUID of the created global pool volume GVOL are stored, and the volume number of the internal volume IVOL associated with the global pool volume GVOL is stored in the internal volume number column 32C (FIG. 7) of the line.

Further, the front end unit 20 stores the node number of the storage node 3 in which the active storage control unit 21 of the storage control unit pair 28 selected in step S3 is arranged in the node number column 32D (FIG. 7) of the row. At the same time, the storage control unit number of the active storage control unit 21 is stored in the storage control unit number column 32E (FIG. 7) of that line.

Further, the front end unit 20 adds a new row to the global pool volume table 35 (FIG. 10), and in the global pool volume number column 35A (FIG. 10) of that row, the volume number of the global pool volume GVOL created at that time. To store.

Then, in the master cluster control unit 25, as described above, the front end unit 20 in the own node sets the global pool volume GVOL and the internal volume IVOL in the front end table volume table 31, the mapping table 32 and the global pool volume in the own node. When the registration in the table 35 is completed, the differences before and after the update of the front-end table volume table 31, the mapping table 32, and the global pool volume table 35 are used as difference data, except for the local node in the cluster 7 via the back-end network 6. It is transmitted to each storage node 3 of (S8).

Thus, the front-end unit 20 of each storage node 3 that has received the difference data sets the front-end table volume table 31, the mapping table 32, and the global pool volume table 35 in the own node as the master cluster based on the difference data. The front-end table volume table 31, the mapping table 32, and the global pool volume table 35 in the storage node 3 in which the control unit 25 is arranged are updated in the same manner (S9). In this way, the front-end table volume table 31, mapping table 32, and global pool volume table 35 in each storage node 3 are updated synchronously, and these front-end table volume table 31, mapping table 32, and global pool volume table are updated. The contents of 35 are always maintained in the same state among the storage nodes 3 in the cluster 7.

Further, when these front-end units 20 finish updating the front-end table volume table 31, the mapping table 32, and the global pool volume table 35 in the own node in this way, the back-end network 6 notifies the completion to that effect. Is transmitted to the master cluster control unit 25 via (S10).

Then, when the front end unit 20 of all the storage nodes 3 other than the own node in the cluster 7 gives the completion notification of step S10 described above, the master cluster control unit 25 finishes creating the requested global pool volume GVOL. A completion notification to that effect is transmitted to the management node 4 (S11).

On the other hand, when the completion notification of step S11 is given from the master cluster control unit 25, the management node 4 requests to create a target TG (FIG. 4) associated with the host volume HVOL to be created at that time (hereinafter, , This is called a target creation request) is transmitted to the master cluster control unit 25 via the storage service network 5 (S12). The target creation request includes the target name of the target TG to be created at that time, the volume number of the host volume HVOL associated with the target TG, and the IP address of the target TG.

When the master cluster control unit 25 receives this target creation request, the master cluster control unit 25 receives the target creation request, and the target requested in the two storage nodes 3 in which each storage control unit 21 constituting the storage control unit pair 28 selected in step S3 is arranged. Instruct the front end unit 20 in the own node to create each TG (S13). However, since these target TGs are virtual ones having no substance, the creation of these target TGs is performed by newly registering the information of these target TGs in the front-end target table 33 (FIG. 8). ..

In practice, the front-end unit 20 that receives such an instruction from the master cluster control unit 28 adds two new rows to the front-end target table 33 (FIG. 8) stored in the memory 11 of the own node. The target name column 33B (FIG. 8), the target IP column 33D (FIG. 8), and the host volume number column 33E (FIG. 8) of these lines are each specified in the target creation request given by the management node 4 in step S12. The corresponding information of the target name and network IP of the new target TG and the volume number of the host volume HVOL is stored respectively. Further, the front end unit 20 stores the UUID unique to the cluster 7 assigned to the corresponding target TG in the UUID column 33C (FIG. 8) of these rows.

After that, the master cluster control unit 28 in the own node so as to map the global pool volume GVOL created in step S5A, the internal volume IVOL created in step S5B, and the host volume HVOL to be created at that time. An instruction is given to the front end portion 20 of the above (S14).

Thus, the front-end unit 20 that receives this instruction requests the target creation in the volume number column 31A (FIG. 6) of the row newly added to the front-end table volume table 31 (FIG. 6) in the own node in step S7. Stores the volume number of the host volume HVOL to be created at that time specified in. Further, the front end unit 20 assigns a unique UUID in the cluster 7 to the host volume HVOL, and stores the assigned UUID in the UUID column 31B (FIG. 6) of the row.

Further, the front end unit 20 is designated in the target creation request in the target number column 35B of the row newly added to the global pool volume table 35 (FIG. 10) in the own node in step S7 in the same manner as described above. In addition to storing the target number of each target TG created at that time, the volume number of the host volume HVOL to be created at that time is stored in each host volume number column 35C of that line.

Then, the master cluster control unit 25 creates the target TG requested as described above, the global pool volume GVOL created in step S5A, the internal volume IVOL created in step S5B, and the host to be created at that time. When the mapping with the volume HVOL is completed, the difference before and after the update of the front end target table 33 updated in step S13 and the difference before and after the update of the front end table volume table 31 and the global pool volume table 35 updated in step S14 are displayed. The difference data is transmitted to each storage node 3 other than the own node in the cluster 7 via the back-end network 6 (S15).

Thus, the front-end unit 20 of each storage node 3 that has received the difference data includes the front-end target table 33 in the own node, the front-end table volume table 31, and the global pool volume table 35 based on the difference data. Is updated in the same manner as the front-end target table 33, the front-end table volume table 31, and the global pool volume table 35 in the storage node 3 in which the master cluster control unit 25 is arranged (S16).

Further, when these front-end units 20 finish updating the front-end target table 33, the front-end table volume table 31, and the global pool volume table 35 in the own node in this way, they back up a completion notification to that effect. It is transmitted to the master cluster control unit 25 via the end network 6 (S17).

Then, when the front end unit 20 of all the storage nodes 3 other than the own node in the cluster 7 gives the completion notification of step S17 described above, the master cluster control unit 25 has completed creating the requested target TG. The completion notification of is transmitted to the management node 4 (S18).

On the other hand, when the completion notification of step S18 is given from the master cluster control unit 25, the management node 4 sends the target TG created by the processes of steps S12 to S18 and the host volume HVOL associated with the target TG. A request for registering an initiator to the effect that it should be associated with the compute node 2 (initiator) that is allowed to access the above is transmitted to the master cluster control unit 25 (S19). The initiator registration request includes the node name of the compute node 19 and the target name of the target TG.

Then, when the master cluster control unit 25 receives the initiator registration request, the master cluster control unit 25 is a front end unit in its own node so as to register the compute node 2 specified in the initiator registration request as an initiator that can access the target TG created in step S13. Give instructions to 20. Thus, the front-end unit 20 that receives this instruction registers the compute node 2 as an initiator that can access the target TG (S20). Specifically, the front-end unit 20 is the compute node 2 specified in the initiator registration request in the initiator column 33F (FIG. 8) of the front-end target table 33 stored in the memory 11 (FIG. 2) of the own node. Stores the node name of.

Then, when the master cluster unit 25 finishes registering the compute node 2 as an initiator that can access the target TG as described above, the difference before and after the update of the front-end target table 33 is used as the difference data in the back-end network. It is transmitted to each storage node 3 other than the own node in the cluster 7 via 6 (S21).

Thus, the front-end unit 20 of each storage node 3 that has received the difference data sets the front-end target table 33 in its own node in the storage node 3 in which the master cluster control unit 25 is arranged based on the difference data. It is updated in the same manner as the front end target table 33 (S22).

Further, when these front-end units 20 finish updating the front-end target table 33 in the own node in this way, they transmit a completion notification to that effect to the master cluster control unit 25 via the back-end network 6. (S23).

Then, when the front end unit 20 of all the storage nodes 3 other than the own node in the cluster 7 gives the completion notification of step S23 described above, the master cluster control unit 25 indicates that the registration of the requested initiator has been completed. The completion notification is transmitted to the management node 4 (S24). As described above, this series of processing is completed.

(4) Flow of processing when a failure of a storage node occurs or when an expansion occurs Next, in the information processing system 1, when a failure occurs in any of the storage nodes 3 constituting the cluster 7, or a storage node in the cluster 7. The flow of processing when 3 is added will be described.

(4-1) Flow of processing when a storage node failure occurs FIG. 13 shows a series of processing flows executed in the cluster 7 when a failure occurs in any of the storage nodes 3 constituting the cluster 7. show.

In the case of this embodiment, the master cluster control unit 25 executes a periodic health check for each storage node 3. Then, when the master cluster control unit 25 detects that a failure has occurred in any of the storage nodes 3 during such a health check, the master cluster control unit 25 determines the failed storage node (hereinafter, this is referred to as a failed storage node) 3. The other storage control unit (passive storage control unit) 21 that constitutes the same storage control unit pair 28 as the arranged active storage control unit 21 is specified (S30).

Specifically, the master cluster control unit 25 refers to the storage control unit pair table 30 (FIG. 5), and among the rows of the storage control unit pair table 30, the node number column 30BA of the active column 30B (FIG. 5). A row in which the node number of the failed storage node 3 is stored is specified in (FIG. 5), and the storage control stored in the storage control unit number column 30CB (FIG. 5) of the passive column 30C (FIG. 5) in that row is specified. The part number and the node number stored in the node number column 30CA (FIG. 5) of the passive column 30C are acquired.

Subsequently, the master cluster control unit 25 switches the state of the storage control unit 21 specified in step S30 to the active mode (S31). Specifically, the master cluster control unit 25 has the node number column 30BA (FIG. 5) and the storage control unit number of the active column 30B (FIG. 5) of the row specified in step S30 among the rows of the storage control unit pair table 30. The corresponding number of the node number and the storage control unit number acquired in step S30 is stored in the column 30BB (FIG. 5). Further, the master cluster control unit 25 transmits the difference data of the difference before and after the update of the storage control unit pair table 30 updated in this way to each storage node 3 other than the own node in the cluster 7, respectively. The storage control unit pair table 30 in the storage node 3 is updated in the same manner.

Further, the master cluster control unit 25 activates the storage control unit 21 to which the storage control unit number acquired as described above is assigned in the storage node 3 to which the node number acquired in step S30 is assigned (S32). The storage control unit 21 is made to execute the necessary failure processing (S33).

On the other hand, the master cluster control unit 25 also issues a failure processing execution instruction similar to the failure processing execution instruction in step S31 to the front end unit 20 mounted on the storage node 3 having the node number specified in step S30. It is transmitted via the back-end network 6 (S34).

Then, the front end unit 20 that has received the failure processing execution instruction changes the corresponding destination of the internal volume IVOL associated with the storage control unit 21 that was switched to the active mode in step S32 to the active mode in step S32. A failure time process for switching to the switched storage control unit 21 is executed (S35).

Specifically, regarding the front-end table volume table 31 in the own node, the front-end unit 20 first stores the storage of the active storage control unit 21 which has previously configured the own storage control unit 21 and the storage control unit pair 28. The line in which the control unit number is stored in the storage control unit number column 31F (FIG. 6) is specified.

Then, the front-end unit 20 updates the node identification information stored in the active side node identification information column 31E (FIG. 6) of the row to the node identification information of its own node, and at the same time, updates the storage control unit number column 31F of that row. The storage control unit number stored in is updated to the storage control unit number of the storage control unit 21 whose active mode is switched in step S31. Further, the front-end unit 20 switches the port number stored in the virtual port number column 31G (FIG. 6) of the row to the active mode in step S31, and the virtual port VPT of the storage control unit 21 (FIG. 4). The port number of the virtual port VPT associated with the corresponding host volume HVOL is updated.

Further, the front-end unit 20 first controls the storage of the active storage control unit 21 which has previously configured the storage control unit 21 and the storage control unit pair 28 with respect to the mapping table 32 (FIG. 7) in the own node. The line in which the unit number is stored in the storage control unit number field 32E (FIG. 6) is specified. Then, the front end unit 20 rewrites the node number stored in the node number column 32D (FIG. 6) of that line to the node number of its own node. Further, the front-end unit 20 rewrites the storage control unit number stored in the storage control unit number column 32E of that line to the storage control unit number of the storage control unit 21 switched to the active mode in step S31.

Further, in the front-end unit 20, the volume number of the host volume HVOL associated with the corresponding internal volume IVOL defined in the fault storage node 3 is the host volume number list of the front-end target table 33 (FIG. 8). Identify the line stored in column 33E. Then, the front-end unit 20 sets the IP address stored in the target ID column 33D of that line to the physical port PPT (FIG. 4) in the own node and is associated with the host volume HVOL. Rewrite to the IP address of.

Then, the master cluster control unit 25 stores the difference data of the differences before and after the update of the front-end table volume table 31, the mapping table 32, and the front-end target table 33 updated in this way for each storage other than the own node in the cluster 7. By transmitting to the front-end unit 20 of the node 3, the front-end table volume table 31, the mapping table 32, and the front-end target table 33 in each of the storage nodes 3 are similarly updated.

Although not shown in FIG. 13, the master cluster control unit 25 replaces the active storage control unit 21 arranged in the failure storage node 3 with the processing of steps S31 and S32 in step S32. Select the storage control unit 21 to newly configure the storage control unit pair 28 together with the storage control unit 21 switched to the active mode in.

Such a storage control unit 21 is arranged in any storage node 3 other than the failed storage node 3 in the cluster 7, and is a storage control unit 21 that is unused at that time (both storage control units 21 are paired with the storage control unit 21). Among the storage control units 21) not set in 28, among the storage control units 21 arranged in the storage node 3 other than the storage node 3 in which the storage control unit 21 switched to the active mode in step S32 is arranged. Be selected.

Then, the master cluster control unit 25 sets the storage control unit 21 selected in this way and the storage control unit 21 switched to the active mode in step S32 as a new storage control unit pair 28 in the storage control unit pair table. 30 (FIG. 5) and front-end table volume table 31 (FIG. 6) are updated.

Specifically, the master cluster control unit 25 steps the storage control unit pair table 30 into the node number column 30BA (FIG. 5) and the storage control unit number column 30BB (FIG. 5) of the active column 30B (FIG. 5), respectively. The row in which the node number and the storage control unit number acquired in S30 are stored is specified, and the storage control unit stored in the storage control unit number column 30CB (FIG. 5) of the passive column 30C (FIG. 5) of that row is specified. The number is rewritten to the storage control unit number of the storage control unit 21 selected as described above, and the storage control unit 21 replaces the node number stored in the node number column 30CA (FIG. 5) of the passive column 30C with the storage control unit number. Rewrite with the node number of the arranged storage node 3.

Further, the master cluster control unit 25 sets the front-end table volume table 31 to either the first or second node specific information fields 31C or 31D of the corresponding row via the front-end unit 20 in the own node. The node number of the stored fault storage node 3 is rewritten to the node number of the storage node 3 in which the storage control unit 21 selected as described above is arranged.

Then, the master cluster control unit 25 transfers the difference data of the differences before and after the update of the storage control unit pair table 30 and the front end table volume table 31 updated in this way to the storage nodes 3 other than the own node in the cluster 7. By transmitting each of them, the storage control unit pair table 30 and the front-end table volume table 31 in the storage node 3 are similarly updated.

(4-2) Flow of processing at the time of adding a storage node On the other hand, FIG. 14 shows a flow of a series of processing executed in the cluster 7 when the storage node 3 is added to the cluster 7.

When the storage node 3 is added to the cluster 7, the system administrator notifies the master cluster control unit 25 via the management node 4. Then, the master cluster control unit 25 that has received this notification should move the storage control unit (hereinafter, referred to as an additional storage node) 3 to the added storage node 3 in order to distribute the load between the storage nodes 3 (hereinafter, this is referred to as an additional storage node). Hereinafter, this is referred to as a movement target storage control unit) 21 is selected (S40).

For example, the master cluster control unit 24 selects the active storage control unit 21 arranged on the storage node 3 having the highest load among the storage control units 21 in the cluster 7 as the storage target storage control unit 21 described above.

Then, the master cluster control unit 25 then replaces any one of the storage control units 21 that is unused at that time, which is arranged in the expansion storage node 3, with the movement target storage control unit 21 selected in step S40. The process is executed (S41 to S45).

Specifically, the master cluster control unit 25 first activates the storage control unit (hereinafter, referred to as the selected storage control unit) 21 selected in step S40 of the expansion storage node 3 (S41), and controls the selected storage. (S42) causes the unit 21 to execute a predetermined expansion processing for taking over the processing previously executed by the movement target storage control unit 21 (S42).

Further, the master cluster control unit 25 causes the movement target storage control unit 21 to execute a predetermined expansion processing for allowing the selection storage control unit 21 to take over the processing previously executed by the movement target storage control unit 21 (the movement target storage control unit 21). S41, S42). For this expansion processing, the storage control unit pair table 30 (FIG. 5), the front-end table volume table 31 (FIG. 6), the mapping table 32 (FIG. 7), the front-end target table 33 (FIG. 8), and the global A process of transferring each data of the pool volume table 35 (FIG. 10) to the front end portion 20 of the expansion storage node 3 is also included.

After that, the master cluster control unit 25 switches the correspondence destination of the internal volume IVOL associated with the movement target storage control unit 21 to the selection storage control unit 21 for the front end unit 20 in the own node. Including, an instruction is given to set the environment of the selected storage control unit 21 in the expansion storage node 3 to be the same as the environment of the movement target storage control unit 21 in the storage node 3 in which the movement target storage control unit 21 is arranged ( S43).

The front-end unit 20 that receives this instruction thus receives the storage control unit pair table 30 (FIG. 5), the front-end table volume table 31 (FIG. 6), and the mapping table stored in the memory 11 (FIG. 2) of the own node. 32 (FIG. 7), the front-end target table 33 (FIG. 8), and the global pool volume table 35 (FIG. 10) are updated as necessary (S44).

Specifically, regarding the storage control unit pair table 30, the front-end unit 20 first has a storage control unit number column 30BB (FIG. 5) in which the storage control unit number of the storage control unit 21 to be moved is the active column 30B (FIG. 5). Identify the line stored in 5). Then, the front-end unit 20 rewrites the node number stored in the node number column 30BA (FIG. 5) of the active column 30B of the row to the node number of the additional storage node 3 and stores the active column 30B. The storage control unit number stored in the control unit number field 30BB (FIG. 5) is rewritten with the storage control unit number of the selected storage control unit 21.

Further, regarding the front end table volume table 31, the front end unit 20 first identifies a row in which the storage control unit number of the storage control unit 21 to be moved is stored in the storage control unit number column 31F (FIG. 6).

Then, the front end unit 20 is a storage node 3 in which the storage target storage control unit 21 to be moved is arranged, which is stored in any of the first and second node specific information fields 31C and 31D (FIG. 6) of the row. The node identification information and the node identification information of the storage node 3 stored in the active side node identification information column 31E (FIG. 6) are rewritten into the node identification information of the additional storage node 3, respectively.

Further, the front end unit 20 rewrites the storage control unit number stored in the storage control unit number field 31F of the line to the storage control unit number of the selected storage control unit 21. Further, the front-end unit 20 corresponds to the virtual port number stored in the virtual port number column 31G (FIG. 6) of the row to the corresponding host volume HVOL to which the storage control unit 21 to be moved has been associated. Then, the virtual port number of the virtual port VPT (FIG. 4) defined in the selected storage control unit 21 is rewritten.

Further, in the mapping table 32, the front end unit 20 first identifies a row in which the storage control unit number of the storage control unit 21 to be moved is stored in the storage control unit number column 32E (FIG. 7). Then, the front end unit 20 rewrites the storage control unit number stored in the storage control unit number field 32E of the line to the selected storage control unit number. Further, the front end unit 20 rewrites the node number stored in the node number column 32D (FIG. 7) of that line to the node number of the additional storage node 3.

Further, in the front-end target table 33, the front-end unit 20 first identifies a row in which the volume number of the host volume HVOL corresponding to the host volume number list column 33E (FIG. 8) is stored. The volume number can be obtained, for example, as the volume number stored in the host volume number column 32A of the row specified in the mapping table 32 as described above.

Then, the front end unit 20 associates the IP address stored in the target IP column 33D of that line with the corresponding host volume HVOL in the expansion storage node 3 and stores the storage service network 5 (FIG. 4) of the physical port PPT (FIG. 4). 1) Rewrite to the above IP address.

Then, when the master cluster control unit 25 completes the update of the storage control unit pair table 30, the front-end table volume table 31, the mapping table 32, and the front-end target table 33 in this way, the updated storage control unit pair Storage control by transmitting the difference data of the difference before and after the update of the table 30, the front-end table volume table 31, the mapping table 32) and the front-end target table 33 to the storage nodes 3 other than the own node in the cluster 7, respectively. The part pair table 30, the front-end table volume table 31, the mapping table 32, and the front-end target table 33 are updated in the same manner.

Further, the master cluster control unit 25 accesses the selected storage control unit 21 and holds the virtual port number column 34A (FIG. 9) and the internal volume number column 34B of the storage control unit configuration information table 34 held by the selected storage control unit 21. (FIG. 9) stores the same values as those stored in the virtual port number column 31G (FIG. 6) and the internal volume number column 31H (FIG. 6) of the front-end table volume table 31 in the own node.

By the above processing, the environment of the selected storage control unit 21 in the expansion storage node 3 is set to be the same as the environment of the movement target storage control unit 21 in the storage node 3 in which the movement target storage control unit 21 is arranged.

On the other hand, after that, the compute node 2 responds to the instruction given from the management node 4 based on the operation input of the system administrator, and corresponds to the corresponding definition defined in the expansion storage node 3 by the expansion process in step S44 from the own node. The path expansion process for setting the path to the target TG is executed (S46). Further, the compute node 2 transmits a login request to the front end unit 20 of the additional storage node 3 to log in to the target TG via the path set in this way (S47).

When the front-end unit 20 of the additional storage node 3 receives this login request, it executes a login process according to the login request (S48), and transmits the processing result to the compute node 2 that is the source of the login request. (S49).

(5) Flow of write processing and read processing in this information processing system (5-1) Flow of write processing Next, the flow of write processing in this information processing system 1 will be described. FIG. 15 shows a flow of write processing executed in the cluster 7 when a write request is given to any storage node 3 constituting the cluster 7 from the compute node 2. In the following, the storage node 3 that has received the write request in the cluster 7 will be appropriately referred to as a “write request receiving node”.

In the write request, as described above, the volume number of the host volume HVOL of the write destination, the start address of the storage area of the write destination in the host volume HVOL, and the data to be written (hereinafter, this is referred to as write data). The data length of is specified.

Then, when the first communication device 13 of the "write request receiving node" receives the write request (S50), the first communication device 13 transfers the received write request to the front end unit 20 in the own node (S51). Further, the front end unit 20 secures a buffer area having a required capacity on the memory 11 (FIG. 2) based on the transferred write request (S52), and then completes the process of securing the buffer area. The notification is transmitted to the compute node 2 that is the source of the write request via the first communication device 13 (S53, S54).

Further, when the write data is transferred from the compute node 2 (S55), the first communication device 13 of the "write request receiving node" transfers the write data to the front end unit 20 in the own node. (S56).

The front end unit 20 that has received the write data stores the received write data in the buffer area secured in step S52 (S57), and after that, the write process according to the write request should be executed in the own node. It is determined whether or not it is a process (S58).

This determination is made by referring to the node number column 32D (FIG. 7) of the row corresponding to the host volume HVOL specified as the write destination in the write request among the rows of the mapping table 32 (FIG. 7). Specifically, when the node number stored in the node number field 32D is the node number of the own node, the front end unit 20 should execute the write process according to the write request in the own node. Judge that it is a process. Further, when the node number stored in the node number field 32D is the node number of the other storage node 3, the front end unit 20 executes the write process based on the write request on the other storage node 3. It is determined that the process should be performed.

Then, when it is determined in step S58 that the write process corresponding to the write request should be executed in the own node, the front end unit 20 is designated as the write destination in the write request in the mapping table 32 in the own node. The storage control unit number stored in the storage control unit number field 32E (FIG. 7) of the row corresponding to the host volume HVOL is acquired, and the storage control unit number (active storage) in the own node to which the storage control unit number is assigned is acquired. The write request is transferred to the control unit) 21 (S59). At this time, the front end unit 20 rewrites the volume number of the host volume HVOL of the write destination included in the write request to the volume number of the internal volume IVOL associated with the host volume HVOL.

Further, the storage control unit 21 that has received the write request is associated with the internal volume IVOL designated as the write destination in the write request and the storage area in the internal volume IVOL with reference to a management table (not shown). The logical chunk LC (FIG. 4) and the storage area in the logical chunk LC are specified. Then, the storage control unit 21 generates an I / O command that specifies the logical chunk LC of the write destination of the write data specified in this way and the storage area in the logical chunk LC, and uses this as the back end in the own node. It is transmitted to the unit 23 (S60).

Upon receiving this I / O command, the back-end unit 23 refers to a management table (not shown) and sets each physical chunk PC (FIG. 4) associated with the logical chunk LC specified in the I / O command. The storage device 12 (FIG. 2) to be provided is specified. Further, when one of the specified storage devices 12 exists in the own node, the back-end unit 23 includes the logical chunk LC specified by the I / O command in the storage device 12 and the logical chunk LC in the logical chunk LC. Write data to the storage area of the physical chunk PC corresponding to the storage area (S61). Then, when the writing of the write data to the storage device 12 is completed, the back-end unit 23 transmits a write completion notification to that effect to the storage control unit (active storage control unit) 21 of the own node (S62).

Further, the back-end unit 23 includes a storage node 3 in which a storage device 12 that provides the remaining physical chunk PCs associated with the logical chunk LC is present in addition to the processing in step S61 (hereinafter, appropriately referred to as “another node”. The I / O command and write data are transferred to the corresponding back-end unit 23 in (referred to as "2") via the back-end network 6 (S63). As a result, the same write process as in step S61 is executed in the storage node 3 (“other node 2”) (S64), and when the write process is completed, a write completion notification to that effect is sent via the back-end network 6. It is given to the storage control unit 21 of the “write request receiving node”.

The ledge control unit 21 of the “write request receiving node” sends such a write completion notification from the backend unit 23 of its own node and such a write completion notification from the backend unit 23 of all other necessary storage nodes 3. Upon receipt, the front end unit 20 is notified of the completion of the requested write processing (S76). Further, the front-end unit 20 that has received the completion notification sends a completion notification to the effect that the write processing corresponding to the write request is completed to the sender of the write request via the first communication device 13 and the storage service network 5. It is transmitted to the compute node 2 (S77, S78). As a result, a series of write processing corresponding to the write request is completed.

On the other hand, the front end unit 20 of the "write request receiving node" has obtained a determination in step S58 that the write process corresponding to the write request should be executed in the other storage node 3. In this case, the storage node 3 (hereinafter referred to as the storage node 3) to which the node number stored in the node number column 32D (FIG. 7) of the row corresponding to the host volume HVOL specified as the write destination in the write request in the mapping table 32 (FIG. 7) is assigned. , As appropriate, this is referred to as "another node 1") to transfer the write request (S66). This write request transfer is done over the backend network 6.

Then, the front end portion 20 of the "other node 1" that has received the write request secures a buffer area having a required capacity on the memory 11 (FIG. 2) in the own node based on the received write request (FIG. 2). S67).

Further, the front end unit 20 then refers to the mapping table 32 stored in the memory 11 in the own node, and corresponds to the host volume HVOL specified as the write destination in the write request in the mapping table 32. The storage control unit number stored in the storage control unit number column 32E (FIG. 7) of the row to be stored is acquired, and the storage control unit number (active storage control unit) 21 in the own node to which the storage control unit number is assigned is assigned. Transfer the write request (S68). At this time, the front end unit 20 also notifies the storage control unit 21 of information on whether or not the buffer area having the required capacity has been secured in step S67 (hereinafter, this is referred to as buffer reservation correctness information).

Then, the storage control unit 88 that has received this write request determines whether or not the buffer area can be secured based on the buffer reservation correctness / rejection information given from the front end unit 20 together with this write request as described above. S69), the determination result is transmitted to the front end unit 20 of the own node (S70). Thus, the front-end unit 20 that has received this completion notification responds via the back-end network 6 whether or not the buffer area of the required capacity can be secured on the memory 11 in step S67. Is transmitted to the front end unit 20 (S71).

If the front-end unit 20 of the "write request receiving node" that has received this response cannot secure a buffer area of the required capacity on the memory 11 by the front-end unit 20 of the "other node 1", that effect. Error notification is transmitted to the compute node 2 that is the source of the write request via the first communication device 13, and this series of write processing is completed.

On the other hand, the front end portion 20 of the "write request receiving node" is a back end when the front end portion 20 of the "other node 1" can secure a buffer area having a required capacity on the memory 11. The write data is transferred to the "other node 1" via the network 6 (S72).

Further, the front end unit 20 of the "other node 1" that has received the write data stores the received write data in the buffer area secured in step S67 (S73). After that, in the "other node 1", the same data write processing as the above-mentioned data write processing is executed for steps S59 to S61 and steps S63 to S64 (S74), and when this data write processing is completed, that fact is achieved. The write completion notification is transmitted from the front end unit 20 of the “other node 1” to the storage control unit 21 of the “write request receiving node” via the back end network 6 (S75).

Thus, at this time, the storage control unit 21 of the "write request receiving node" receives the write completion notification from the backend unit 23 of the "other node 1" and the backend unit 23 of all the other storage nodes 3 required. Upon receiving the write completion notification, the front end unit 20 of the own node is notified of the completion notification that the requested write processing is completed (S76). Further, the front-end unit 20 that has received the completion notification sends a completion notification to the effect that the write processing corresponding to the write request is completed to the sender of the write request via the first communication device 13 and the storage service network 5. It is transmitted to the compute node 2 (S77, S78). As a result, a series of write processing corresponding to the write request is completed.

(5-2) Flow of Read Processing On the other hand, FIG. 16 shows a read executed in the cluster 7 when a read request is given to any storage node 3 constituting the cluster 7 from the compute node 2. The processing flow is shown. In the following, the storage node 3 that has received the read request in the cluster 7 will be appropriately referred to as a “read request receiving node”.

In the read request, the volume number of the host volume HVOL of the read destination, the start address of the storage area of the read destination in the host volume HVOL, and the data length of the data to be read (hereinafter, this is referred to as read data). Is specified.

Then, when the first communication device 13 of the "read request receiving node" receives the read request (S80), the first communication device 13 transfers the received read request to the front end unit 20 in the own node (S81). Further, the front end unit 20 analyzes the transferred read request and determines whether or not the read process according to the read request is a process to be executed in the own node (S82).

This determination is made by referring to the node number column 32D (FIG. 7) of the row corresponding to the host volume HVOL specified as the read destination in the read request among the rows of the mapping table 33 (FIG. 7). Specifically, when the node number stored in the node number field 32D is the node number of the own node, the front end unit 20 should execute the read process according to the read request in the own node. Judge that it is a process. Further, when the node number stored in the node number column is the node number of the other storage node 3, the front end unit 20 executes read processing based on the read request in the other storage node 3. Judge that it is a process to be performed.

Then, when the front end unit 20 determines in step S82 that the read process corresponding to the read request should be executed in the own node, the front end unit 20 stores the mapping table in the memory 11 (FIG. 2) of the own node. Acquires the storage control unit number stored in the storage control unit number column 32E (FIG. 7) of the row corresponding to the host volume HVOL specified as the read destination in the read request in 32 (FIG. 7), and obtains the storage control unit number. The read request is transferred to the storage control unit (active storage control unit) 21 in the local node to which is assigned (S83). At this time, the front end unit 20 rewrites the volume number of the read destination host volume HVOL included in the read request to the volume number of the internal volume IVOL associated with the host volume HVOL.

Further, the storage control unit 21 that has received the read request is associated with the internal volume IVOL designated as the read destination in the read request and the storage area in the internal volume IVOL with reference to a management table (not shown). The logical chunk LC and the storage area in the logical chunk LC are specified. Then, the storage control unit 21 generates an I / O command that specifies the logical chunk LC of the read destination specified in this way and the storage area in the logical chunk LC, and sends this to the back end unit 23 in the own node. Transmit (S84).

When the back-end unit 23 receives this I / O command, it refers to a management table (not shown) and refers to one physical unit from each physical chunk PC associated with the logical chunk LC specified in the I / O command. Identify the storage device 12 that provides the chunk PC. Here, it is assumed that the selected physical chunk PC is provided by the storage device 12 mounted on the “read request receiving node”. Then, the back-end unit 23 is stored in the storage area of the logical chunk LC specified by the I / O command and the storage area of the physical chunk PC corresponding to the storage area in the logical chunk LC in the specified storage device 12. Is read (S85). Then, the back-end unit 23 transmits the read data (read data) to the storage control unit 21 of the own node (S86).

The storage control unit 21 that has received the read data transfers the read data to the front end unit 20 of its own node (S87). Further, the front-end unit 20 that has received the read data transmits the read data to the compute node 2 that is the source of the read request via the first communication device 13 and the storage service network 5 (S93, S94). ). As a result, a series of read processing corresponding to the read request is completed.

On the other hand, in the determination in step S82, the front-end unit 20 of the "read request receiving node" has obtained a determination that the read process corresponding to the read request is a process to be executed in the other storage node 3. In this case, the storage node 3 (hereinafter referred to as the storage node 3) to which the node number stored in the node number column 32D (FIG. 7) of the row corresponding to the host volume HVOL specified as the read destination in the read request in the mapping table 32 (FIG. 7) is assigned. , As appropriate, this is referred to as an "other node") to transfer the read request (S88). This read request transfer is done over the backend network 6.

Then, the front end unit 20 of the "other node" that has received this read request analyzes the content of the received read request (S89). Further, the front end unit 20 refers to the mapping table 32 stored in the memory 11 of the own node, and stores the row corresponding to the host volume HVOL specified as the read destination in the read request in the mapping table 32. Acquires the storage control unit number stored in the control unit number field 32E (FIG. 7), and transfers the read request to the storage control unit (active storage control unit) 21 in the own node to which the storage control unit number is assigned. (S90). At this time, the volume number of the read destination host volume HVOL included in the read request is rewritten to the volume number of the internal volume IVOL associated with the host volume HVOL.

After that, in the "other node", the same data read processing as in steps S84 to S86 is executed (S91), and the data (read data) read from the corresponding storage device 12 by this data read processing is "read data". It is transmitted from the front-end unit 20 of the "other node" to the front-end unit 20 of the "read request receiving node" via the back-end network 6 (S92).

Then, the front end unit 20 of the "read request receiving node" that has received the read data transfers the read data to the compute node 2 that is the source of the read request via the first communication device 13 and the storage service network 6. (S93, S94). As a result, a series of read processing corresponding to the read request is completed.

(6) Effect of the present embodiment As described above, in the information processing system 1 of the present embodiment, the internal volume IVOL is created and created in association with the storage control unit 21 arranged in each storage node 3. The internal volume IVOL is associated with the host volume 2 provided to the compute node 2 as a storage area for reading / writing data.

Further, in the information processing system 1, when an I / O request from the compute node 2 is given to the storage node 3, the front end portion 20 of the storage node 3 serves as a read / write destination in the I / O request. The storage node 3 in which the internal volume IVOL associated with the specified host volume HVOL is located is specified, and if the specified storage node 3 is the own node, the I / O request is made in the own node. Transfer to the corresponding storage control unit 21, and if the specified storage node 3 is another storage node 3, the I / O request is transferred to the storage node 3.

Therefore, according to the information processing system 1, the compute node 2 accesses the desired data without being aware of the storage node 3 to which the I / O request is issued, regardless of whether the storage node 3 is scaled out or not. Can be done. Therefore, according to the present embodiment, the highly expandable information processing system 1 can be realized.

Further, according to the information processing system 1, the front end unit 20 is in charge of all the processing related to scale-out, and the storage control unit 21 executes only the processing completed in its own node, so that it already exists as the storage control unit 21. It is also possible to divert the control software used in the storage device of.

(7) Other Embodiments In the above-described embodiment, the case where the internal volume IVOL and the host volume HVOL are associated with each other on a one-to-one basis has been described, but the present invention is not limited to this. A plurality of host volumes HVOL may be associated with the internal volume IVOL.

Further, in the above-described embodiment, the case where the present invention is applied to the information processing system 1 configured as shown in FIG. 1 has been described, but the present invention is not limited to this and has various other configurations. It can be widely applied to information processing systems.

Further, in the above-described embodiment, the case where the front end unit 20, the storage control unit 21, the back end unit 23, and the like in each storage node 2 are configured as software has been described, but the present invention is not limited to this. Instead, these may be configured as hardware.

Further, in the above-described embodiment, the cluster control unit 25 is arranged in each storage node 3, one of the cluster control units 25 is selected as the master cluster control unit 25, and the master cluster control unit 25 is selected. However, the present invention is not limited to this, and the cluster control unit 25 is not arranged in each storage node 3, and the function as the master cluster control unit 25 is provided. The server device or the like may be provided separately from the storage node 3.

Further, the hypervisor may be operated on the server, one or a plurality of virtual machines may be operated on the hypervisor, and various programs shown in FIG. 3 may be operated on the virtual computer. That is, various programs (control software 20, redundancy unit 22, cluster control unit 23) may operate on the hardware of the physical computer or may operate on the virtual computer. Similarly, the compute node 2 may be an application program (host program) running on a virtual computer or a physical host computer (host computer). When the information processing system 1 has a plurality of servers, some of the servers may be located at different sites. Further, a part or all of the server of the information processing system 1 may be on the cloud and the service may be provided to the user via the network.

A configuration (hyper-converged infrastructure) in which a virtual computer on which various programs (control software 20, redundancy unit 22, cluster control unit 23) operates and a virtual computer on which a host program operates are on the same server (node). However, the configuration may be on different servers connected via a network.

The present invention can be widely applied to systems having various configurations including a plurality of storage nodes.

1 ... Information processing system, 2 ... Compute node, 3 ... Storage node, 4 ... Management node, 5 ... Storage service network, 6 ... Backend network, 7 ... Cluster, 10 ... CPU, 12 Storage device, 20 ... Front-end unit, 21 ... Storage control unit, 22 ... Capacity control unit, 23 ... Back-end unit, 25 ... Cluster control unit, 28 ... Storage control unit pair, 30 ... … Storage control unit pair table, 31 …… front-end table volume table, 32 …… mapping table, 33 …… front-end target table, 34 …… storage control unit configuration information management table, 35 …… global pool volume table, GVOL …… Global pool volume, HVOL …… Host volume, IVOL …… Internal volume, LC …… Logical chunk, PC …… Physical chunk, PPT …… Physical port, TG …… Target, VPT …… Virtual port.

Claims (9)

A system that includes a cluster consisting of multiple storage nodes
Each storage node mentioned above
It has a storage device and a logical storage area corresponding to the storage device.
Upon receiving a logical volume creation instruction based on a request from the management device, the storage node is associated with a host volume for reading and writing data provided to a higher-level device, which is a virtual global identifiable within the cluster. Create a pool volume in your node and
By associating the logical storage area in the local node with the global pool volume created in the local node, a virtual internal volume that can be identified in the local node is created in the local node.
And said global pool volumes created, is registered as the corresponding relationship Gama mappings information with the storage node to which the global pool volume was created, the mapping information is updated in synchronization among the storage nodes in the cluster ,
When each storage node receives an I / O request from the higher-level device, the global pool is associated with the host volume designated as the I / O destination in the I / O request based on the mapping information. The storage node of the I / O destination is specified from the information of the storage node corresponding to the volume, and if the specified storage node is the own node, the I / O destination specified in the I / O request is used. When the volume number of the host volume is rewritten to the volume number of the corresponding internal volume, I / O processing is executed on the internal volume, and the specified storage node is another storage node, the relevant storage node is concerned. A system characterized in distributing the I / O request to a storage node. Each said global pool volume
The system according to claim 1, wherein the system is managed as one pool that traverses each storage node that constitutes the cluster. Each storage node includes a storage control unit that executes I / O request processing from the higher-level device, and the storage control unit is a storage control unit and storage arranged in another storage node different from the own node. Set in the control unit pair,
One of the storage control units constituting the storage control unit pair is set to the active system, and the other storage control unit is set to the standby system.
The system according to claim 1, wherein the internal volume is created by the storage node having the storage control unit set in the working system. The third aspect of claim 3, wherein the logical storage area of each storage node belonging to the storage control unit pair is managed as one pool, and the internal volume is created in association with the pool. System. Further, it is provided with a master cluster control device that controls the entire cluster.
When the master cluster control device receives a request for creating the global pool volume from the management device, the master cluster control device selects the storage control unit pair suitable for creating the global pool volume, and any one belonging to the storage control unit pair. Outputs a logical volume creation instruction to the two storage nodes,
Wherein the storage node that received the logical volume creation instruction, transmits the identification information of the global pool volumes created on the master cluster controller,
The system according to claim 4, wherein the system is characterized by the above. The master cluster control device is characterized in that the storage node having the lowest load selects the storage control unit pair set in the active system and outputs the logical volume creation instruction to the storage node. The system according to 5. The master cluster controller receives the global pool volume identification information beam transmitted from the storage node to output the logical volume creation command is registered as the mapping information in association with the identification information and the storage node At the same time, the difference data of the mapping information before and after the update is transmitted to each storage node in the cluster.
Each storage node updates the mapping information in its own node based on the received difference data.
The system according to claim 5, wherein the system is characterized by the above. A method executed by the storage node in a system including a cluster composed of a plurality of storage nodes.
Each storage node mentioned above
It has a storage device and a logical storage area corresponding to the storage device.
Upon receiving a logical volume creation instruction based on a request from the management device, the storage node is associated with a host volume for reading and writing data provided to a higher-level device, which is a virtual global identifiable within the cluster. The first step to create a pool volume in your node,
In association with the logical storage area in the local node, and the global pool volume that is created in this node, the second to create a virtual internal volume that can be identified in this node in the local node Steps and
And said global pool volumes created, third to register the correspondence between the storage node that created the global pool volume as mapping information, updating the mapping information in synchronization among the storage nodes in the cluster With steps and
When each storage node receives an I / O request from the host device, the global pool associated with the host volume designated as the I / O destination in the I / O request based on the mapping information. The storage node of the I / O destination is specified from the information of the storage node corresponding to the volume, and if the specified storage node is the own node, the I / O destination specified in the I / O request is used. When the volume number of the host volume is rewritten to the volume number of the corresponding internal volume, I / O processing is executed on the internal volume, and the specified storage node is another storage node, the relevant storage node is concerned. A system control method characterized by distributing the I / O request to a storage node. A program executed by the storage node in a system including a cluster composed of a plurality of storage nodes.
Each storage node mentioned above
It has a storage device and a logical storage area corresponding to the storage device.
Upon receiving a logical volume creation instruction based on a request from the management device, the storage node is associated with a host volume for reading and writing data provided to a higher-level device, which is a virtual global identifiable within the cluster. The first step to create a pool volume in your node,
A second that creates a virtual internal volume that can be identified in the local node by associating the logical storage area in the local node with the global pool volume created in the local node. Steps and
And said global pool volumes created, third to register the correspondence between the storage node that created the global pool volume as mapping information, updating the mapping information in synchronization among the storage nodes in the cluster With steps and
When each storage node receives an I / O request from the host device, the global pool associated with the host volume designated as the I / O destination in the I / O request based on the mapping information. The storage node of the I / O destination is specified from the information of the storage node corresponding to the volume, and if the specified storage node is the own node, the I / O destination specified in the I / O request is used. When the volume number of the host volume is rewritten to the volume number of the corresponding internal volume, I / O processing is executed on the internal volume, and the specified storage node is another storage node, the relevant storage node is concerned. A program characterized in that the storage node executes a process of distributing the I / O request to the storage node.

Images