JP4618777B2 - Data processing system and storage subsystem provided in the data processing system - Google Patents

Description

  The present invention relates to data storage processing technology, for example, data replication between a plurality of storage subsystems.

  Conventionally, even if a failure occurs in the first storage subsystem, the data stored in the first storage subsystem is transferred to the second and third storage subsystems so that the data processing system including the failure can provide services. The technology to duplicate is considered. As such a technique, for example, there is a technique disclosed in Japanese Patent Laid-Open No. 2003-122509. According to this technique, the first storage subsystem holds the first logical volume, the second storage subsystem holds the second logical volume, and the third storage subsystem And the first storage subsystem writes data (hereinafter referred to as “data A”) to the first logical volume at the timing of data A write and synchronization. Transfer data A to the second storage subsystem, write data A to the second logical volume, and transfer data A to the third storage subsystem at a timing asynchronous with the data A write. Data A is written to the third logical volume.

JP 2003-122509 A

  In the above prior art, the function of grasping the progress status of data transfer / reception / update between the second storage subsystem and the third storage subsystem is the second storage subsystem and the third storage subsystem. Provided in each of the systems. However, there may be a case where it is desirable to perform data-consistent replication between the second and third storage subsystems without providing such a function.

  Accordingly, an object of the present invention is to provide a data processing system having at least three or more storage subsystems, in which data written in the first storage subsystem is transferred to and retained in each of the second and third storage subsystems. The second and third storage subsystems can ensure data consistency between the second and third storage subsystems without having to query each other's progress regarding data updates. It is in.

  Other objects of the present invention will become clear from the following description.

  A data processing system according to a first aspect of the present invention includes a first storage subsystem that receives the write data from a first host terminal that transmits write data that is data to be written, and stores the received write data. A second storage subsystem connected to the first storage subsystem and a third storage subsystem connected to the first storage subsystem.

  The first storage subsystem includes a first storage device assigned with an attribute as a first replication source, and one or more first storage devices associated with the first storage device and assigned with an attribute as a second replication source. And two storage devices. The first storage subsystem stores the write data received from the first host terminal in the first storage device. The first storage subsystem issues an update number indicating the order in which the first storage device is updated by storing the write data in the first storage device, and the issued update number and the A data set including write data is generated, and the generated data set is stored in the one or more second storage devices. The first storage subsystem reads a data set from the one or more second storage devices, and transmits the read data set to the second storage subsystem. The first storage subsystem reads a data set from the one or more second storage devices, and transmits the read data set to the third storage subsystem.

  A third storage in which the second storage subsystem is associated with at least one of the one or more second storage devices and assigned an attribute as a second replication destination that is a pair of the second replication source A fourth storage device associated with the third storage device and assigned with an attribute as a first replication destination that is a pair of the first replication source. The second storage subsystem receives the data set from the first storage subsystem and stores the received data set in the third storage device. The second storage subsystem selects a data set to be read from the third storage device based on an update number included in each of the one or more data sets in the third storage device. The selected data set is read from the three storage devices, and the write data in the read data set is stored in the four storage devices.

  A fifth storage device in which the third storage subsystem is associated with at least one of the one or more second storage devices and assigned an attribute as the second replication destination; and the fifth storage device And a sixth storage device to which the attribute as the first replication destination is assigned. The third storage subsystem receives the data set from the first storage subsystem and stores the received data set in the fifth storage device. Further, the third storage subsystem selects a data set to be read from the fifth storage device based on an update number included in each of the one or more data sets in the fifth storage device. The selected data set is read from the five storage devices, and the write data in the read data set is stored in the six storage devices.

  From the third or fifth storage device, for example, a data set with the smallest update number is read out of data sets that have not been read out yet. That is, reading of the data set from the third or fifth storage device is performed in the order of update numbers.

  In the first embodiment of the data processing system, when the second storage subsystem is connected to a second host terminal that transmits write data, and a failure occurs in the first host terminal, the first host The storage subsystem reads an unsent data set to at least the second storage subsystem among the one or more data sets in the one or more second storage devices, and the read unsent data set is To the second storage subsystem. The second storage subsystem receives the untransmitted data set from the first storage subsystem, stores the received data set in the third storage device, and is still in the third storage device One or more data sets that have not been read are read in the order of update numbers, and the write data in the read data sets are stored in the four storage devices. The data processing system replaces the first replication source that is the attribute of the first storage device and the first replication destination that is the attribute of the fourth storage device, and the one or more second storage devices The second replication source that is the attribute of the second storage device and the second replication destination that is the attribute of the third storage device are switched. The second storage subsystem receives write data from the second host terminal and stores the received write data in the fourth storage device. Further, the second storage subsystem issues an update number indicating an order in which the fourth storage device is updated by storing the write data in the fourth storage device, and the issued update number and the A data set including write data is generated, and the generated data set is stored in the third storage device. The second storage subsystem reads a data set from the third storage device and transmits the read data set to the first storage subsystem. The first storage subsystem receives the data set from the second storage subsystem and stores the received data set in the one or more second storage devices. In addition, the first storage subsystem is a read target from the one or more second storage devices based on update numbers respectively included in the one or more data sets in the one or more second storage devices. The selected data set is read from the one or more second storage devices, and the write data in the read data set is stored in the one storage device.

  In the second embodiment of the data processing system described above, in the first embodiment, the third storage subsystem is connected to the second storage subsystem. The first storage subsystem or the second storage subsystem reads a data set that has not been transmitted to at least the third storage subsystem among the one or more data sets in the one or more second storage devices, The read untransmitted data set is transmitted to the third storage subsystem. The third storage subsystem receives the untransmitted data set from the first storage subsystem, stores the received data set in the fifth storage device, and is still in the fifth storage device One or more data sets that have not been read are read in the order of update numbers, and the write data in the read data sets are stored in the six storage devices. In the data processing system, the fifth storage in which the second replication destination is allocated to the third storage device to which the second replication source is allocated by exchanging the second replication destination and the second replication source. Associate devices. The second storage subsystem reads a data set from the third storage device and transmits the read data set to the third storage subsystem. The third storage subsystem receives the data set from the second storage subsystem and stores the received data set in the fifth storage device.

  In the third embodiment of the data processing system, the first storage subsystem and the third storage subsystem may be configured to stop communication between the first storage subsystem and the second storage subsystem. Independently, communication between the first storage subsystem and the third storage subsystem is maintained.

  Specifically, for example, at least one of the second storage subsystem and the first storage subsystem reads the data set having the update number next to the update number of the data set that has been recently read from the third storage device. If this is not possible, communication between the first storage subsystem and the second storage subsystem is centered. The first storage subsystem and the third storage subsystem can store a data set recently read from the fifth storage device even if communication between the first storage subsystem and the second storage subsystem is interrupted. If the data set of the update number next to the update number can be read, the communication between the first storage subsystem and the third storage subsystem is maintained.

  In the fourth embodiment of the data processing system, at least one of the second storage subsystem and the third storage subsystem transmits a read command to the first storage subsystem. The first storage subsystem receives the read command, and transmits the read data set to at least one of the second storage subsystem and the third storage subsystem in response to the read command.

  In the fifth embodiment of the data processing system, the one or more second storage devices correspond to the first second storage device associated with the third storage device and the fifth storage device. The first storage subsystem stores the generated data set in both the first and second second storage devices.

  In the sixth embodiment of the above data processing system, at least one of the first to third storage subsystems is free on a storage device to which the second replication source or the second replication destination is assigned. When the storage area runs out, the data set with the oldest update number is deleted from one or more data sets stored in the storage device.

  In a seventh embodiment of the above data processing system, when the second storage subsystem is connected to the third storage subsystem and a failure occurs in the first storage subsystem, the second storage subsystem At least one of the system and the third storage subsystem includes the third storage device to which the attribute as the second replication source is assigned, and the fifth storage device to which the attribute as the second replication destination is assigned. Associate. The second storage subsystem transmits the data set read from the third storage device to the third storage subsystem. The third storage subsystem receives a data set from the second storage subsystem and stores the received data set in the fifth storage device.

  The storage subsystem according to the second aspect of the present invention is connectable to a host terminal that transmits write data that is data to be written, a second storage subsystem, and a third storage subsystem, and the first storage A device; one or more second storage devices; and a control device that controls access to the first storage device and the one or more second storage devices. The control device stores the write data received from the first host terminal in the first storage device. Further, the control device issues an update number indicating an order in which the first storage device is updated by storing the write data in the first storage device, and the issued update number, the write data, Is generated, and the generated data set is stored in the one or more second storage devices. The control device reads a data set from the one or more second storage devices, and transmits the read data set to the second storage subsystem. In addition, the control device reads a data set from the one or more second storage devices, and transmits the read data set to the third storage subsystem.

  A data processing method according to a third aspect of the present invention includes a first storage subsystem that receives the write data from a first host terminal that transmits write data that is data to be written, and stores the received write data. The data processing system includes a second storage subsystem connected to the first storage subsystem and a third storage subsystem connected to the first storage subsystem. The first storage subsystem includes a first storage device and one or more second storage devices. The second storage subsystem includes a third storage device and a fourth storage device. The third storage subsystem includes a fifth storage device and a sixth storage device. The data processing method includes a step of storing write data transmitted from the first host terminal in a first storage device, and the first storage device is updated by storing the write data in the first storage device. Issuing an update number representing the order, and generating a data set including the issued update number and the write data, and storing the generated data set in one or more second storage devices; Reading a data set from the one or more second storage devices; storing the read data set in the third storage device and the fifth storage device; and one in the third storage device Based on the update number included in each of the above data sets, the data set is stored from the third storage device. And storing the write data in the read data set in the fourth storage device, and the update number included in each of the one or more data sets in the fifth storage device, Reading a data set from the fifth storage device, and storing write data in the read data set in the sixth storage device.

  According to the present invention, in a data processing system including first to third storage subsystems, data written to the first storage subsystem is transferred to and retained in each of the second and third storage subsystems. In this case, data consistency can be obtained between the second and third storage subsystems without each of the second and third storage subsystems inquiring about the progress of the data update with respect to each other.

  Hereinafter, with reference to the drawings, an embodiment of the present invention and some examples based on the embodiment will be described.

  FIG. 1A shows an overview of a first replication process performed by a data processing system according to an embodiment of the present invention, and FIG. 1B shows an overview of a second replication process performed by the data processing system.

  The data processing system 1 includes at least three storage subsystems 100, for example, a first storage subsystem 100A, a second storage subsystem 100B, and a third storage subsystem 100C. Hereinafter, in order to make the explanation easier to understand, elements related to the first storage subsystem 100A are given a branch code “A” in addition to the reference number, and elements related to the second storage subsystem 100B are added to the reference number. The branch code “B” is attached, and the element related to the third storage subsystem 100C is assigned the branch code “C” in addition to the reference number.

  Each of the storage subsystems 100A, 100B, and 100C can communicate with the host terminal 180. The host terminal 180 is a computer that includes a CPU, memory, and the like as hardware resources, and is specifically a personal computer or a server machine, for example. Hereinafter, the host terminal 180 serving as the communication partner of the first storage subsystem 100A is referred to as “first host terminal 180A”, and the host terminal 180 serving as the communication partner of the second storage subsystem 100B is referred to as “second host terminal 180B”. The host terminal 180 that is the communication partner of the third storage subsystem 100C is referred to as “third host terminal 180C”.

  Each of the storage subsystems 100A, 100B, and 100C includes one or a plurality of physical storage devices (for example, hard disk drives) (not shown), and a logical storage device is included in the one or the plurality of storage devices. A plurality of logical volumes (hereinafter simply referred to as “VOL”) 6 are provided. For example, the first storage subsystem 100A includes three VOLs 6A1 to 6A3, the second storage subsystem 100B includes two VOLs 6B1 to 6B2, and the third storage subsystem 100C includes two VOLs 6C1 to 6C2. Is provided.

  At least one type of VOL attribute among a plurality of types of VOL attributes is assigned to VOL6. There are four types of VOL attributes that can be assigned, for example, PVOL, SVOL, PJNLVOL, and SJNLVOL.

  VOL 6 having a VOL attribute “PVOL” is a VOL (hereinafter, write data VOL) that is a write destination of write target data (hereinafter, write data) 2 received by the storage subsystem 100 from the host terminal 180, and , Primary write data VOL that is a copy source of write data 2. Hereinafter, VOL6 whose VOL attribute is “PVOL” will be referred to as “PVOL6”.

  A VOL 6 with a VOL attribute “SVOL” is a secondary write data VOL that is a copy destination of the write data 2 stored in the PVOL 6. Hereinafter, VOL6 having the VOL attribute “SVOL” is referred to as “SVOL6”.

  A VOL 6 having a VOL attribute “PJNLVOL” is a VOL (hereinafter referred to as “JNLVOL”) 3 in which a journal (hereinafter abbreviated as “JNL”) 3 generated based on the write data 2 is stored, and a JNL replication source Is the primary JNLVOL. Hereinafter, VOL6 whose VOL attribute is “PJNLVOL” will be referred to as “PJNLVOL6”.

  VOL6 with a VOL attribute “SJNLVOL” is a secondary JNLVOL that is a replication destination of JNL3 stored in PJNLVOL6. Hereinafter, VOL6 whose VOL attribute is “SJNLVOL” will be referred to as “SJNLVOL6”.

  VOL 6 may have both VOL attributes “SVOL” and “PVOL”. In this case, the write data 2 may be written to the VOL 6 as the SVOL 6 (that is, the copy destination of the write data 2), or the write data 2 may be read out as the PVOL 6 (that is, the copy source of the write data 2). is there. Similarly, VOL 6 may have both VOL attributes “SJNLVOL” and “PJNLVOL”. In this case, JNL3 may be written as SJNLVOL6 (that is, the copy destination of JNL3), and JNL3 may be read out as PJNLVOL6 (that is, the copy source of JNL3). Hereinafter, VOL6 having both VOL attributes “SVOL” and “PVOL” will be referred to as “SPVOL6”, and VOL6 having both VOL attributes “SJNLVOL” and “PJNLVOL” will be referred to as “SPJNLVOL6”.

  JNL 3 is a data group generated by the storage subsystem 100 based on the write data 2. JNL 3 includes write data 2 and update data 4. The update data 4 is data for managing the storage position and update order of the write data 2.

  FIG. 2 shows a configuration example of the update data 4. FIG. 3 is a configuration example of the write data VOL and JNLVOL, and particularly shows a configuration example represented by the update data 4 illustrated in FIG. Hereinafter, the position from the beginning of the storage area of the VOL 6 (in other words, the position relative to the beginning) will be referred to as “address” for convenience. The write data 2 included in the JNL 3 including the update data 4 is referred to as “JNL write data 2”, and the write data 2 written according to the write command is referred to as “original write data 2”.

As data element items constituting the update data 4, for example, the following data element items (1) to (5),
(1) Time when the write command is received,
(2) Update number,
(3) Logical address of write instruction (for example, set of VOLID of VOL where original write data 2 is written and address of VOL),
(4) Write data size,
(5) JNLVOL logical address storing the write data,
There is. According to the example of the update data 4 shown in FIG. 2, the JNL write data 2 corresponds to the original write data 2 written according to the write command received at 22:20:10 on March 17, 1999. It can be seen that it is. Further, as shown in FIG. 3, the write command is a command that means to start writing the original write data 2 from the address 700 of the write data VOL6 whose VOLID (identifier of VOL6) is “1”. It can be seen that the data size of data 2 is 300 KB. Further, it can be seen that the JNL write data 2 corresponding to the original write data 2 can be written from the address 1500 of JNLVOL 6 whose VOLID is “4” as shown in FIG. It can be seen that the JNL write data 2 is the write data 2 written at the time of the fourth update. The update data 4 may hold only one of the time when the write command is received and the update number. Further, when the write instruction creation time exists in the write instruction from the host terminal 180, the creation time in the write instruction may be held in the update data 4 instead of the time when the write instruction is received.

  For example, as shown in FIG. 3, the JNLVOL 6 is divided into a storage area (update data area) 7 for storing update data 4 and a storage area (write data area) 8 for storing write data. Update data is stored in the update data area 7 in the order of update numbers from the top of the update data area 7. When the end of the update data area 7 is reached, the next update number is stored from the top of the update data area 7. . JNL write data 2 is stored in the write data area 8 in order from the top of the write data area 8. When the end of the write data area 8 is reached, the next JNL write data 2 is stored from the top of the write data area 8. . The ratio of the sizes of the update data area 7 and the write data area 8 may be a fixed value or may be set and changed by a specific terminal such as an SVP (maintenance terminal) or a host terminal 180 described later. Information regarding these configurations can be included in pointer management data 700 described later. In the following description, the JNLVOL 6 is used by being divided into the update data area 7 and the write data area 8, but a method of continuously storing the set of update data 4 and write data 2 from the beginning of the JNLVOL 6 is adopted. (That is, it may not be divided into the update data area 7 and the write data area 8).

  Now refer to FIGS. 1A and 1B again.

  In the data processing system 1, the VOL pair 14 is configured by the PVOL 6 and the SVOL 6, and another VOL pair (hereinafter referred to as “mirror pair” for convenience) 12 is configured by the PJNLVOL 6 and the SJNLVOL 6. A PVOL 6 constituting a certain VOL pair 14 is associated with the PJNLVOL 6 of the mirror pair 12, and an SVOL 6 constituting the VOL pair 14 is associated with the SJNLVOL 6 of the same mirror pair 12. Thus, a VOL group 16 composed of PVOL6, PJNLVOL6, SVJNLVOL6, and SVOL6, in other words, a VOL group 16 including the VOL pair 14 and the mirror pair 12 is constructed. Under this configuration, the original write data 2 is stored in the following order, that is, PVOL6, JNL3 is generated based on the original write data 2, stored in PJNLVOL6, and JNL3 is read from PJNLVOL6. The original write data 2 stored in the PVOL 6 is copied to the SVOL 6 in the order in which the original write data 2 is restored in the SVOL 6 based on the JNL 3 stored in the SJNLVOL 6 and stored in the SJNLVOL 6. It should be noted that the number of PVOLs versus the number of SVOLs (and / or the number of SVOLs versus the number of PVOLs) can be increased to one or more by the combination of VOL pairs. Further, the number of PJNLVOLs versus the number of SJNLVOLs (and / or the number of SJNLVOLs vs. the number of PJNLVOLs) can be increased to one or more by the combination of mirror pairs.

  In the present embodiment, PVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, PJNLVOL6, which PJNLVOL6 is associated with PJNLVOL6 A VOL group 16 including SJNLVOL6 and SVOL6 can be defined. Further, in the present embodiment, depending on how the plurality of VOL groups 16 are combined, specifically, for example, a PVOL of one VOL group 16 and a PVOL of another VOL group 16 are made the same, or Depending on whether the PVOL of one group 16 and the SVOL of another VOL group 16 are the same, the data replication path and the replication direction (in other words, the path and direction in which the write data 2 flows) are set to two or more. You can define whether or not Hereinafter, a data replication method having two or more data replication paths and replication directions is referred to as a “multi-target method” for convenience, and a data replication method having one data replication path and replication direction is referred to as “ This is referred to as “multi-hop method”.

  First, with reference to FIG. 1A, an outline of data replication processing by the multi-target method will be described.

  As illustrated in FIG. 1A, the first PJNLVOL 6A2 and the second PJNLVOL 6A3 included in the storage subsystem 100A are associated with one PVOL 6A1 included in the first storage subsystem 100A. The first SJNLVOL 6B1 included in the second storage subsystem 100B is associated with the first PJNLVOL 6A2, and the first SVOL 6B2 included in the storage subsystem 100B is associated with the first SJNLVOL 6B1. On the other hand, the second SJNLVOL6C1 provided in the third storage subsystem 100C is associated with the second PJNLVOL6A3, and the second SVOL6C2 provided in the storage subsystem 100C is associated with the second SJNLVOL6C1. . With this configuration, the first and second data replication paths and the replication direction are defined, and the data processing system 1 performs the following processing.

  The first storage subsystem 100A receives the original write data 2 from the first host terminal 180A and stores the original write data 2 in the PVOL 6A1. Also, the first storage subsystem 100A generates a duplicated JNL (hereinafter referred to as the first and second JNL) 3 including an update number indicating the update order to the PVOL 6A1 at that time, and the first JNL3 Are stored in the first PJNLVOL 6A2 and the second JNL3 is stored in the second JNLVOL 6A3 (JNL duplication can be performed, for example, on the cache memory 130 described later). The update numbers included in the first and second JNL 3 are the same update number. Note that the first and second JNLs 3 including the same update number may be generated by a process different from the duplex process.

  After this processing, replication processing is performed in accordance with the first replication path and the replication direction, that is, the path with PVOL6A1 as the replication start VOL, first PJNLVOL6A2 and first SJNLVOL6B1 as the relay VOL, and first SVOL6B2 as the replication goal VOL. As a result, the original write data 2 written in the PVOL 6A1 is restored to the first SVOL 6B2. Specifically, the second storage subsystem 100B generates a JNL read instruction for reading the first JNL3 from the first PJNLVOL6A2 at a predetermined or arbitrary timing, and stores the JNL read instruction in the first storage Transmit to subsystem 100A. The JNL read instruction may mean, for example, an instruction for simply reading JNL, or the youngest update number of the first JNL 3 that has not yet been read (in other words, read in the immediately preceding round). An instruction for reading the first JNL3 including the update number next to the update number included in the first JNL3 (specifically, for example, including the VOLID of the first PJNLVOL6A2 and the update number) Read command). In response to the read command, the first storage subsystem 100A reads the first JNL3 having the update number specified in the read command from the first PJNLVOL6A2, and transmits the read first JNL3 to the read command. Return to the original second storage subsystem 100B. The second storage subsystem 100B stores the first JNL3 received from the first storage subsystem 100A in the first SJNLVOL 6B1. In addition, the second storage subsystem 100B has the smallest update number (in other words, the first read out in the immediately preceding time) among the one or more first JNL3s that have not been restored at a predetermined or arbitrary timing. The first JNL3 including the update number next to the update number included in the JNL3 is read from the first SJNLVOL6B1, and the JNL write data 2 included in the read first JNL3 is used as the original write data 2 as the first SVOL6B2. To store. The timing at which the first JNL3 is written to the first SJNLVOL 6B1 and the timing at which the first JNL3 is read from the first SJNLVOL 6B1 may be the same or different. In other words, for example, the update number in the first JNL3 written in the first SJNLVOL 6B1 at a certain timing and the first JNL3 in the first JNL3 read out from the first SJNLVOL 6B1 at the same timing or the latest timing. The update number may be the same or different.

  Also, by performing replication processing according to the second replication path and the replication direction, that is, the path in which PVOL6A1 is the replication start VOL, the second PJNLVOL6A3 and the second SJNLVOL6C1 are the relay VOL, and the second SVOL6C2 is the replication goal VOL. The original write data 2 written in the PVOL 6A1 is restored to the second SVOL 6C2. A specific processing flow example is the same as the replication processing according to the first replication path and the replication direction. In at least one of the first and second duplication paths and duplication directions, the first storage subsystem 100A sends the write target JNL3 and the write command to the SJNLVOL 6B1 or 6B2 as the destination. It may be stored in SJNLVOL 6B1 or 6B2.

  The above is the outline of the duplication processing according to the multi-target method. According to this duplication processing, a plurality of JNLs 3 having the same update number are generated based on the original write data 2 written in the PVOL 6A1, and the plurality of JNLs 3 are a plurality of PJNLVOLs 6A2 associated with the PVOL 6A1. And 6A3, respectively. Thereafter, JNL3 is read from PJNLVOL6A2 and 6A3 to SJNLVOL6B1 and 6C1, respectively, which constitute mirror pair 12 from PJNLVOL6A2 and 6A3 in the order according to the update number in JNL3, and PVOL6A1 and VOL pair are based on JNL3. The same original write data 2 as the original write data 2 written in the PVOL 6A1 is restored to the SVOLs 6B2 and 6C2 (in other words, the SVOLs 6B2 and 6C2 respectively associated with the JNL3 replication destinations SJNLVOL 6B1 and 6C1). The As a result, the consistency of data between the two storage subsystems 100B and 100C can be maintained without the second storage subsystem 100B and the third storage subsystem 100C inquiring about the data update status.

  Next, with reference to FIG. 1B, the outline of the data replication processing by the multi-hop method will be described.

  As illustrated in FIG. 1B, PJNLVOL6B2 included in the storage subsystem 100B is associated with PVOL6B2 included in the second storage subsystem 100B. The SJNLVOL6A2 included in the first storage subsystem 100A is associated with the PJNLVOL6B1, and the SPVOL6A1 included in the storage subsystem 100A is associated with the SJNLVOL6A2 as an SVOL (SVOL that constitutes the VOL pair 14 with the PVOL6B2). . Further, the SPVOL 6A1 is associated with the PJNLVOL 6A3 provided in the storage subsystem 100A, assuming that the SPVOL 6A1 is a PVOL. The SJNLVOL6B1 provided in the third storage subsystem 100C is associated with the PJNLVOL6A3, and the SVOL provided in the storage subsystem 100C (the SVOL constituting the PVOL and the SVOL constituting the VOL pair 14) 6C2 is associated with the SJNLVOL6B1. Attached. With this configuration, the PVOL 6B2 included in the second storage subsystem 100B is set as the replication start VOL, the SVOL 6C2 included in the third storage subsystem 100C is set as the replication goal VOL, and the VOLs 6B2, 6A2, 6A1, 6B3, and 6A2 are set as the relay VOL. A data replication path and a replication direction are defined, and the data processing system 1 performs the following processing.

  The second storage subsystem 100B receives the original write data 2 from the second host terminal 180B and stores the original write data 2 in the PVOL 6B2. The second storage subsystem 100B generates JNL3 including an update number indicating the update order of the PVOL 6B2 at that time, and stores the JNL3 in the PJNLVOL 6B1.

  The first storage subsystem 100A generates a JNL read instruction for reading JNL3 from the PJNLVOL6B1 at a predetermined or arbitrary timing, and transmits the JNL read instruction to the second storage subsystem 100B. The JNL read instruction may mean, for example, an instruction for simply reading the JNL, or the smallest update number among the JNLs 3 that have not yet been read (in other words, included in the JNL read in the previous round) An instruction for reading JNL including the update number next to the update number (specifically, for example, a read instruction including the VOLID of PJNLVOL6B1 and the update number) may be used. In response to the read command, the second storage subsystem 100B reads JNL3 having the update number specified in the read command from PJNLVOL6B1, and reads the read JNL3 from the first storage sub that is the source of the read command. Return to system 100B. The first storage subsystem 100A stores the JNL3 received from the second storage subsystem 100B in the SJNLVOL6A2. Further, the first storage subsystem 100A has the smallest update number (in other words, one or more JNLs 3 that have not been restored) at a predetermined or arbitrary timing (for example, at the same timing as the storage of JNL in the SJNLVOL 6A2). Then, JNL3 including the update number next to the update number included in the first JNL3 read immediately before is read from SJNLVOL6A2, and the JNL write data 2 included in the read JNL3 is the original write. The data 2 is stored in the SPVOL 6A1. Note that the timing at which JNL3 is written to SJNLVOL6A2 and the timing at which JNL3 is read from SJNLVOL6A2 may be the same or different. In other words, for example, the update number in JNL3 written to SJNLVOL6A2 at a certain timing and the update number in JNL3 read from SJNLVOL6A2 at the same timing or the latest timing are the same. It may be different.

  A process similar to the data replication process performed in the VOL group 16 having PVOL6B2, PJNLVOL6B1, SJNLVOL6A2 and SPVOL6A1 is similar to the SPVOL6A1, PJNLVOL6A3, SJNLVOL6C1 and the other VOL of the 16th. That is, the first storage subsystem 100A generates JNL3 including an update number indicating the update order of SPVOL6A1 from SPVOL6A1, and stores the JNL3 in PJNLVOL6A3. Thereafter, the first storage subsystem 100A receives the JNL3 read command from the third storage subsystem 100C, and in response to this, reads the JNL3 from the PJNLVOL 6A3 and transfers it to the third storage subsystem 100C. Thereby, in the third storage subsystem 100C, the JNL3 is written to the SJNLVOL6C1, and the JNL write data 2 included in the JNL3 is restored to the SVOL6C2 as the original write data 2.

  The above is the outline of the duplication processing according to the multi-hop method. Note that JNL replication from PJNLVOL to SJNLVOL was performed by the storage subsystem 100 having SJNLVOL sending a read instruction to the storage subsystem 100 having PJNLVOL. Conversely, the storage subsystem having PJNLVOL is This may be done by the system 100 sending a write command to the storage subsystem 100 having SJNLVOL.

  According to the replication processing according to this multi-hop method, JNL3 including the update number is generated based on the original write data 2 written in PVOL6B2 of the second storage subsystem 100B, and the JNL3 is associated with PVOL6B2. Stored in the PJNLVOL6B1. Thereafter, the write data 2 is transferred downstream along the defined one replication path and replication direction, and eventually the replication goal VOL in the replication path and the replication direction of the third storage subsystem 100C. The same data as the original write data 2 written in the PVOL 6B2, which is the duplication start VOL, is duplicated in the SVOL 6C2. As a result, the consistency of data between the two storage subsystems 100B and 100C can be maintained without the second storage subsystem 100B and the third storage subsystem 100C inquiring about the data update status.

  The data processing system 1 can dynamically perform switching between the multi-target method and the multi-hop method described above, or combining the multi-target method and the multi-hop method. For example, when the multi-target method illustrated in FIG. 1A is adopted and a failure occurs in the first host terminal 180A, it is possible to switch to the multi-hop method illustrated in FIG. 1B. Further, for example, when the multi-hop method exemplified in FIG. 1B is adopted and a failure occurs in the second host terminal 180B, the multi-target method exemplified in FIG. 1A can be switched. Further, for example, in the multi-target method when the number of storage subsystems 100 is four or more, at least one of the first replication route and the replication direction and the second replication route and the replication direction is a replication route according to the multi-hop method. It can be a replication direction.

  Hereinafter, the basic configuration and processing of the data processing system 1 according to the present embodiment will be described with reference to FIGS. 4 to 20, and then the multi-target method will be described in detail as a first example of the embodiment. The multi-hop scheme will be described in detail as a second example of the embodiment.

  FIG. 4 shows a configuration example of the data processing system 1. In this figure, the configuration of the first storage subsystem 100A is shown in detail, and the configuration of the other storage subsystems 100B and 100C is the same as that of the first storage subsystem 100A, so it is schematically shown. . Hereinafter, the first storage subsystem 100A will be described as a representative example, and the other storage subsystems 100B and 100C will be described as appropriate.

  The first storage subsystem 100A is a disk array system such as a RAID (Redundant Array of Independent Disks) system. The first storage subsystem 100A includes, for example, a control device 101A that controls processing performed by the first storage subsystem 100A, a RAID group 210A, and a service processor (SVP) 281A. The control device 101A includes, for example, a plurality of disk adapters (hereinafter referred to as DKA) 120A, a plurality of channel adapters (hereinafter referred to as CHAs) 110A, a cache memory 130A, a shared memory 140A, and a switching control unit 270A.

  The RAID group 210A includes a plurality of storage devices 150A and provides, for example, redundant storage based on RAID such as RAID1 or RAID5. Each storage device 150A can be composed of a storage device such as a hard disk drive (or hard disk itself), a semiconductor memory device, a magneto-optical disk drive (or magneto-optical disk itself), and the like. At least one or more VOL 6A, which is a logical storage area, can be set on the physical storage area provided by each storage device 150A. A plurality of write data used from the host terminal 180A can be stored in the VOL 6A. Further, in another VOL 6A, first control information 141A and the like described later can be stored and used as a system area. Note that all of the storage devices 150A need not be located within the housing of the first storage subsystem 100A. For example, a VOL that the other storage subsystem 100B or 100C has can be used as a VOL of the first storage subsystem 100A.

  Each DKA 120A controls data exchange with each storage device 150A. Each DKA 120A is configured as a microcomputer system including a CPU, a ROM, a RAM, and the like, for example. A plurality of each DKA 120A is provided in the first storage subsystem 100A. The DKA 120 performs block-level data transfer with the storage device 150A based on, for example, SCSI or iSCSI.

  Each of the plurality of CHAs 110A can be configured as a microcomputer system similarly to the DKA 120A. The plurality of CHAs 110A communicate data with one or more host CHAs 110HA that perform data communication with the host terminal 180A via the connection path 190A, and the other storage subsystems 100B and 100C via the connection paths 200A and 200B, respectively. The systems CHA110SA1 and 110SA2 to be performed are included. Note that at least one of the connection paths 190A, 200A, and 200B may be a communication network or a dedicated path line. The host CHA 110HA may be prepared according to the type of the host terminal 180 (for example, a server or a mainframe).

  The cache memory 130A can be composed of, for example, a volatile or non-volatile semiconductor memory. The cache memory 130A stores write data 2 (data written to the VOL) from the host terminal 180A and write data 2 read from the VOL 6A.

  The shared memory 140A can be configured from, for example, a nonvolatile or volatile semiconductor memory. The shared memory 140A stores, for example, various commands received from the host terminal 180A, first control information 141A used for controlling the first storage subsystem 100A, and the like. Commands, first control information 141A described later, and the like may be redundantly stored by the plurality of shared memories 140A. Note that the cache memory 130A and the shared memory 140A can be configured as separate memories, respectively, or a part of one memory can be used as a cache memory area and the rest can be used as a shared memory area. it can.

  The switching control unit 270A connects each DKA 120A, the host CHA 110HA, the systems CHA 110SA1 and SA2, the cache memory 130A, and the shared memory 140A to each other. The switching control unit 270A can be composed of, for example, an ultrahigh-speed crossbar switch.

  The SVP (Service Processor) 281A collects and monitors the state of each unit in the first storage subsystem 100A via, for example, an internal network (for example, LAN) 282A. The SVP 281A outputs the collected internal state information as raw data or as statistically processed data to an external management terminal (not shown). Examples of information that can be collected by the SVP 281A include device configuration, power supply alarm, temperature alarm, input / output speed, and the like. The system administrator can change the setting of the RAID configuration, block various types of packages (for example, CHA 110A and DKA 120A), etc. from the management terminal via the SVP 281A. Further, the SVP 281A may be remotely operated from the management terminal 109 via a communication network (for example, a LAN or the Internet) 108.

  Next, an example of processing performed by the first storage subsystem 100A will be described. The host CHA 110HA receives the write command and the write data 2 from the host terminal 180A via the connection path 190A. The received write command is stored in the shared memory 140A, and the received write data 2 is stored in the cache memory 130A. The DKA 120A refers to the shared memory 140A as needed. When the DKA 120A finds an unprocessed write instruction stored in the shared memory 140A, the DKA 120A reads the write data 2 from the cache memory 130A according to the write instruction, and performs address conversion and the like. The DKA 120A stores the write data 2 in each storage device 150A constituting the VOL 6A designated by the write command.

  A case where a read command from the host terminal 180A is processed will be described. When the host CHA 110HA receives a read command from the host terminal 180A, the host CHA 110HA stores the read command in the shared memory 140A. When the DKA 120A finds an unprocessed read instruction in the shared memory 140A, the DKA 120A reads the write data 2 from each storage device 150A constituting the VOL 6A designated by the read instruction. The DKA 120A stores the read write data 2 in the cache memory 130A. Further, the DKA 120A notifies the host CHA 110HA via the shared memory 140A that the reading of the requested write data 2 has been completed. The host CHA 110HA reads the write data 2 from the cache memory 130A and transmits it to the host terminal 180A.

  Data replication (hereinafter sometimes referred to as “remote copy”) performed between the first storage subsystem 100A and the second storage subsystem 100B via the connection path 200A (which may be referred to as a remote copy line). An example will be described. Note that the explanation is based on data replication performed between the first storage subsystem 100A and the third storage subsystem 100C via the connection path 200B, and between the second storage subsystem 100B and the third storage subsystem 100C. The present invention can also be applied to data replication that can be performed via the connection path 200C.

  Remote copy is performed in response to a write command or a read command transmitted / received between the storage subsystems 100A and 100B, not a write command or a read command from the host terminal 180A, thus eliminating the need for the host terminal 180A. Data replication processing.

  Specifically, for example, every time the PVOL 6A is updated, the control device 101A of the first storage subsystem 100A generates the above-described JNL3, stores it in the PJNLVOL 6A, and receives a read command from the second storage subsystem 100B. In the case (or when the write command is issued to the second storage subsystem 100B), JNL3 in the PJNLVOL 6A is transmitted to the second storage subsystem 100B through the connection path 200A. As a result, the JNL 3 is stored in the second storage subsystem 100B asynchronously with the storage of the JNL 3 in the first storage subsystem 100A. Further, the restore process using the JNL 3 is performed in the second storage subsystem 100B, so that the SVOL 6B becomes a copy of the PVOL 6A.

  The above is the configuration example of the storage subsystem 100 in the present embodiment. Needless to say, the storage subsystem 100 need not be limited to the configuration described above. For example, the connection path 200C may not be provided. Also, the management terminal 109 remotely operates the SVPs 281A to 281C of the storage subsystems 100A to 100C via the communication network 108, and the first to third control information 141A to 141C are stored in the storage subsystems 100A to 100C. May be registered respectively. The control device 101 is not limited to the above-described configuration, for example, a memory that can store control information, write data, and the like, an interface device for the host terminal 180 (hereinafter abbreviated as “I / F”), An I / F for other storage subsystems, an I / F for the storage device 150, and a control unit (for example, a CPU) that controls communication via those I / Fs based on information on the memory. Also good. In the storage subsystem 100, the first data transfer from the host terminal 180 (or another storage subsystem) to the cache memory 130 via the CHA 110 and the switching control unit 270, the switching control unit 270 from the cache memory 130, and The second data transfer to the host terminal (or other storage subsystem) via the CHA 110 is performed if the CHA 110 that controls the first data transfer is different from the CHA 110 that controls the second data transfer and / or data. If the transfer source cache and the transfer destination cache are different (for example, if the transfer source cache memory address and the transfer destination cache memory address are different), it can be performed simultaneously. Similarly, in the storage subsystem 100, third data transfer is performed from the storage device 150 to the cache memory 130 via the DKA 120 and the switching control unit 270, and from the cache memory 130 to the storage device 150 via the switching control unit 270 and the DKA 120. The fourth data transfer to be performed is different if the DKA 120 that controls the third data transfer is different from the DKA 120 that controls the fourth data transfer and / or the data transfer source and the transfer destination cache are different (for example, transfer If the original cache memory address and the transfer destination cache memory address are different), they can be performed simultaneously. In addition, if the transfer destination in the first data transfer or the transfer source in the second data transfer is different from the transfer destination in the third data transfer and / or the transfer source in the fourth data transfer, the first data The transfer or the second data transfer and the third data transfer and / or the fourth data transfer can be performed simultaneously. Further, in order to perform the simultaneous transfer as described above, each transfer path (for example, a transfer path between the CHA 110 and the switching control unit 270, a transfer path between the DKA 120 and the switching control unit 270, and the switching control unit 270) An appropriate bandwidth (transfer speed) is required for the transfer path to and from the cache memory 130. For example, when one switching control unit 270 is connected to two CHAs 110, two DKAs 120, and two cache memories 130, the bandwidth between the switching control unit 270 and the cache memory 130 is CHA 110 ( Alternatively, if the bandwidth between the DKA 120) and the switching control unit 270 is not 1 or more times, there is no merit of multiplex transmission, and it is desirable that there is a bandwidth that is 2 or more times. Further, the writing speed and reading speed of the storage device 150 and the writing speed and reading speed of the cache memory 130 are faster than the transfer speed of the path between the storage device 150 and the cache memory 130, and It is preferable that the speed is such that no underrun error or overrun error occurs. Further, the host terminals 180A to 180C and the storage subsystems 100A to 100C may be connected to the same communication network (for example, SAN (Storage Area Network)).

  Incidentally, the first control information 141A is stored in a memory that can be referred to from the CHA 110A and the DKA 120A, for example, the shared memory 140A. The first control information 141A may be content specific to the first storage control system 100A, or may be content common to all the storage subsystems 100A to 100C provided in the data processing system 1. The first control information 141A may be input from the SVP 281A, or may be input from the management terminal 109 via the communication network 108 and the SVP 281A. All or part of the first control information 141A input from the SVP 281A is stored in a centralized or distributed manner in at least one of the shared memory 140, the cache memory 130, the CHA 110, the DKA 120, and the storage device 150, for example. Also good. In the present embodiment, the first control information 141A is registered in the shared memory 140A from the CHA 110A or DKA 120A via the internal network 282A, for example. For example, as shown in FIG. 5, the first control information 141A includes VOL management data 400A, path management data 500A, and pointer management data 700A. These will be described below.

  FIG. 6A shows a configuration example of the VOL management data 400A when the multi-target method exemplified in FIG. 1A is adopted. FIG. 6B shows a configuration example of the VOL management data 400A when the multi-hop method exemplified in FIG. 1B is adopted. Note that, in FIG. 6B, a location different from the content shown in FIG. 6A is indicated by a dotted frame.

  The VOL management data 400A is data for managing a plurality of VOLs 6A. For example, VOL ID, VOL status, format format, VOL capacity (for example, gigabytes) and physical address associated with each VOL 6A are data. Contains as an element.

  The VOLID is an identifier for identifying the VOL 6A. The identifier is, for example, a number. 6A and 6B, the VOLID “1” is the VOL6A1 illustrated in FIGS. 1A and 1B, the VOLID “4” is the VOL6A2 illustrated in FIGS. 1A and 1B, and the VOLID “5”. Is the VOL6A3 illustrated in FIGS. 1A and 1B.

  The VOL state is a data element that represents the state of the VOL 6A, and can be represented by, for example, “normal”, “normal”, “secondary”, “abnormal”, or “unused”. The VOL 6A whose VOL state is “normal” or “normal” is a VOL 6A that can be normally accessed from the host terminal 180A. The VOL 6A whose VOL state is “secondary” is a VOL 6A to which access from the host terminal 180A may be permitted. The VOL 6A in which the VOL state is “positive” is a PVOL or a PJNLVOL. The VOL 6A in which the VOL state is “secondary” is SVOL or SJNLVOL. The VOL 6A whose VOL state is “abnormal” is a VOL 6A that cannot be normally accessed due to a failure. The “failure” mentioned here is, for example, a failure of the storage device 150A that holds the VOL 6A. A VOL 6A whose VOL status is “unused” indicates that the VOL 6A is not in use. Note that whether the VOL 6A whose VOL state is “primary” is PVOL or PJNLVOL, and whether the VOL 6A whose VOL state is “secondary” is SVOL or SJNLVOL, will be described later. It can be specified by referring to 501A.

  The VOL capacity represents the storage capacity of VOL 6A.

  The physical address is a physical storage position in the first storage subsystem 100A. For example, as shown in the figure, an ID (for example, a number) for identifying the storage device 150 in the storage subsystem 100A and the storage device 150 Of the storage area (for example, the position from the beginning of the storage area of the storage device 150). One VOL 6A may be a storage area provided in one storage device 150A or a plurality of storage devices 150A by converting logical addresses and physical addresses (in other words, associating). It may be a storage area.

  According to the VOL management data 400A illustrated in FIG. 6A, for example, a VOL 6A having a VOLID “1” has a VOL capacity of 3 GB and a storage device ID “1” from the beginning of the storage area of the storage device 150A. It can be seen that the data is stored and is PVOL or PJNLVOL. Further, according to the VOL management data 400A illustrated in FIG. 6B, it can be seen that the VOL 6A whose VOLID is “1” is SPVOL or SPJNLVOL.

  FIG. 7A shows a configuration example of the path management data 500A when the multi-target method exemplified in FIG. 1A is adopted. FIG. 7B shows a configuration example of the path management data 500A when the multi-hop method exemplified in FIG. 1B is adopted. In FIG. 7B, a different location from the content shown in FIG. 7A is indicated by a dotted frame.

  The path management data 500A includes pair management sub data 501A for managing a VOL pair and mirror management sub data 502A for managing a mirror pair.

  The pair management sub-data 501A includes, for example, a VOL pair ID, pair status, primary storage subsystem ID, PVOL-ID, PJNLVOL-ID, secondary storage subsystem ID, SVOL-ID, SJNLVOL associated with each VOL pair. -Includes ID and copied address as sub-data elements.

  The VOL pair ID is an identifier (for example, a number) for identifying the VOL pair.

  The VOL pair status is a sub-data element indicating the status of the VOL pair, and can be represented by, for example, “normal”, “abnormal”, “unused”, “not copied”, or “copying”. The VOL pair status “normal” indicates that data replication of the PVOL 6A is normally performed. The VOL pair status “abnormal” indicates that the PVOL 6A cannot be replicated due to a failure. The “failure” mentioned here is, for example, disconnection of the connection path 200. The VOL pair status “unused” indicates that the information on the pair number corresponding to the VOL pair status is not valid. The VOL pair status “copying” indicates that an initial copy process described later is being performed. The VOL pair status “copy not yet” indicates that an initial copy process described later has not been performed yet.

  The primary storage subsystem ID is an identifier for specifying the storage subsystem 100 that holds the PVOL. As the identifier, for example, at least one of a number, a WWN (World Wide Name), an iSCSI name, and a MAC address can be adopted.

  PVOL-ID is an identifier of PVOL.

  PJNLVOL-ID is an identifier of PJNLVOL associated with the PVOL.

  The secondary storage subsystem ID is an identifier for specifying the storage subsystem 100B that holds the SVOL.

  SVOL-ID is an identifier of SVOL.

  SJNLVOL-ID is an identifier of SJNLVOL associated with the SVOL.

  The copied address is used in an initial copy process (see FIG. 10) described later. The copied address will be described later.

  The mirror management sub-data 502A includes, for example, a mirror ID, PJNLVOL-ID, SJNLVOL-ID, JNL generation update number, JNL replication update number, and restore update number associated with each mirror pair as sub-data elements. .

  The mirror ID is an identifier (for example, a number) for identifying a mirror pair.

  PJNLVOL-ID is an identifier (for example, a number) for specifying PJNLVOL.

  SJNLVOL-ID is an identifier (for example, a number) for specifying SJNLVOL.

  The JNL generation update number indicates how many JNLs have been generated in the corresponding mirror pair (in other words, indicates the newest number among the update numbers in the generated JNL). For example, in FIG. 7A, the JNL generation update number corresponding to the mirror ID “1” is “12”. This is because the update number “11” is assigned to the PJNLVOL in the mirror pair corresponding to the mirror ID “1”. This means that next time, JNL3 including the update number “12” should be generated and stored in PJNLVOL.

  The JNL replication update number indicates how many JNL replications have been performed in the corresponding mirror pair (in other words, the JNL replication update number indicates the newest number among the update numbers in the replicated JNL). For example, in FIG. 7A, the JNL replication update number corresponding to the mirror ID “1” is “9”. This is because the update number “8” is assigned to the SJNLVOL in the mirror pair corresponding to the mirror ID “1”. This means that JNL3 including the update number “9” should be read from PJNLVOL next time.

  The restore update number indicates how many JNL-based restores have been performed in the corresponding mirror pair (in other words, indicates the newest update number in the JNL read out for restore). For example, when the restore update number is “8”, this means that the restore process based on the JNL3 with the update number “7” has already been completed. Therefore, next time, the JNL3 including the update number “8” will be SJNLVOL. Means that the restoration processing should be performed. Note that FIG. 7A is an example of the mirror management sub-data 502A provided in the first storage subsystem 100A that does not include the SVOL, so there is no meaning of the restore update number. In FIG. 7B, for the mirror ID “1”, a JNL generation update number is meaningless because no JNL is generated in the first storage subsystem 100A.

  Each CHA 110A and each DKA 120A of the first storage subsystem 100A store the write data from the host terminal 180A in which VOL by referring to the path management data 500A as described above, and from which VOL the update number is what number It is possible to specify which VOL3 should be read out and stored in which VOL, from which VOL the update number of JNL3 should be read out, and in which VOL the restoration process should be performed. For example, the path management data 500A in FIG. 7A is illustrated in FIG. 1A by referring to the sub data elements associated with the VOL pair IDs “1” and “2” and the mirror IDs “1” and “2”. The configuration of the multi-target method can be specified. Further, in the path management data 500A in FIG. 7B, the sub-data elements associated with the VOL pair IDs “1” and “2” and the mirror IDs “1” and “2” are referred to in FIG. 1B. The configuration of the multi-hop method can be specified.

  For example, when a failure occurs in the first host terminal 180A, the first control device 101A of the first storage subsystem 100A changes the content of the path management data 500A from the content illustrated in FIG. 7A to the content illustrated in FIG. 7B. (In other words, by inverting PVOL and SVOL corresponding to VOLID “1” and inverting PJNLVOL and SJNLVOL corresponding to mirror ID “1”) as illustrated in FIG. 1A The multi-target method can be switched to the multi-hop method illustrated in FIG. 1B. For example, when a failure occurs in the second host terminal 180B, the first control apparatus 101A changes the content of the path management data 500A from the content illustrated in FIG. 7B to the content illustrated in FIG. 7A ( In other words, by reversing the PVOL and SVOL corresponding to VOLID “1” and reversing the PJNLVOL and SJNLVOL corresponding to mirror ID “1”), the multi-hop method illustrated in FIG. 1B is illustrated. It is possible to switch to the multi-target method exemplified in 1A.

  For example, when the number of storage subsystems 100 provided in the data processing system 1 is four or more, if the contents of the path management data 500A are specific to the first storage subsystem 100A, the first storage subsystem 100A Cannot specify where the replication start VOL or the replication goal VOL exists, and therefore it may not be possible to specify all configurations of the multi-target method or the multi-hop method. However, each of the other storage subsystems 100 is provided with path management data 500 unique to the storage subsystem 100, and the path management data 500 includes information related to the JNL replication source and replication destination JNLVOL (for example, VOLID and storage subsystem ID) are registered, so that regardless of the number of storage subsystems 100, it is possible to realize replication processing according to at least one of the multi-target method, multi-hop method, and combinations thereof. .

  FIG. 8 shows a configuration example of the pointer management data 700. FIG. 9 shows a JNLVOL configuration specified from the pointer management data 700 illustrated in FIG.

  As shown in FIG. 8, the pointer management data 700 is data prepared for each JNLVOL. The pointer management data 700 includes, for example, an update data area start address, write data area start address, update data latest address, update data oldest address, write data latest address, write data oldest address, read start address, and retry start address. Include as a data element.

  The update data area start address is the top logical address of the storage area (update data area) for storing the update data 4 of JNLVOL.

  The write data area start address is the logical address of the start of the storage area (write data area) for storing the write data 2 of JNLVOL.

  The latest update data address indicates the beginning logical address used for storing the update data 4 in the JNL 3 when the JNL 3 is stored next (that is, where the update data 4 in the next JNL 3 starts to be written). Information).

  The update data oldest address is the first logical address in the area for storing the update data 4 of the oldest (small update number) JNL3.

  The latest write data address is the first logical address used to save the write data 2 in the JNL3 when storing the next JNL3 (that is, information indicating where to start writing the write data 2 in the next JNL3. ).

  The oldest write data address is the first logical address in the area for storing the write data 2 in the oldest (small update number) JNL3.

  The read start address and the retry start address are data elements used only in the first storage subsystem 100A, and are used in journal read reception processing described later. A detailed description of the read start address and the retry start address will be described later.

  According to the pointer management data 700 illustrated in FIGS. 8 and 9, the update data area 7 is a range from address 0 (start) to address 699 in VOL 6 with JNLVOL-ID “4”. , JNLVOL-ID “4” in VOL6, address 700 to address 2699. The update data 4 of a certain JNL3 is stored in the range of the address 200 to the address 499 in the VOL6 of the JNLVOL-ID “4”, and the update data 4 of the next update number JNL3 is the JNLVOL-ID “ It can be seen that data is written from the address 500 in the VOL6 of “4”. The write data 2 of a certain JNL3 is stored in the range of the address 1300 to the address 2199 in the VOL6 of the JNLVOL-ID “4”, and the write data 2 of the next JNL3 is the JNLVOL-ID “4”. As can be seen from the address 2200 in the VOL6.

  Now, referring to FIG. 10 to FIG. 20, the first storage subsystem 100A is a primary storage subsystem (that is, a storage subsystem having a PVOL), and the second storage subsystem 100B is a secondary storage subsystem (that is, a SVOL). The processing related to data replication from the first storage subsystem 100A to the second storage subsystem 100B will be described. In the following description, the PVOL included in the first storage subsystem 100A is “PVOL6A1,” the PJNLVOL included in the first storage subsystem 100A is “PJNLVOL6A2,” and the SJNLVOL included in the second storage subsystem 100B is “SJNLVOL6B1. The SVOL included in the second storage subsystem 100B is “SVOL6B2.”

  FIG. 10 is a flowchart of the initial copy process.

  The initial copy process is a process for preparing JNL3 for PVOL6A1 that has not been copied. In the initial copy process, JNL3 is created for each unit size in order from the beginning of the storage area using the copied address of the path management data 500A for all the storage areas of PVOL6A1. The copied address has an initial value of 0, and the created data amount is added every time JNL3 is created. From the beginning of the storage area of the VOL 6A to the previous address before the copied address, the JNL 3 has been created in the initial copy process. By performing the initial copy processing, it is possible to transfer the unupdated write data 2 of the PVOL 6A1 to the SVOL 6B2 that constitutes the VOL pair with the PVOL 6A1. In the following description, it is described that the host CHA 110HA in the first storage subsystem 100A performs the processing, but the DKA 120A may perform the processing instead.

  Based on the path management data 500A in the first storage subsystem 100A, the host CHA 110HA in the first storage subsystem 100A grasps the PVOL 6A1 whose VOL pair status is “not copied”, and the VOL pair related to the grasped PVOL 6A1 The state is changed to “copying”, and the following processing is repeated (steps 1010 and 1020). If there is no PVOL 6A1 whose VOL pair status is “not copied”, the host CHA 110HA ends the process (step 1030).

  If there is a PVOL 6A1 in the VOL pair status “not copied” in step 1020, the host CHA 110HA creates JNL 3 for the data of the unit size (for example, 1 MB). The journal creation process will be described later (step 1040).

  The host CHA 110HA adds the created JNL3 data size to the copied address (step 1050).

  The host CHA 110HA repeats the above processing until the copied address reaches the capacity of the PVOL 6A1 (Step 1060). When the copied address becomes equal to the capacity of PVOL6A1, JNL3 has been created for all storage areas of PVOL6A1, so the VOL pair status is updated to “normal” and processing of other PVOLs is started (step 1070). .

  In the flowchart described above, the PVOL is described as being targeted one by one. However, the JNL 3 may be simultaneously generated for a plurality of data respectively stored in the plurality of PVOLs.

  FIG. 11 shows an outline of the flow of the command reception process 210 performed by the first storage subsystem 100A. FIG. 12 shows a flowchart of the command reception process 210. FIG. 13 shows a flowchart of the JNL creation process performed by the first storage subsystem 100A. Hereinafter, the processing performed when the first storage subsystem 100A receives an access command for the PVOL 6A1 from the host terminal 180A will be described using these.

  The host CHA 110HA receives an access command from the host terminal 180A (step 1200). The access command includes, for example, an identifier indicating the type of command (for example, read, write, or JNL read described later), a logical address of a command target (for example, a write destination or a read source), a data amount, and the like. Hereinafter, the logical address specified from the access command received in step 1200 is “logical address“ A ””, the VOLID specified from the access command is “VOLID“ A ””, and the VOL specified from the access command is specified. The internal position is “VOL internal position“ A ””, and the data amount specified from the access command is “data amount“ A ””. Further, it is assumed that the VOL specified from the VOLID “A” is the VOL “A”.

  The host CHA 110HA checks the access command (Steps 1210 and 1215). If it is determined in step 1215 that the access instruction is a JNL read instruction, a JNL read reception process described later is performed (step 1220). If the access instruction is other than a JNL read instruction and a write instruction, for example, a read instruction, a read process corresponding to the read instruction is performed (step 1230).

  If the access instruction is a write instruction in the examination in step 1210, the host CHA 110HA refers to the VOL management data 400A and examines the VOL state of the VOL “A” identified from the write instruction (step 1240). If the VOL state of the VOL “A” is other than “normal” or “normal” in step 1240, the access to the VOL “A” is impossible, and the host CHA 110HA has an error in the host terminal 180. The end is reported (step 1245).

  If it is determined in step 1240 that the VOL state of the instruction specific VOL is “normal” or “normal”, the host CHA 110HA allocates a storage area of a certain size (hereinafter referred to as a cache area) on the cache memory 130. And notifies the host terminal 180A that it is ready to receive data. The host terminal 180A receives the notification and transmits the write data 2 to the first storage subsystem 100A. The host CHA 110HA receives the write data 2, and stores the write data 2 in the secured cache area (step 1250, 1100 in FIG. 11).

  The host CHA 110HA refers to the VOL management table 400A and the path management table 500A and checks whether or not the VOL “A” is PVOL 6A1 (step 1260). If an affirmative result is obtained in step 1260, the host CHA 110HA performs a JNL creation process described later (step 1265).

  If a negative result is obtained in the examination in Step 1260 (or after the JNL creation process in Step 1265 is completed), the host CHA 110HA instructs the DKA 120A to write the write data 2 to the VOL “A” ( In FIG. 11, 1140), an end report is sent to the host terminal 180A (steps 1270 and 1280). Thereafter, the DKA 120A that has received the write command for the write data 2 executes the read / write process 220, thereby saving the write data 2 in the cache area in the VOL “A” (1110 in FIG. 11).

  Next, the JNL creation process will be described with reference to FIG.

  The host CHA 110HA checks the VOL status of the JNLVOL 6A2 associated with the PVOL 6A1 based on the VOL management data 400A and the path management data 500A (Step 1310). If the VOL status of the JNLVOL 6A2 is “abnormal” as determined in step 1310, the host CHA 110HA terminates the process because the JNL 3 cannot be stored in the JNLVOL 6A2 (step 1315). In this case, the host CHA 110HA may perform processing such as changing the JNLVOL 6A2 to a normal VOL.

  If it is determined in step 1310 that JNLVOL 6A2 is normal, the host CHA 110HA continues the JNL creation process. The contents of the JNL creation process differ depending on whether the process is in the initial copy process or the instruction reception process 210 (step 1320). If the JNL creation process is a process in the instruction reception process 210, the host CHA 110HA performs the process from step 1330. If the JNL creation process is in the initial copy process, the host CHA 110 HA performs the process from step 1370.

  When the JNL creation process is a process in the instruction reception process 210, the host CHA 110HA checks whether the write target logical address “A” is a process target of the initial copy process (step 1330). If the VOL pair status of VOL “A” is “not copied”, the JNL creation process will be performed later in the initial copy process, so the host CHA 110 HA ends the process without creating JNL 3 (step 1335). When the VOL pair status of VOL “A” is “Copying”, if the copied address is equal to (or smaller than) the position “A” in the logical address, the JNL creation process is performed later in the initial copy process. Therefore, the host CHA 110HA ends the process without creating the JNL 3 (step 1335). Other than the above, that is, when the VOL pair status of VOL “A” is “copying” and the copied address is greater than or equal to the position “A” in the logical address (or the VOL pair status of VOL “A” is “normal” ) Since the initial copy process has already been completed, the host CHA 110HA continues the JNL creation process.

  Next, the host CHA110HA checks whether JNL3 can be stored in JNLVOL6A2. Specifically, the host CHA 110HA refers to the pointer management data 700 and checks whether there is an unused area in the update data area (step 1340). If the update data latest address of the pointer management data 700 is equal to the update data oldest address, there is no unused area in the update data area, so the host CHA 110HA terminates the process as JNL creation failure (step 1390).

  If there is an unused area in the update data area in step 1340, the host CHA 110HA uses the pointer management data 700 to check whether write data can be stored in the write data area (step 1345). When the sum of the write data latest address and the data amount “A” is equal to (or larger than) the write data oldest address, the host CHA 110 HA cannot store the data in the write data area, and terminates the processing as a JNL creation failure. (Step 1390).

  When JNL3 can be stored, the host CHA 110HA uses the latest update number (specifically, the latest update number of one or more update numbers stored in JNLVOL 6A2) and the storage location of the update data 4. A logical address and a logical address as a storage destination of the write data 2 are acquired, and update data 4 is created in the cache area. In addition, the host CHA 110HA sets a numerical value obtained by adding 1 to the acquired update number as a new update number in the pair management table 500A. The logical address that is the storage destination of the update data 4 is the latest update data address of the pointer management data 700, and the host CHA 110HA uses the value obtained by adding the size of the update data 4 as a new update data latest address in the pointer management data 700A. Set. The logical address that is the storage destination of the write data 2 is the latest write data address of the pointer management data 700A, and the host CHA 110HA uses a value obtained by adding the data amount “A” to the write data latest address as a new write data latest address. The pointer management data 700A is set.

  The host CHA 110HA sets the obtained numerical value, the time when the write command is received, the logical address A and the data amount “A” in the write command to the update data 4 (step 1350, 1120 in FIG. 11).

  The host CHA 110HA instructs the DKA 120A to write the update data 4 and the write data 2 of JNL3 into the JNLVOL 6A2, and terminates normally (step 1360, 1130, 1140, 1150 in FIG. 11).

  If the JNL creation process is a process within the initial copy process, the process from step 1370 is performed. The host CHA 110HA checks whether JNL 3 can be created. Specifically, the host CHA 110HA uses the pointer management data 700 to check whether there is an unused area in the update data area (step 1370). If the update data latest address of the pointer management data 700 is equal to the update data oldest address, there is no unused area in the update data area, so the host CHA 110HA terminates the process as JNL creation failure (step 1390). In the case of the initial copy processing shown in the present embodiment, JNL write data is read from the primary VOL and the write data area is not used, so it is not necessary to check the unused area of the write data area.

  If the JNL 3 can be created in step 1370, the host CHA 110HA acquires the update number set in the update data 4 (for example, the update number written in the pair management table 500A) and caches the update data 4 Create in the area. The host CHA 110HA sets a numerical value obtained by adding 1 to the latest update number as a new update number in the pair management table 500A. The logical address for storing the update data 4 is the position of the update data latest address of the pointer management data 700, and the host CHA 110HA adds a numerical value obtained by adding the size of the update data 4 to the pointer management data 700 as a new update data latest address. Set.

  The host CHA 110HA sets the acquired update number, the start time of this process, the logical address of the initial copy process target, and the like in the update data 4 created in the cache area (step 1380, 1120 in FIG. 11).

  The host CHA 110HA commands the DKA 120A to write the update data 4 to the JNLVOL 6A2 (in other words, commands to write from the new update data latest address in the JNLVOL 6A2), and ends normally (step 1385, FIG. 11). 1140, 1160).

  The above is an explanation of FIGS. 11 to 13.

  FIG. 14 is a diagram for explaining the operation (JNL read reception processing) of the host CHA 110HA of the first storage subsystem 100A that has received the JNL read command, and FIG. 15 is a flowchart of the JNL read reception processing. Hereinafter, the operation when the first storage subsystem 100A receives a JNL read command from the second storage subsystem 100B will be described using these.

  System CHA110SA1 receives an access command from system CHA110SB2. The access command includes an identifier indicating that it is a JNL read command, a read source (for example, VOLID of PPJNLVOL6A2), and presence / absence of a retry instruction (step 1220, 1410 in FIG. 14).

  System CHA110SA1 checks the state of PJNLVOL6A2 (step 1520). If it is determined in step 1520 that the VOL state of the PJNLVOL 6A2 is not “normal”, for example, “failure”, the system CHA 110SA1 ends the process (step 1525). The system CHA110SB2 ends the JNL read process.

  If the VOL state of the PJNLVOL 6A2 is “normal” in the check in step 1520, the system CHA110SA1 checks whether the JNL read instruction is a retry instruction (step 1530).

  If it is determined in step 1530 that the JNL read instruction is a retry instruction, the system CHA110SA1 transmits the previously transmitted JNL3 to the second storage subsystem 100B again. The system CHA 110SA1 secures the cache area, and instructs the DKA 120 to read the size information of the update data 4 from the retry start address of the pointer management data 700 into the cache area (1420 in FIG. 14).

  The DKA 120 executes the read / write process 220 in response to the instruction from the system CHA 110SA1, thereby reading the update data 4 from the PJNLVOL 6A2, storing it in the cache area, and notifying the instruction source system CHA 110SA1 of the end of reading ( 14 of FIG. 14).

  The system CHA110SA1 receives the notification of the completion of reading of the update data 4, acquires the logical address of the write data 2 and the data size of the write data from the update data 4 stored in the cache area, and then secures the cache area. The DKA 120 is instructed to read the write data from the acquired logical address into the secured cache area (step 1540, 1440 in FIG. 14).

  The DKA 120 reads the write data 2 from the PJNLVOL 6A2 (specifically, the instructed logical address) by the read / write processing 220, stores the write data 2 in the secured cache area, and the system CHA110SA1 that is the instruction source Is notified of the end of reading (1450 in FIG. 14).

  The system CHA110SA1 receives the read end notification of the write data, transmits the update data 4 and the write data 2 to the second storage subsystem 100B (that is, transmits JNL3), and releases the cache area holding the JNL3. Then, the process ends (step 1545, 1460 in FIG. 14).

  If it is not a retry instruction in the check in step 1530, the system CHA 110SA1 checks whether there is a JNL 3 that has not been transmitted, and if it exists, transmits the JNL 3 to the second storage subsystem 100B. The system CHA110SA1 compares the read start address of the pointer management data 700 with the latest update data address (step 1550).

  When the read start address is equal to the update data latest address, all the JNL3s have been transmitted to the second storage subsystem 100B, and therefore the system CHA110SA1 transmits "no JNL" to the second storage subsystem 100B (step 1560), the storage area of JNL3 transmitted to the second storage subsystem 100B at the time of the previous JNL read command is released (step 1590).

  In the JNL storage area release processing, the system CHA 110SA1 sets a retry start address as the oldest update data address of the pointer management data 700. When the update data oldest address becomes the write data area start address, the system CHA 110SA1 sets the update data oldest address to 0. The system CHA110SA1 changes the write data oldest address of the pointer management data 700 to a value obtained by adding the size of the write data transmitted in response to the previous read JNL command. When the oldest address of the write data becomes a logical address greater than the capacity of JNLVOL, the system CHA 110SA1 subtracts and corrects the write data area head address.

  If there is an untransmitted JNL in step 1550, the system CHA 110SA1 reserves a cache area and reads update data from the read start address of the pointer management data 700 into the secured cache area (in other words, , Reading information for a predetermined size from the read start address) (step 1420 in FIG. 14).

  The DKA 120 executes the read / write process 220 in response to the instruction, thereby reading the update data from the PJNLVOL 6A2, storing it in the cache memory 130, and notifying the instruction source system CHA110SA1 of the end of reading (1430 in FIG. 14). ).

  The system CHA 110SA1 receives the notification of the completion of reading the update data, acquires the logical address of the write data and the size of the write data from the read update data, secures the cache area, and is acquired from the acquired logical address The DKA 120 is instructed to read the write data of the specified size into the secured cache area (step 1570, 1440 in FIG. 14).

  The DKA 120 reads and writes the write data from the PJNLVOL6A2 (in other words, the instructed logical address of the JNLVOL6A2) in accordance with the instruction by performing the read / write process 220, and stores the write data in the reserved cache area. Then, the end of reading is notified to the system CHA110SA1 (1450 in FIG. 14).

  The system CHA 110SA1 receives the write data read completion notification, transmits the update data and the write data to the second storage subsystem 100B (step 1580), and releases the cache area holding the JNL (1460 in FIG. 14). ). Then, the system CHA 110SA1 sets a read start address as the retry start address of the pointer management data 700, and adds a new value to the pointer management data 700 by adding the update data size of the transmitted JNL to the read start address. Set as start address.

  The system CHA 110SA1 releases the storage area of the JNL transmitted to the second storage subsystem 100B when the previous JNL read command was processed (step 1590).

  The above is an explanation of FIGS. 14 to 15. In the JNL read reception process described above, the first storage subsystem 100A transmits JNL one by one to the second storage subsystem 100B, but simultaneously transmits a plurality of JNLs to the second storage subsystem 100B. May be. The number of JNLs to be transmitted in one JNL read command may be specified by the second storage subsystem 100B in the JNL read command, or specified by the user in the first storage subsystem 100A or the second storage subsystem 100B. May be. Further, the number of JNLs transmitted by one JNL read command is determined by the transfer capacity or load of the connection path 200A between the first storage subsystem 100A and the secondary storage subsystem 100B, and the first storage subsystem 100A or the second storage subsystem. 100B may change dynamically. In addition, the JNL transfer amount may be specified in consideration of the size of JNL write data instead of the number of JNLs. The transfer amount may also be changed dynamically.

  In the above-described JNL read reception process, JNL is read from the storage device 150 into the cache memory 130. However, when the JNL exists in the cache memory 130, the process may not be performed.

  Further, the JNL storage area release processing in the JNL read reception processing described above is performed at the time of processing of the next JNL read instruction, but it may be performed immediately after the JNL is transmitted to the second storage subsystem 100B. Further, the second storage subsystem 100B may set an update number that may be released in the JNL read instruction, and the first storage subsystem 100A may release the JNL storage area in accordance with the instruction.

  FIG. 16 is a diagram for explaining the outline of the JNL read instruction process 240, FIG. 17 is a flowchart of the JNL read instruction process 240, and FIG. 18 is a flowchart of the JNL storage process. Hereinafter, using these, the system CHA110SB2 of the second storage subsystem 100B reads the JNL from the PJNLVOL6A2 of the first storage subsystem 100A, and stores the JNL in the SJNLVOL6B1 in the second storage subsystem 100B based on the JNL The operation to be performed will be described.

  The system CHA110SB2 secures a cache area for storing the JNL, and an access instruction for a JNL read instruction (for example, an instruction including an identifier indicating a JNL read instruction, a VOLID of the PJNLVOL6A2, and a retry instruction) And the access command is transmitted to the first storage subsystem 100A (step 1700, 1610 in FIG. 16).

  The system CHA110SB2 receives the response and JNL of the first storage subsystem 100A (1620 in FIG. 16). The system CHA110SB2 examines the received response. If the response is “JNL absent”, the first storage subsystem 100A does not have JNL3 in the PJNLVOL6A2, and therefore after a certain time, the system CHA110SB2 A read JNL command is transmitted (steps 1720 and 1725).

  If the response of the first storage subsystem 100A is, for example, normal termination, the system CHA 110SB2 refers to the VOL management data 400B and checks the VOL status of the SJNLVOL 6B1 that is the replication destination (step 1740). If the VOL status of the SJNLVOL 6B2 is “abnormal”, the system CHA110SB2 ends the process because the JNL cannot be stored in the SJNLVOL 6B2 (step 1745).

  If it is determined in step 1740 that the VOL state of the SJNLVOL 6B1 is “normal”, the system CHA 110SB2 performs a JNL storage process 1800 described later. If the JNL storage process 1800 is completed normally, the system CHA 110SB2 transmits the next JNL read command (step 1760). Instead, the system CHA 110SB2 may generate and transmit the next JNL read command after a predetermined time has elapsed after the JNL storage processing 1800 has ended normally. The system CHA 110SB2 may periodically transmit the timing for transmitting the next JNL command at regular time intervals, or the number of received JNLs, the communication amount of the connection path 200, the second storage subsystem 100B. May be determined by the storage capacity of the JNLVOL held by the memory, the load of the second storage subsystem 100B, or the storage capacity of the JNL held by the first storage subsystem 100A (or the first storage subsystem) 100A pointer management data 700) may be acquired and determined based on the acquired storage capacity. The information transfer may be performed by a dedicated command or may be included in a response to the JNL read command. The subsequent processing is the same as that after step 1710.

  If the JNL storage process in step 1800 does not end normally, the unused area of SJNLVOL6B1 is insufficient, and therefore the system CHA110SB2 discards the received JNL and transmits a JNL read command for a retry instruction after a certain time (step 1755). ). Alternatively, the system CHA110SB2 may hold the JNL in the cache area and perform the JNL storage process again after a certain time. This is because an unused area may increase in the SJNLVOL 6B1 after a certain time by performing a restore process 250 described later. In the case of this method, it is not necessary to include the retry instruction in the JNL read instruction.

  Next, the JNL storage process 1800 shown in FIG. 18 will be described.

  System CHA110SB2 checks whether JNL can be stored in SJNLVOL6B1. Specifically, the system CHA110SB2 checks whether there is an unused area in the update data area of the SJNLVOL6B1 using the pointer management data 700B (see FIG. 5) provided in the second storage subsystem 100B (step 1810). ). If the update data latest address and the update data oldest address of the pointer management data 700 are equal, there is no unused area in the update data area, and the system CHA 110SB2 ends the process as JNL creation failure (step 1820).

  If there is an unused area in the update data area of SJNLVOL6B1 as a result of the check in step 1810, the system CHA110SB2 uses the pointer management data 700 to check whether write data can be stored in the write data area (step 1830). . If the sum of the data amount of the write data latest address and the received JNL write data is equal to or larger than the write data oldest address, the write data cannot be stored in the write data area, and the system CHA110SB2 treats this as a JNL creation failure. Is finished (step 1820).

  When the JNL can be stored, the system CHA 110SB2 changes the write data logical address included in the JNL update data received by the JNL read process 240. Specifically, the system CHA 110SB2 changes the write data logical address in the update data 4 to the latest write data address of the pointer management data 700B provided in the second storage subsystem 100B. The system CHA110SB2 changes the update data latest address of the pointer management data 700B to a value obtained by adding the size of the update data to the current update data latest address (step 1840).

  The system CHA110SB2 secures a cache area, stores the updated data updated and the received write data in the JNL in the secured cache area, and writes the updated data and write data to the SJNLVOL6B1. This is instructed to the DKA 120, and the processing is terminated as a successful JNL creation (step 1850, 1630 in FIG. 16). Thereafter, the DKA 120 writes the update data and write data stored in the cache area to the SJNLVOL 6B1 by the read / write process 220, and then releases the secured cache area (1640 in FIG. 16).

  In the JNL storage process described above, JNL is stored in SJNLVOL6B1 (in other words, storage device 150 having the JNL), but a certain amount of cache area is prepared in advance for JNL, and all the cache areas are used. Then, the JNL may be stored in the SJNLVOL 6B1 from all the cache areas. The cache area size for JNL may be specified from, for example, SVP 281B.

  FIG. 19 is a diagram for explaining the restore process 250, and FIG. 20 is a flowchart of the restore process 250. Hereinafter, an operation in which the host CHA 110HB of the second storage subsystem 100B updates data using JNL will be described using these. The restore processing 250 may be performed by another CHA 110B (for example, the system CHA 110SB2), or may be performed by the DKA 120 of the second storage subsystem 100B.

  The host CHA 110HB refers to the VOL management data 400B, the path management data 500B, etc., and checks the VOL state corresponding to the SJNLVOL 6B1 (step 2020). If the VOL status of JNLVOL 6B1 is “abnormal” as determined in step 2020, the host CHA 110HB ends the processing because access is impossible (step 2025).

  If the VOL status of the SJNLVOL 6B1 is “normal” in step 2020, the host CHA 110HB checks whether the JNL to be restored exists in the SJNLVOL 6B1. Specifically, the host CHA 110HB acquires the update data oldest address and the update data latest address of the pointer management data 700B and compares them. When the update data oldest address and the update data latest address are equal, the host CHA 110HB does not have JNL, so the restore process is temporarily ended, and after a certain time, the restore process is resumed (step 2030).

  If it is determined in step 2030 that there is a JNL to be restored, the host CHA 110HB performs the following process on the JNL having the oldest (minimum) update number. The JNL update data having the oldest (minimum) update number is stored from the oldest update data address of the pointer management data 700B. The host CHA 110HB reserves a cache area, and instructs the DKA 120B to read information corresponding to the update data size (that is, the update data itself) from the SJNLVOL 6B1 from the oldest address of the update data (1910 in FIG. 19).

  In response to the instruction, the DKA 120B reads the update data from the SJNLVOL 6B1 by the read / write processing 220, stores the update data in the secured cache area, and notifies the host CHA 110HB of the read completion (FIG. 19). 1920).

  The host CHA 110HB receives the notification of the completion of the read of the update data, acquires the logical address of the write data and the size of the write data from the update data in the cache area, secures the cache area, and allocates the size of the size from the logical address. The DKA 120B is instructed to read data (that is, one write data) from the SJNLVOL 6B1 (1930 in FIG. 19).

  In response to the command, the DKA 120B reads the write data from SJNLVOL6B2 (in other words, the specified logical address) by the read / write processing 220, stores the write data in the cache area, and notifies the host CHA 110HB of the end of the read. (Step 2040, 1940 in FIG. 19).

  The host CHA 110HB obtains the logical address of the SVOL 6B2 to be updated (in other words, the logical address of the write instruction (see FIG. 2)) from the update data, and writes the write data to the address of the SVOL 6B2 specified from the logical address. (Step 2050, 1950 in FIG. 19). In response to the instruction, the DKA 120 writes the write data stored in the cache area to the storage area of the storage device 150 corresponding to the logical address of the SVOL6B2 (logical address of the write instruction) by the read / write process 220, and The area is released, and the end of the write process is notified to the host CHA 110HB (1960 in FIG. 19).

  The host CHA 110HB receives a write processing completion notification from the DKA 120B and releases the JNL storage area. In the JNL storage area release processing, the host CHA 110HB adds the update data oldest address of the pointer management data 700B provided in the second storage subsystem 100B to the current update data oldest address plus the size of the update data. Change to The host CHA 110HB sets the write data area start address to 0 when the update data oldest address becomes the write data area start address. The host CHA 110HB changes the oldest write data address of the pointer management data 700B to a value obtained by adding the size of the written data to the oldest write data address. When the oldest address of the write data becomes a logical address greater than or equal to the capacity of SJNLVOL6B1, the host CHA 110HB subtracts and corrects the write data area start address. Thereafter, the host CHA 110HB starts the next restore process (step 2060).

  The above is an explanation of FIGS. In the restore process 250 described above, the JNL is read from the SJNLVOL 6B1 into the cache memory 130. However, if the JNL exists in the cache memory 130, the process may not be performed.

  In the JNL read reception process and the JNL read instruction process 240 described above, the second storage subsystem 100B may determine the JNL to be received. For example, the system CHA110SB2 adds an update number to the JNL read command. In this case, the system CHA 110SA1 that receives the JNL read command obtains the logical address of the update data including the update number designated by the second storage subsystem 100B in the JNL read reception process. In the shared memory 140 of 100A, a table or search method for obtaining a logical address storing update data from an update number may be provided.

  In the JNL read reception process and the JNL read instruction process 240 described above, the JNL read instruction is used, but a normal read instruction may be used. For example, the pointer management data 700A of the first storage subsystem 100A is transferred to the second storage subsystem 100B in advance, and the second storage subsystem 100B reads the JNL of the PJNLVOL6A2 of the first storage subsystem 100A. Good.

  Further, in the above-described JNL read reception process, it has been described that JNL is transmitted from the first storage subsystem 100A to the second storage subsystem 100B in the order of update numbers. A plurality of JNL read instructions may be transmitted from the first storage subsystem 100A to the second storage subsystem 100B. In this case, in order to process JNL in the order of update numbers in the restore process, the second storage subsystem 100B can be provided with a table or search method for obtaining a logical address storing update data from the update number.

  The above is the description of the embodiment relating to the basics of data processing using JNL. In the method described so far, JNL generation based on the original write data 2 stored in the PVOL, storage of the generated JNL in the PJNLVOL, JNL replication from the PJNLVOL to the SJNLVOL, and storage in the SJNLVOL Write data is restored to SVOL based on JNL. By applying such a mechanism, replication processing according to the multi-target method, replication processing according to the multi-hop method, dynamic switching between the multi-target method and the multi-hop method, and the like can be realized. Hereinafter, the multi-target method will be described in detail as a first example of the above-described embodiment, and then the multi-hop method will be described in detail as a second example of the embodiment.

  FIG. 21A shows an overview of replication processing that is normally performed in the data processing system according to the first example of one embodiment of the present invention, and FIG. 21B shows a failure in the first host terminal in the data processing system. FIG. 22 shows a flow of a switching process from the multi-target method to the multi-hop method performed when a failure occurs in the first host terminal. Hereinafter, differences from the above-described embodiment will be mainly described, and description of common points will be omitted or simplified.

  As shown in FIG. 21A, in the data processing system 1 according to the first embodiment, the first host terminal 180A and the first storage subsystem 100A connected thereto exist in the first site 840A and the second host terminal 180B. The second storage subsystem 100B connected to the second storage subsystem 100B exists at the second site 840B, and the third host terminal 180C and the third storage subsystem 100C connected thereto exist at the third site 840C.

  At normal times (for example, when no failure has occurred in the data processing system 1), as shown in FIG. 21A, the original data 2 written in the PVOL 6A1 of the first storage subsystem 100A is stored in the PVOL 6A1. As the replication start VOL, replication is performed along the two replication paths and the replication direction to the first SVOL 6B2 and the second SVOL 6C2, which are the respective replication goal VOLs in the two replication paths and the replication direction. That is, in the normal state, the VOL 6A1 of the first storage subsystem 100A is set as the replication start VOL, the VOL 6B2 of the second storage subsystem 100B, and the VOL 6C2 of the third storage subsystem 100C is set as the replication goal VOL. Processing is performed.

  In this case, when a failure occurs in the first host terminal 180A connected to the first storage subsystem 100A having the replication start VOL, the replication processing according to the multi-target method is switched to the replication processing according to the multi-hop method. Hereinafter, with reference to FIG. 21B and FIG. 22, a processing flow performed when switching from a replication process according to the multi-target method to a replication process according to the multi-hop method will be described.

  When a failure occurs in the first host terminal 180A (step S100), the data processing system 1 detects it. Specifically, for example, if the first storage subsystem 100A does not receive any response even if a predetermined signal is periodically transmitted to the first host terminal 180A, the first host terminal 180A may fail. The second host terminal 180B (or another device) may detect that a failure has occurred in the first host terminal 180A by a method such as heartbeat communication.

  When a failure occurs in the first host terminal 180A, a takeover process is performed in which the process of the first host terminal 180A is taken over by the second host terminal 180B (or the third host terminal 180C). The host terminal that takes over the processing may be determined in advance, or may be a host terminal connected to the storage subsystem having the SVOL with the slowest or fastest progress of the restore process. Hereinafter, the case where the process takeover destination is the second host terminal 180B will be described as an example.

  When a failure occurs in the first host terminal 180A, the processing of the first host terminal 180A is taken over by the second host terminal 180B (S101). Thereafter, the second host terminal 180B transmits a processing start instruction to the second storage subsystem 100B (S102).

  In response to the processing start instruction, the second storage subsystem 100B transmits a JNL read command for the first PJNLVOL6A2 to the first storage subsystem 100A, reads JNL3 from the first PJNLVOL6A2, and reads the read JNL3 to the first It stores in SJNLVOL6B1 (S103A). The second storage subsystem 100B repeats this process until all the JNL3 stored in the first PJNLVOL 6A2 has been read. The first storage subsystem 100A searches the first PJNLVOL 6A2 for a JNL that includes the same number as the JNL replication update number of the mirror management sub-data 502A, transmits the JNL to the second storage subsystem 100B, and the JNL replication update number. Increase the value of by 1. Further, when the JNL replication update number of the mirror management sub-data 502A and the JNL generation update number (for example, update number “16”) become the same, the first storage subsystem 100A stores the replication target in the first PJNLVOL 6A2. The second storage subsystem 100B may notify the second storage subsystem 100B that there is no JNL, so that the second storage subsystem 100B may recognize that the replication of all JNLs has been completed. In addition, when JNL3 is read from the first PJNLVOL 6A2, the first storage subsystem 100A may erase the read JNL3 from the first PJNLVOL 6A2. That is, when all JNLs are read from the first PJNLVOL 6A2, the first PJNLVOL 6A2 may be emptied.

  The second storage subsystem 100B restores to the first SVOL6B2 based on JNL3 stored in the first SJNLVOL6B1 at the same (or different) timing as writing the JNL3 read from the first PJNLVOL6A2 to the first SJNLVOL6B1. Is executed (S103B). The second storage subsystem 100B repeats this process until all the JNL3 stored in the first SJNLVOL 6B1 has been read.

  By performing the processes of S103A and S103B described above, the content of the replication goal VOL6B2 can be made completely the same as the content of the replication start VOL6A2.

  Thereafter, the second storage subsystem 100B executes a copy inversion process for inverting the replication direction in the VOL group including the SJNLVOL 6B1 (S104A). Specifically, for example, the second storage subsystem 100B generates a JNL copy inversion instruction including the mirror ID “1” of the mirror pair that the SJNLVOL 6B1 configures, and refers to the path management data 500B, and the SJNLVOL 6B1 The PJNLVOL constituting the mirror pair and the primary storage subsystem including the same are identified, and the generated JNL copy inversion instruction is transmitted to the identified primary storage subsystem (that is, the first storage subsystem) 100A. Further, for example, the second storage subsystem 100B includes the PJNLVOL-ID, PVOL-ID and primary storage subsystem ID related to the mirror ID “1”, the SJNLVOL-ID, SVOL-ID, and the secondary in the path management data 500B. Replace storage subsystem ID. Further, for example, the second storage subsystem 100B sets the JNL generation update number received by the copy inversion process described later of the first storage subsystem 100A in the path management data 500B in association with the inverted PJNLVOL 6B1.

  The first storage subsystem 100A that has received the JNL copy inversion instruction from the second storage subsystem 100B executes a copy inversion process in response to the instruction (S104B). Specifically, for example, the first storage subsystem 100A refers to the path management data 500A and notifies the second storage subsystem 100B of the JNL generation update number (for example, update number “16”) related to the PJNLVOL6A2. Further, for example, the first storage subsystem 100A includes the PJNLVOL-ID, PVOL-ID and primary storage subsystem ID related to the mirror ID “1”, the SJNLVOL-ID, SVOL-ID, and the secondary in the path management data 500A. Replace storage subsystem ID. Also, the first storage subsystem 100A sends a JNL read instruction for the inverted PJNLVOL6B1 to the second storage subsystem 100B, thereby reading the JNL in the PJNLVOL6B1, and storing the JNL in the inverted SJNLVOL6A2. Also good.

  The second storage subsystem 100B transmits an availability notification to the second host terminal 180B (S105). Note that the second storage subsystem 100B may notify the VOL-ID of the SVOL 6B2 to the second host terminal 180B so that the second host terminal 180B can send a write command to the VOL 6B2.

  Thereafter, the duplication processing according to the multi-hop method shown in FIG. 21B is performed. Specifically, for example, in the VOL group including the inverted PVOL 6B2, the following replication process is performed.

  That is, the second host terminal 180B generates the write data after receiving the usable notification by the processing of S104 of the second storage subsystem 100B, and has the generated write data and the VOL-ID of the inverted PVOL6B2. The write command is transmitted to the second storage subsystem 100B. The second storage subsystem 100B stores the received write data in the inverted PVOL 6B2, and the write data and the JNL generation update number (for example, update number “16”) notified from the first storage subsystem 100A. Is generated, and the JNL is stored in the inverted PJNLVOL6B1.

  Further, the first storage subsystem 100A reads the JNL in the SJNLVOL 6A2 at the same timing (or different) from the timing when the JNL is stored in the inverted SJNLVOL 6A2, and the write data included in the JNL is read The data is stored in SVOL (specifically, SVPOL) 6A1.

  With the above flow, in the VOL group including the inverted PVOL 6B2, the write data is duplicated from the inverted PVOL 6B2 to the inverted SVOL 6A1. In another VOL group that has not been subjected to the copy inversion process, the same process as in the normal state is performed as shown in FIGS. 21A and 21B.

  As described above, in the first embodiment, when a failure occurs in the first host terminal 180A connected to the first storage subsystem 100A having the replication start VOL, the path management data 500A of the first storage subsystem 100A, and the first Based on the path management data 500B of the secondary storage subsystem 100B, the VOL attributes of the JNLVOL 6A2 associated with the replication start VOL 6A1 and the JNLVOL 6B1 associated with the replication goal VOL 6B2 are inverted, and the replication is performed accordingly. The VOL attributes of the start VOL 6A1 and the replication goal VOL 6B2 are also inverted. By this process, the duplication direction is reversed. At that time, until the JNL generation update number for the PJNLVOL before the inversion matches the JNL replication update number and the restore update number for the SJNLVOL before the inversion, the JNL replication processing from the PJNLVOL6A2 to the SJNLVOL6B1, and the SJNLVOL6B1 to SVOL The contents of the replication start VOL before inversion and the contents of the replication goal VOL before inversion are made the same. As described above, in the first embodiment, when a failure occurs in the first host terminal 180A, the replication direction of one of the two VOL groups 16 and 16 including the PVOL 6A1 is automatically set. Inverted to maintain a reliable and redundant replication process.

  The above is the description of the first embodiment. An abstract description of the first embodiment is as follows. That is, the first storage subsystem 100A includes a first storage device 6A1 and one or more second storage devices 6A2 and 6A3, and the second storage subsystem 100B includes a third storage device 6B1 and a fourth storage device. The third storage subsystem 100C includes a fifth storage device 6C1 and a sixth storage device 6C2. The first storage subsystem 100A generates a data set including an update number indicating the update order of the first storage device 6A1 and the write data stored in the first storage device 6A1, and one or more second storage devices 6A2 and 6A3, and the data set is transmitted to the second and third storage subsystems 100B and 100C. Each of the second and third storage subsystems 100B, 100C stores the received data set in the third or fifth storage device 6B1, 6C1, and the data from the third or fifth storage device 6B1, 6C1 according to the update number. The set is read, and the write data in the data set is stored in the fourth or sixth storage device 6B2, 6C2.

  By the way, several modifications can be considered for the first embodiment. Hereinafter, these modifications will be described.

  (A) A first modification of the first embodiment.

  FIG. 23 shows an outline of replication processing after a failure occurs in the first host terminal 180A in this first modification example when replication processing according to the multi-target method shown in FIG. 21A is performed. Reference numeral 24 denotes a flow of switching processing from the multi-target method to another multi-target method performed when a failure occurs in the first host terminal 180A. Hereinafter, differences from the above-described first embodiment will be mainly described, and description of common points will be omitted or simplified.

  In “another multi-target method” referred to in the first modification, a plurality of PJNLVOLs are not associated with one PVOL, but as shown in FIG. 23, one PJNLVOL 6B1 is associated with one PVOL6B2. Are associated with each other, and a plurality of SJNLVOLs 6A2 and 6C1 are associated with one PJNLVOL6B1.

  In this first modification, as shown in FIG. 24, when a failure occurs in the first host terminal 180A, the same processing as S101 to S105 is performed, and in addition, the VOL group restructuring processing (S106) is performed. Done. This will be described in detail below.

  The first storage subsystem 100A identifies the VOLID of the PJNLVOL6B1 after copy inversion and the VOLID of the second SJNLVOL6C1 from the path management data 500A, and in other words, a mirror formation instruction for associating these two VOLIDs, in other words , An instruction to form a mirror pair of PJNLVOL6B1 and second SJNLVOL6C1 is transmitted to the second storage subsystem 100B and the third storage subsystem 100C. Further, the first storage subsystem 100A identifies the VOLID of PJNLVOL6A3 and the VOLID of SJNLVOL6C1 from the path management data 500A, and cancels the correspondence between these two VOLIDs, in other words, PJNLVOL6A3. An instruction to release the mirror pair of SJNLVOL6C1 is transmitted to the third storage subsystem 100C.

  In response to the mirror formation instruction from the first storage subsystem 100A, the second storage subsystem 100B sets information representing the mirror pair of the PJNLVOL 6B1 and SJNLVOL 6C1 in the path management data 500B.

  In response to the mirror cancellation instruction from the first storage subsystem 100A, the third storage subsystem 100C removes information representing the mirror pair of the PJNLVOL 6A3 and SJNLVOL 6C1 from the path management data 500B. Further, in response to the mirror formation instruction from the first storage subsystem 100A, the third storage subsystem 100C sets information representing the mirror pair of PJNLVOL 6B1 and SJNLVOL 6C1 in the path management data 500C.

  Through the series of processes described above, a new VOL group is constructed in which the replication start VOL is set to VOL6B2, the relay JNLVOL is set to 6B1 and 6C1, and the replication goal VOL is set to VOL6C2.

  Thereafter, replication processing according to another multi-target method is executed.

  For example, the second host terminal 180B generates the write data after receiving the usable notification by the process of S104 of the second storage subsystem 100B, and generates the generated write data and the write command for the PJNLVOL6B2 in the second storage. Transmit to subsystem 100B. The second storage subsystem 100B stores the received write data in the PVOL 6B2, and includes the write data and the JNL generation update number (for example, update number “16”) notified from the first storage subsystem 100A. JNL is generated, and the JNL is stored in PJNLVOL6B1.

  Based on the updated path management data 500A, the first storage subsystem 100A transmits a JNL read command for the PJNLVOL 6B1 to the second storage subsystem 100B, and in response to the command, the JNL in the PJNLVOL 6B1 is stored in the second storage subsystem 100A. The received JNL is received from the subsystem 100B and stored in the SJNLVOL 6A2 constituting the mirror pair with the PJNLVOL 6B1. The first storage subsystem 100A reads the JNL in the SJNLVOL 6A2 at the same (or different) timing as the JNL is stored in the SJNLVOL 6A2, and stores the write data included in the JNL in the SVOL 6A1.

  Based on the updated path management data 500C, the third storage subsystem 100C transmits a JNL read command for the PJNLVOL 6B1 to the second storage subsystem 100B, and in response to the command, stores the JNL in the PJNLVOL 6B1 in the second storage The received JNL is received from the subsystem 100B and stored in the SJNLVOL 6C1 that constitutes the mirror pair with the PJNLVOL 6B1. Further, the third storage subsystem 100C reads the JNL in the SJNLVOL 6C1 at the same timing (or different) from the timing at which the JNL is stored in the SJNLVOL 6C1, and stores the write data included in the JNL in the SVOL 6C2.

  As described above, in the first modification of the first embodiment, when a failure occurs in the first host terminal 180A, the replication direction of one of the two VOL groups 16 and 16 including the PVOL 6A1. Are automatically inverted, and a plurality of SJNLVOLs are associated with the inverted PJNLVOL, whereby another multi-target method is constructed and the duplication process is continued. Thereby, even if a failure occurs in the first host terminal 180A, the entire data processing system 1 does not go down, so that highly reliable replication processing is maintained.

  (B) A second modification of the first embodiment.

  FIG. 25 shows an overview of replication processing after a failure occurs in the first storage subsystem 100A in the replication processing according to the multi-target method shown in FIG. 21A. FIG. 26 shows a failure in the first storage subsystem 100A. The flow of processing performed when it occurs is shown.

  When a failure occurs in the first storage subsystem 100A (S110), the processing of the first host terminal 180A is taken over by the second host terminal 180B (S111). Note that a failure has occurred in the first storage subsystem 100A, for example, a certain time after the second storage subsystem 100B or the third storage subsystem 100C sends a JNL read command to the first storage subsystem 100A. If JNL cannot be received in response to the elapse of time, it can be determined that a failure has occurred in the first storage subsystem 100A. In this case, the storage subsystem that made the determination notifies the host terminal connected to the storage subsystem to that effect, and the host terminal that received the notification performs a takeover process with the first host terminal 100A. Also good.

  Thereafter, a VOL group reconstruction process is performed (S112). Hereinafter, a specific example of this process will be described.

  The second storage subsystem 100B connected to the second host terminal 100B that has taken over the processing from the first host terminal 100A changes the attribute of JNLVOL6B1 from SJNLVOL to PJNLVOL, and accordingly the attribute of VOL6B2 is changed from SVOL to PVOL. The information about the new mirror pair by PJNLVOL6B1 and SJNLVOL6C1 is added to the path management data 500B. Further, the second storage subsystem 100B changes the following pair partner change instruction, that is, a pair of changing the mirror pair partner of SJNLVOL6C1 from the second PJNLVOL6A3 to PJNLVOL6B1, and changing the VOL pair partner of SVOL6C2 from PVOL6A1 to PVOL6B2. A partner change instruction is transmitted to the third storage subsystem 100C. In response to the pair partner change instruction from the second storage subsystem 100B, the third storage subsystem 100C changes the contents of the path management data 500C, the mirror pair partner of SJNLVOL6C1 is PJNLVOL6B1, and the VOL pair partner of SVOL6C2 The content is updated to indicate that it is PVOL6B2.

  Through the series of processes described above, a new VOL group is constructed in which the replication start VOL is set to VOL6B2, the relay JNLVOL is set to 6B1 and 6C1, and the replication goal VOL is set to VOL6C2. Thereafter, processing according to the new VOL group is executed. For example, when the second storage subsystem 100B generates a JNL to be stored in the PJNLVOL6B1 for the first time after a new VOL group is constructed, the JNL includes the next update number of the one or more JNLs in the PJNLVOL6B1. Include update number. Further, when the second storage subsystem 100B first reads the JNL in the PJNLVOL6B1 and transmits it to the third storage subsystem 100C after constructing a new VOL group, the second storage subsystem 100B has the update number specified for the third storage subsystem 100C. The JNL or the JNL having the oldest update number among one or more JNLs in the PJNLVOL6B1 is read and transmitted to the third storage subsystem 100C. When the third storage subsystem 100C receives the JNL having the oldest update number, if the JNL including the update number exists in the SJNLVOL6C1 (or if the restore process based on the JNL has been completed) The received JNL is discarded (whether or not this is the case can be determined by referring to the path management data 500C, for example). Further, when the third storage subsystem 100C receives the JNL, the update number in the JNL is a number next to the latest update number (eg, “7”) of one or more JNLs in the SJNLVOL6C1 (eg, “8”). ”), The JNL is stored in SJNLVOL6C1. In addition, when the third storage subsystem 100C receives the JNL having the oldest update number, the oldest update number is the next update number (for example, “7”) of one or more JNLs in the SJNLVOL6C1. If the number is larger than the number (for example, “9”), the replication process is stopped. This is because restoration processing in the order of consecutive update numbers cannot be performed.

  As described above, in the second modification of the first embodiment, when a failure occurs in the first storage subsystem 100A, a new VOL group from which the VOL provided in the first storage subsystem 100A is removed is reconstructed. The replication process is continued in accordance with the new VOL group.

  (C) A third modification of the first embodiment.

  FIG. 27 shows an overview of the duplication processing according to the multi-target method according to the third modification of the first embodiment.

  In this third modified example, a plurality of PJNLVOLs are not associated with one PVOL6A1, but as shown in FIG. 27, one PJNLVOL6A2 is associated with one PJNLVOL6B1, and one PJNLVOL6B1 is associated with one PJNLVOL6B1. A plurality of SJNLVOLs 6B1 and 6C1 are associated with each other. In this case, both the second storage subsystem 100B and the third storage subsystem 100C transmit a JNL read command to the PJNLVOL 6A2, thereby receiving the JNL read from the PJNLVOL 6A2 from the first storage subsystem 1000A.

  According to the third modification, the first storage subsystem 100A does not need to create a plurality of JNLs for one original write data 2, so the load on the first storage subsystem 100A can be reduced. Further, according to the third modification, it is sufficient to prepare one PJNLVOL 6A2 for one PVOL 6A1, so that the storage capacity can be reduced.

  (D) Fourth modification of the first embodiment.

  FIG. 28A shows an example of a case where JNL replication from the first storage subsystem 100A to the second storage subsystem 100B can no longer be performed in the replication processing according to the multi-target method shown in FIG. 21A.

  In the replication process according to the multi-target method shown in FIG. 21A, there may be a case where JNL cannot be replicated from the first storage subsystem 100A to the second storage subsystem 100B. In this case, for example, when a failure occurs in the connection path 200A that connects the first storage subsystem 100A and the second storage subsystem 100B to each other, or as shown in FIG. 28B, the second storage subsystem A case where the JNL that is the read target of 100B does not exist in the PJNLVOL 6A2 can be considered. In addition, the situation where the JNL that is the read target of the second storage subsystem 100B does not exist in the PJNLVOL6A2 is such that, for example, the JNL with the oldest update number is deleted from the PJNLVOL6A2 after the JNL becomes full in the PJNLVOL6A2. (This also applies to the replication process according to the multi-hop method).

  In such a case, as shown in FIG. 28, even if the first storage subsystem 100A stops replicating the JNL to the second storage subsystem 100B, the JNL to the third storage subsystem 100C is stopped. Replication continues.

  According to the fourth modified example of the first embodiment, even if the replication process according to a certain replication path and the replication direction is stopped, the replication process according to another replication path and the replication direction is not affected.

  (E) A fifth modification of the first embodiment.

  FIG. 29 shows an overview of a replication process according to the multi-target method according to the fifth modification of the first embodiment.

  According to the multi-target method according to the fifth modification, in the first storage subsystem 100A, the third PJNLVOL 6A4 is prepared in the PVOL 6A1, and the third SJNLVOL 6D1 provided in the fourth storage subsystem 100D is associated with the third PJNLVOL 6A4. It is done. The third SJNLVOL 6D1 is associated with the third SVOL 6D2.

  Based on one embodiment and first example of the present invention, depending on which PJNLVOL is associated with which PVOL, which SJNLVOL is associated with which PJNLVOL, a mirror pair is configured, and which SVOL is associated with which SJNLVOL Regardless of the number of subsystems, replication processing according to the multi-target method can be realized.

  Next, the second example of the embodiment of the present invention, that is, the multi-hop method will be described in detail.

  FIG. 30A shows an overview of replication processing according to the multi-hop method that is normally performed in the data processing system according to the second example of one embodiment of the present invention, and FIG. 30B shows the second host terminal in the data processing system. FIG. 31 shows a flow of a switching process from the multi-hop method to the multi-target method performed when a failure occurs in the second host terminal.

  As shown in FIG. 30A, in normal times (for example, when no failure has occurred in the data processing system 1), the original data 2 written in the PVOL 6B2 of the second storage subsystem 100B is stored in the PVOL 6B2. As a replication start VOL and VOL6C2 of the third storage subsystem 100C as a replication goal VOL, the data flows downstream along one replication path and replication direction. Specifically, for example, the second storage subsystem 100B generates JNL3 based on the original write data 2 written in PVOL6B2, and stores the JNL3 in PJNLVOL6B1. The first storage subsystem 100A transmits a JNL read command to the PJNLVOL 6B1, and in response, receives the JNL read from the PJNLVOL 6B1 from the second storage subsystem 100B and stores it in the SJNLVOL 6A2. For example, the first storage subsystem 100A restores the write data 2 included in the JNL3 to the SPVOL 6A1 at the same timing as the storage of the JNL3 in the SJNLVOL 6A2. Further, for example, the first storage subsystem 100A generates a JNL including the write data 2 and the update number corresponding thereto at the same (or different) timing as the restore processing of the write data 2 to the SPVOL 6A1. The JNL is stored in the PJNLVOL6A3. That is, in the same site 140A, the JNL3 replication to the SJNLVOL6A2, the restoration of the write data 2 in the JNL3 to the SPVOL6A1, the generation of the JNL3 including the restored write data 2 and the storage to the PJNLVOL6A3 are as follows: It is performed at substantially the same timing (may be performed at different timings). The JNL stored in the PJNLVOL 6A3 is read by the JNL read instruction from the third storage subsystem 100C, and the read data 2 in the read JNL is the replication goal VOL (SVOL) that constitutes the VOL pair with the SPVOL 6A1. Restored to 6C2.

  In this case, when a failure occurs in the second host terminal 180B connected to the second storage subsystem 100B having the replication start VOL, the replication processing according to the multi-hop method is switched to the replication processing according to the multi-target method. Hereinafter, with reference to FIG. 30B and FIG. 31, a processing flow performed when switching from a replication process according to a multi-hop scheme to a replication process according to a multi-target scheme will be described.

  When a failure occurs in the second host terminal 180B (S200), the data processing system 1 detects it. As the detection method, the same method as in the first embodiment can be adopted.

  When a failure occurs in the second host terminal 180B, a takeover process is performed in which the process of the second host terminal 180B is taken over by the first host terminal 180A (or the third host terminal 180C). The host terminal that takes over the processing may be determined in advance, or may be a host terminal connected to the storage subsystem having the SVOL with the slowest or fastest progress of the restore process. Hereinafter, a case where the process takeover destination is the first host terminal 180A will be described as an example.

  When a failure occurs in the second host terminal 180B, the processing of the second host terminal 180B is taken over by the first host terminal 180A (S201). Thereafter, the first host terminal 180A transmits a process start instruction to the first storage subsystem 100A (S202).

  In response to the processing start instruction, the first storage subsystem 100A transmits a JNL read command for the PJNLVOL6B1 to the second storage subsystem 100B, reads JNL3 from the PJNLVOL6B1, and stores the read JNL3 in the SJNLVOL6A2 (S203A ). The first storage subsystem 100A repeats this process until all the JNL3 stored in the PJNLVOL6B1 has been read. The second storage subsystem 100B searches the PJNLVOL6B1 for a JNL containing the same number as the JNL replication update number of the mirror management sub-data 502B, sends it to the first storage subsystem 100A, and the value of the JNL replication update number Increase by one. Further, when the JNL replication update number of the mirror management sub-data 502B and the JNL generation update number (for example, update number “16”) become the same, the second storage subsystem 100B stores the JNL to be replicated in the PJNLVOL6B1. May be notified to the first storage subsystem 100A, so that the first storage subsystem 100A may recognize that all of the JNL replication has been completed. Further, when the JNL3 is read from the PJNLVOL 6B1, the second storage subsystem 100B may erase the read JNL3 from the PJNLVOL 6B1. That is, when all JNLs are read from the PJNLVOL 6B1, the PJNLVOL 6B1 may be emptied.

  The first storage subsystem 100A executes a restore process to SPVOL6A1 based on JNL3 stored in SJNLVOL6A2 at the same (or different) timing as writing JNL3 read from PJNLVOL6B1 to SJNLVOL6A2 (S203B). The first storage subsystem 100A repeats this process until all the JNL3 stored in the SJNLVOL 6A2 has been read.

  By performing the processes of S203A and S203B described above, the contents of SPVOL6A1 can be made completely the same as the contents of replication start VOL6B2.

  Thereafter, the first storage subsystem 100A transmits an availability notification to the first host terminal 180A (S205). The first storage subsystem 100A may notify the first host terminal 180A of the VOL-ID of the VOL 6A1 so that the first host terminal 180A can send a write command to the VOL 6A1.

  Thereafter, the duplication process according to the multi-target method shown in FIG. 30B is performed. Specifically, for example, in the VOL group including the inverted SVOL 6B2, the following replication process is performed.

  That is, the first host terminal 180A receives the usable notification by the process of S204 of the first storage subsystem 100A, generates write data, and has the generated write data and the VOL-ID of the inverted PVOL6A1. The write command is transmitted to the first storage subsystem 100A. In that case, the duplication processing according to the multi-target method described with reference to FIGS. 1A and 21A is executed.

  As described above, in the second embodiment, when a failure occurs in the second host terminal 180A connected to the second storage subsystem 100B having the replication start VOL, the path management data 500B of the second storage subsystem 100B, Based on the path management data 500A of the single storage subsystem 100A, the VOL attributes of the JNLVOL 6B1 associated with the replication start VOL 6B2 and the JNLVOL 6A2 associated with the SPVOL 6A1 are inverted, and accordingly, the replication start VOL 6B2 And the VOL attribute of SPVOL6A1 are also inverted. By this process, the duplication direction is reversed. At that time, until the JNL generation update number for the PJNLVOL before the inversion matches the JNL replication update number and the restore update number for the SJNLVOL before the inversion, the JNL replication processing from the PJNLVOL6B1 to the SJNLVOL6A2, and the SJNLVOL6A1 to SPVOL The contents of the replication start VOL before inversion and the contents of SPVOL are made the same. As described above, in the second embodiment, when a failure occurs in the second host terminal 180A, the replication direction of the VOL group 16 including the PVOL 6B2 is automatically reversed, so that there is a highly reliable redundancy. Replication processing is maintained.

  The above is the description of the second embodiment. Note that several variations of the second embodiment are possible. Hereinafter, these modifications will be described.

  (A) A first modification of the second embodiment.

  FIG. 32 shows an overview of replication processing after a failure has occurred in the second host terminal 180B in this first modification when replication processing according to the multi-hop method shown in FIG. 30A is being performed. 33 shows the flow of the switching process from the multi-hop method to another multi-hop method performed when a failure occurs in the second host terminal 180B. Hereinafter, differences from the above-described second embodiment will be mainly described, and description of common points will be omitted or simplified.

  In the “another multi-hop scheme” referred to in the first modification, the replication route is the same as the replication route illustrated in FIG. 30A as shown in FIG. 32, but the replication direction is the replication example illustrated in FIG. 30A. It is opposite to the direction. In other words, when the replication direction of each VOL group is reversed, the replication start VOL and the replication goal VOL are reversed, and the attributes of each VOL in each VOL group are also reversed.

  In the first modification, when a failure occurs in the second host terminal 180B connected to the second storage subsystem 100B having the replication start VOL, the first storage subsystem 100C connected to the third storage subsystem 100C having the replication goal VOL. The processing of the second host terminal 180B is taken over by the three host terminals 180C (S211). Thereafter, the third host terminal 180C transmits a processing start instruction to the first storage subsystem 100A and the third storage subsystem 100C (S212).

  In response to the processing start instruction to the first storage subsystem 100A, the same processing as S203A and S203B described above is performed (S213A and S213B). In addition, the first storage subsystem 100A generates a JNL based on the write data 2 at the same timing as when the write data 2 included in the JNL3 to the SJNLVOL6A2 is restored to the SPVOL6A1, and generates the JNL as PJNLVOL6A3. (S213C). As a result, the latest update number in one or more JNLs stored in the SJNLVOL 6A2 and the latest update number in one or more JNLs stored in the PJNLVOL 6A3 are the same.

  In response to the processing start instruction, the third storage subsystem 100C transmits a JNL read command for the PJNLVOL6A3 to the first storage subsystem 100A, reads JNL3 from the PJNLVOL6A3, and stores the read JNL3 in the SJNLVOL6C1 (S213D ). The third storage subsystem 100C repeats this process until all the JNL3 stored in the PJNLVOL6A3 has been read. The first storage subsystem 100A searches the PJNLVOL6A3 for a JNL containing the same number as the JNL replication update number of the mirror management sub-data 502A, sends it to the third storage subsystem 100C, and the value of the JNL replication update number Increase by one. Further, when the JNL replication update number of the mirror management sub-data 502A and the JNL generation update number (for example, update number “16”) are the same, the first storage subsystem 100A stores the JNL to be replicated in the PJNLVOL6A3. May notify the third storage subsystem 100C that it does not exist, so that the third storage subsystem 100C may recognize that all JNL replications have been completed. Also, when JNL3 is read from PJNLVOL 6A3, first storage subsystem 100A may erase the read JNL3 from PJNLVOL 6A3. That is, when all the JNLs are read from the PJNLVOL 6A3, the PJNLVOL 6A3 may be emptied.

  The third storage subsystem 100C executes a restore process to SVOL6C2 based on JNL3 stored in SJNLVOL6C1 at the same (or different) timing as writing JNL3 read from PJNLVOL6A3 to SJNLVOL6C1 (S213E). The third storage subsystem 100C repeats this process until all the JNL3 stored in the SJNLVOL6C1 has been read.

  By performing the processes of S213A to S213E described above, the contents of SVOL6C2 can be made completely the same as the contents of replication start VOL6B2.

  The first storage subsystem 100A and the second storage subsystem 100B perform the same processing as the above S204A and S204B, so that the replication direction in the VOL group including the SJNLVOL6A2 is reversed (S214A and S214B). Further, the first storage subsystem 100A (or the second storage subsystem 100B) notifies the third storage subsystem 100C whether the reversal of the VOL group including the SJNLVOL 6A2 has succeeded or failed (S214C).

  The third storage subsystem 100C executes a copy inversion process for inverting the replication direction in the VOL group including the SJNLVOL 6C1 (S214D). Specifically, for example, the third storage subsystem 100C generates a JNL copy inversion instruction including the mirror ID “2” of the mirror pair that the SJNLVOL 6C1 configures, and refers to the path management data 500C, and the SJNLVOL 6C1 The PJNLVOL constituting the mirror pair and the primary storage subsystem including the same are identified, and the generated JNL copy inversion instruction is transmitted to the identified primary storage subsystem (that is, the first storage subsystem) 100A. Further, for example, the third storage subsystem 100C includes the PJNLVOL-ID, PVOL-ID and primary storage subsystem ID related to the mirror ID “2”, the SJNLVOL-ID, SVOL-ID, and the secondary in the path management data 500C. Replace storage subsystem ID. Further, for example, the third storage subsystem 100C sets the JNL generation update number received by the copy inversion process of the first storage subsystem 100B in the path management data 500A in association with the inverted PJNLVOL6C1.

  The first storage subsystem 100A that has received the JNL copy inversion instruction from the third storage subsystem 100C executes a copy inversion process in response to the instruction (S214A). Specifically, for example, the first storage subsystem 100A refers to the path management data 500A and notifies the third storage subsystem 100C of a JNL generation update number (for example, update number “16”) related to the PJNLVOL6A3. Further, for example, the first storage subsystem 100A includes the PJNLVOL-ID, PVOL-ID and primary storage subsystem ID related to the mirror ID “2”, the SJNLVOL-ID, SVOL-ID, and the secondary in the path management data 500A. Replace storage subsystem ID. Also, the first storage subsystem 100A transmits a JNL read instruction for the inverted PJNLVOL6C1 to the third storage subsystem 100C, thereby reading the JNL in the PJNLVOL6C1, and storing the JNL in the inverted SJNLVOL6A3. Also good.

  Thereafter, when the third storage subsystem 100C receives the reversal failure from the first storage subsystem 100A, it notifies the third host terminal 180C of the reversal failure, while receiving the reversal success from the first storage subsystem 100A. In this case, an availability notification is transmitted to the third host terminal 180C (S215). In the latter case, the third storage subsystem 100C may notify the VOL-C ID of the VOL6C2 to the third host terminal 180C so that the third host terminal 180C can send a write command to the VOL6C2.

  When the third host terminal 180C receives the inversion failure from the third storage subsystem 100C, the third host terminal 180C does not generate write data or transmit a write command.

  On the other hand, when the success of inversion is transmitted to the third host terminal 180C, the duplication processing according to the new multi-hop method shown in FIG. 32 is performed. Specifically, for example, in the VOL group including the inverted PVOL6C2, the following replication process is performed.

  That is, the third host terminal 180C generates write data, and transmits the generated write data and the write command having the VOL-ID of the inverted PVOL6C2 to the third storage subsystem 100C. The third storage subsystem 100C stores the received write data in the PVOL6C2, and includes the write data and the JNL generation update number (JNL generation update number corresponding to PJNLVOL6C1) set in the path management data 500C. JNL is generated, and the JNL is stored in PJNLVOL6C1. The first storage subsystem 100A transmits a JNL read command for the PJNLVOL 6C1 to the third storage subsystem 100C, and stores the JNL read in response thereto in the SJNLVOL 6A3. Further, the first storage subsystem 100A stores the write data included in the JNL to the SJNLVOL 6A3 in the SPVOL 6A1, and corresponds to the write data and the JNL generation update number (PJNLVOL 6A2) set in the path management data 500A. JNL generation update number) is generated, and the JNL is stored in PJNLVOL6A2. Thereafter, the JNL is copied from the PJNLVOL 6A2 to the SJNLVOL 6B1, and the write data in the JNL is restored to the SVOL 6B2.

  As described above, in the first modified example of the second embodiment, when a failure occurs in the second host terminal 180B, the replication direction is automatically reversed between all the storage subsystems, so that a new multi A hop method is automatically constructed, and replication processing according to the new multi-hop method is performed. Thereby, even if a failure occurs in the second host terminal 180B, the entire data processing system 1 does not go down, so that highly reliable replication processing is maintained. If the first modification is read, even in the multi-hop method of FIG. 36 described later, the replication direction between all the storage subsystems (in particular, the replication direction of JNL, in other words, the attribute of JNLVOL) is reversed. A new multi-hop scheme can be constructed.

  (B) A second modification of the second embodiment.

  FIG. 34 shows an overview of replication processing after a failure occurs in the first storage subsystem 100A in the replication processing according to the multi-hop method shown in FIG. 30A, and FIG. 35 shows a failure in the first storage subsystem 100A. The flow of processing performed when it occurs is shown.

  When a failure occurs in the first storage subsystem 100A (S220), this is detected by the same method as in the second modification of the first embodiment. When it is detected that a failure has occurred in the first storage subsystem 100A, a VOL group reconstruction process is performed (S221). The process of S221 is the same as S112 (see FIG. 26) described above.

  The second storage subsystem 100B adds information related to the new mirror pair by the PJNLVOL 6B1 and the SJNLVOL 6C1 to the path management data 500B. In addition, the second storage subsystem 100B changes the mirror pair partner of SJNLVOL6C1 from the second PJNLVOL6A3 to PJNLVOL6B1, and the pair partner change instruction for changing the SVOL6C2 VOL pair partner from PVOL6A1 to PVOL6B2. Transmit to system 100C. In response to the pair partner change instruction from the second storage subsystem 100B, the third storage subsystem 100C changes the contents of the path management data 500C, the mirror pair partner of SJNLVOL6C1 is PJNLVOL6B1, and the VOL pair partner of SVOL6C2 The content is updated to indicate that it is PVOL6B2.

  Through the series of processes described above, a new VOL group is constructed in which the replication start VOL is set to VOL6B2, the relay JNLVOL is set to 6B1 and 6C1, and the replication goal VOL is set to VOL6C2. Thereafter, processing according to the new VOL group is executed. For example, when the second storage subsystem 100B generates a JNL to be stored in the PJNLVOL6B1 for the first time after a new VOL group is constructed, the JNL includes the next update number of the one or more JNLs in the PJNLVOL6B1. An update number (eg, “16”) is included. Further, when the second storage subsystem 100B first reads the JNL in the PJNLVOL6B1 and transmits it to the third storage subsystem 100C after constructing a new VOL group, the second storage subsystem 100B has the update number specified for the third storage subsystem 100C. The JNL or the JNL having the oldest update number among one or more JNLs in the PJNLVOL6B1 is read and transmitted to the third storage subsystem 100C. When the third storage subsystem 100C receives the JNL having the oldest update number, if the JNL including the update number exists in the SJNLVOL6C1 (or if the restore process based on the JNL has been completed) The received JNL is discarded. Further, when the third storage subsystem 100C receives the JNL, the update number in the JNL is a number next to the latest update number (eg, “7”) of one or more JNLs in the SJNLVOL6C1 (eg, “8”). ”), The JNL is stored in SJNLVOL6C1. In addition, when the third storage subsystem 100C receives the JNL having the oldest update number, the oldest update number is the next update number (for example, “7”) of one or more JNLs in the SJNLVOL6C1. If the number is larger than the number (for example, “9”), the replication process is stopped. This is because restoration processing in the order of consecutive update numbers cannot be performed.

  As described above, in the second modification of the second embodiment, when a failure occurs in the first storage subsystem 100A, a new VOL group from which the VOL provided in the first storage subsystem 100A is removed is reconstructed. The replication process is continued in accordance with the new VOL group.

  (C) A third modification of the second embodiment.

  FIG. 36 shows an overview of replication processing according to the multi-hop method according to the third modification of the second embodiment.

  In this third modification, JNLVOL6A2 becomes SPJNLVOL having both attributes of SJNLVOL and PJNLVOL, and SVOL6A1 and SJNLVOL6C1 are associated with SPJNLVOL6A2. In this case, the write data included in the JNL from the PJNLVOL 6B1 to the SPJNLVOL 6A2 is restored to the SVOL 6A1, and the JNL to the SPJNLVOL 6A2 is copied to the SJNLVOL 6C1.

  According to the third modification, the first storage subsystem 100A does not need to create a JNL, so the load on the first storage subsystem 100A can be reduced. In addition, according to the third modification, the JNLVOL 6A3 is not necessary, so that the storage capacity can be reduced.

  The above is the description of the modification of the second embodiment. If the description so far is read, it is possible to construct a multi-hop scheme when the number of storage subsystems 100 is four or more. For example, if the description about FIG. 30A is read, it is possible to construct | assemble the multihop system illustrated to FIG. 37A. If the explanation about FIG. 36 is read, it is possible to construct the multi-hop scheme exemplified in FIG. 37B. 32 and 33, in order to switch from the multi-hop method illustrated in FIG. 37A to another multi-hop method illustrated in FIG. 38A, for example, when a failure occurs in the second host terminal 180B. It is possible to construct a mechanism. 34 and FIG. 35, for example, when a failure occurs in the first storage subsystem 100A, the multi-hop method illustrated in FIG. 37A is changed to another multi-hop method illustrated in FIG. 38B. It is possible to build a mechanism for switching.

  Next, a third example of one embodiment of the present invention will be described. The third example relates to a method example for setting various information for realizing at least one of the above-described embodiment and the first to second examples. Hereinafter, a GUI (graphical user interface) screen used when the method example is adopted will be described. Note that the GUI screen described below is a GUI screen provided by software installed in the SVPs 281A to 281C or the management terminal 109. In the following description, a case where a VOL group from the first site 1 to the second site 2 is created is taken as an example. However, in the description, a VOL group is created between another site and another site. It can also be applied to cases.

  FIG. 39A is an example of a first GUI screen.

  This first GUI screen is a screen for specifying a paired VOL and confirming the status of the pair. When the menu “Paircreate” indicated by the reference number 5001 is selected on the first GUI screen, the following second GUI screen is displayed, and a VOL group can be created.

  FIG. 39B is an example of a second GUI screen.

  The second GUI screen is a screen for inputting information related to the VOL group. For example, information relating to the VOL pair partner when the write data VOL6A1 of the first site 840A is set to PVOL, for example, information relating to the write data VOL6B2 of the second site 140B, can be entered in the column indicated by the reference number 5002. In the column indicated by reference number 5003, information regarding JNLVOL (PJNLVOL) 6A2 associated with the write data VOL6A1 (PVOL) of the first site 840A can be input. Further, in a column indicated by reference number 5005, information related to JNLVOL (SJNLVOL) 6B1 that forms a mirror pair with JNLVOL 6A2 and is associated with write data VOL6B2 can be input. In the column indicated by reference number 5004, the mirror ID of the mirror pair can be entered. Further, in the column indicated by reference number 5006, information related to the second storage subsystem 100B including the pair partner VOLs 6B2 and 6B1 can be input. When these pieces of information are input and the “SET” button 5007 is pressed, for example, a copy execution instruction is input from the management terminal 109 to the first and second storage subsystems 100A and 100B via the SVPs 281A and 281B. For example, the initial copy process illustrated in FIG. 10 is executed. Further, the contents of the volume management data 400A and 400B and the path management data 500A and 500B are updated based on the input information.

  The preferred embodiments and some examples of the present invention have been described above, but the present invention is not limited to these embodiments and examples, and various modifications can be made without departing from the scope of the present invention. Needless to say.

  For example, if the number of storage subsystems is four or more, a data processing system employing both the multi-target method and the multi-hop method can be constructed.

  In addition, for example, in the second modification of the first embodiment or the second embodiment, the storage subsystem 100 transmits a JNL read command to a certain storage subsystem 100, so that the storage subsystem 100 A certain JNL is received from the system 100, and the received update number in the JNL is larger than the update number next to the latest update number in the SJNLVOL with which it is provided (or a number below the latest update number) ), The received JNL is discarded, and a JNL read command (which may include designation of a desired update number) is sent to another storage subsystem 100, so that the SNL An attempt may be made to receive a JNL including the update number next to the latest update number (that is, the desired update number). At that time, a certain storage subsystem 100 may transmit a JNL read command to another storage subsystem 100 existing further downstream.

  Further, for example, the control information 141 of each storage subsystem 100 provided in the data processing system 1 includes position data (for example, storage subsystem) indicating where the storage subsystem 100 itself and / or other storage subsystem 100 is located in the replication path Position data associated with the system ID) may be included. The storage subsystem 100 can identify which storage subsystem 100 is in which position of the replication path by referring to one or more pieces of position data. In this case, for example, the storage subsystem 100 receives from the storage subsystem 100 a JNL including the update number next to the latest update number (that is, a desired update number) in the SJNLVOL that the storage subsystem 100 has. In the case of failure, the other storage subsystem (for example, the storage subsystem existing at the most downstream side) 100 existing downstream from the certain storage subsystem 100 is referred to by referring to the control information 141 provided by itself. An attempt may be made to receive a JNL containing the desired update number from another identified storage subsystem 100 that has been identified. This is considered to be particularly effective in the multi-hop method, for example, when the JNL is full and the JNL is deleted from the old JNL. This is because in the multi-hop method, the storage subsystem 100 existing on the downstream side rather than the upstream side is more likely to hold a JNL including an older update number.

  Further, for example, when a failure occurs in the host terminal or the storage subsystem, when at least one of the replication path and the replication direction is reconstructed, the storage subsystem 100 creates a new replication source VOL associated with the VOL provided by itself. Alternatively, it is possible to search for a replication destination VOL, and update the content of the control information 141 included therein to the content in which the new replication source VOL or the replication destination VOL is associated with the VOL. In that case, a new copy source VOL or copy destination VOL can be determined by the following method. For example, in the case where information related to all replication paths and replication directions in the data processing system 1 (for example, a plurality of VOLIDs and storage subsystem IDs arranged in the replication direction) in the data processing system 1 is registered in the control information 141, the storage subsystem Can determine a new copy source VOL or copy destination VOL by referring to the information. Specifically, for example, when a failure occurs in the first storage subsystem 100A, the second storage subsystem 100B uses the second control information 141B in which information on all the above-described replication paths and replication directions is registered. It is possible to refer to and select JNLVOL6C1 having an attribute as a copy destination, and associate SJNLVOL6C1 with PJNLVOL6B1.

  Further, for example, the timing when the storage subsystem 100 transmits a JNL read command (or JNL write command) may be a case where the load (for example, CPU usage rate) of the storage subsystem 100 is below a certain load. .

  In addition, for example, the first site 140A can be taken over to the second site 140B. For example, the following conditions (A) and (B) may be adopted as conditions for executing the takeover. it can.

(A) External (configuration) conditions The infrastructure of the first site 140A and the second site 140B is normal. In addition, at least one host terminal is connected to the second site 140B that is the takeover destination. The host terminal may be any host terminal as long as it can issue a takeover instruction to the second site 140B (for example, the storage subsystem 100B).

(B) Internal (Processing) Conditions The update status in the PVOL at the first site 140A before the takeover is performed (for example, the update number corresponding to the write data last written in the PVOL) and the second site 140B A takeover is performed after the update status in the SVOL (for example, the update number corresponding to the write data restored last in the SVOL) becomes the same. This is to maintain the consistency (identity) of information. For example, if the old JNL has been discarded at the first site 140A and the update status in the PVOL at the first site 140A and the update status in the SVOL at the second site 140B cannot be made the same, The storage subsystem 100B at the two site 140B may abort the takeover, or may access another storage subsystem at another site and replace the old JNL discarded at the first site 140A with the other storage subsystem. It may be acquired from the system so that the update status is the same as that of the first site 140A.

FIG. 1A shows an overview of a first replication process performed by a data processing system according to an embodiment of the present invention. FIG. 1B shows an overview of a second replication process performed by the data processing system. FIG. 2 shows a configuration example of the update data 4. FIG. 3 is a configuration example of the write data VOL and JNLVOL, and particularly shows a configuration example represented by the update data 4 illustrated in FIG. FIG. 4 shows a configuration example of the data processing system 1 according to an embodiment of the present invention. FIG. 5 shows a configuration example of each of the control information 141A to 141C. FIG. 6A shows a configuration example of the VOL management data 400A when the multi-target method exemplified in FIG. 1A is adopted. FIG. 6B shows a configuration example of the VOL management data 400A when the multi-hop method exemplified in FIG. 1B is adopted. FIG. 7A shows a configuration example of the path management data 500A when the multi-target method exemplified in FIG. 1A is adopted. FIG. 7B shows a configuration example of the path management data 500A when the multi-hop method exemplified in FIG. 1B is adopted. FIG. 8 shows a configuration example of the pointer management data 700. FIG. 9 shows a JNLVOL configuration specified from the pointer management data 700 illustrated in FIG. FIG. 10 is a flowchart of the initial copy process. FIG. 11 shows an outline of the flow of the command reception process 210 performed by the first storage subsystem 100A. FIG. 12 shows a flowchart of the command reception process 210. FIG. 13 shows a flowchart of the JNL creation process performed by the first storage subsystem 100A. FIG. 14 is a diagram for explaining JNL read reception processing according to an embodiment of the present invention. FIG. 15 is a flowchart of JNL read reception processing according to an embodiment of the present invention. FIG. 16 is a diagram for explaining JNL read instruction processing according to an embodiment of the present invention. FIG. 17 is a flowchart of JNL read instruction processing according to an embodiment of the present invention. FIG. 18 is a flowchart of JNL storage processing according to an embodiment of the present invention. FIG. 19 is a diagram for explaining restore processing according to an embodiment of the present invention. FIG. 20 is a flowchart of restore processing according to an embodiment of the present invention. FIG. 21A shows an overview of a replication process that is normally performed in the data processing system according to the first example of one embodiment of the present invention. FIG. 21B shows an overview of replication processing after a failure has occurred in the first host terminal in the data processing system. FIG. 22 shows a flow of switching processing from the multi-target method to the multi-hop method performed when a failure occurs in the first host terminal. FIG. 22 shows a flow of switching processing from the multi-target method to the multi-hop method performed when a failure occurs in the first host terminal. FIG. 23 shows an overview of the replication process after a failure occurs in the first host terminal 180A in the first modification example when the replication process according to the multi-target method shown in FIG. 21A is performed. FIG. 24 shows the switching from the multi-target method to another multi-target method that is performed when a failure occurs in the first host terminal 180A during the replication process according to the multi-target method shown in FIG. 21A. The flow of processing is shown. FIG. 25 shows an overview of the replication process after a failure has occurred in the first storage subsystem 100A in the replication process according to the multi-target method shown in FIG. 21A. FIG. 26 shows a flow of processing performed when a failure occurs in the first storage subsystem 100A in the replication processing according to the multi-target method shown in FIG. 21A. FIG. 27 shows an overview of a replication process according to the multi-target method according to the third modification of the first embodiment of the present invention. FIG. 28A shows an example of a case where JNL replication from the first storage subsystem 100A to the second storage subsystem 100B can no longer be performed in the replication processing according to the multi-target method shown in FIG. 21A. FIG. 28B shows a specific example of the cause of this case. FIG. 29 shows an overview of a replication process according to the multi-target method according to the fifth modification of the first embodiment of the present invention. FIG. 30A shows an overview of replication processing according to the multi-hop method, which is normally performed in the data processing system according to the second example of one embodiment of the present invention. FIG. 30B shows an overview of replication processing after a failure has occurred in the second host terminal in the data processing system. FIG. 31 shows the flow of the switching process from the multi-hop method to the multi-target method that is performed when a failure occurs in the second host terminal during the replication process of FIG. 30A. FIG. 32 shows an overview of the replication process after a failure occurs in the second host terminal 180B in the first modification example when the replication process according to the multi-hop method shown in FIG. 30A is performed. FIG. 33 shows the flow of the switching process from the multi-hop method to another multi-hop method that is performed when a failure occurs in the second host terminal 180B during the duplication process of FIG. 30A. FIG. 34 shows an overview of the replication process after a failure occurs in the first storage subsystem 100A in the replication process according to the multi-hop method shown in FIG. 30A. FIG. 35 shows the flow of processing performed when a failure occurs in the first storage subsystem 100A in the replication processing according to the multi-hop method shown in FIG. 30A. FIG. 36 shows an overview of the duplication processing according to the multi-hop method according to the third modification of the second embodiment of the present invention. FIG. 37A shows an overview of a replication process according to the multi-hop method according to the fourth modification of the second embodiment of the present invention. FIG. 37B shows an overview of the duplication processing according to the multi-hop method according to the fifth modification of the second embodiment of the present invention. FIG. 38A shows an overview of replication processing according to the multi-hop method according to the sixth modification of the second embodiment of the present invention. FIG. 38B shows an overview of the duplication processing according to the multi-hop method according to the seventh modification of the second embodiment of the present invention. FIG. 39A shows a configuration example of a first GUI screen according to the third embodiment of the present invention. FIG. 39B shows a configuration example of a second GUI screen according to the third embodiment of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Data processing system 12 ... Mirror pair 14 ... VOL pair 16 ... VOL group 100A, 100B, 100C ... Storage subsystem 101A, 101C, 101C ... Controller 110A, 110B, 110C ... Channel adapter (CHA) 120A, 120B, 120C ... Disk adapter (DKA) 141A, 141B, 141C ... Control information 6A, 6B, 6C ... Logical volume

Claims (10)

A first storage subsystem that receives the write data from a first host terminal that transmits write data that is data to be written, and stores the received write data;
A second storage subsystem connected to the first storage subsystem and connected to a second host terminal for transmitting write data;
A third storage subsystem connected to the first storage subsystem;
The first storage subsystem is
A first storage device to which an attribute as a first replication source is assigned;
One or more second storage devices that are associated with the first storage device and assigned an attribute as a second replication source,
Storing the write data received from the first host terminal in the first storage device;
When the write data is stored in the first storage device, an update number indicating the order in which the first storage device is updated is issued, and a data set including the issued update number and the write data is generated. And storing the generated data set in the one or more second storage devices,
Reading a data set from the one or more second storage devices, sending the read data set to the second storage subsystem;
Reading a data set from the one or more second storage devices, sending the read data set to the third storage subsystem;
The second storage subsystem is
A third storage device that is associated with at least one of the one or more second storage devices and is assigned an attribute as a second replication destination to be a pair of the second replication source;
A fourth storage device associated with the third storage device and assigned with an attribute as a first replication destination to be a pair of the first replication source,
Receiving the data set from the first storage subsystem, storing the received data set in the third storage device;
An update number included in each of the one or more data sets in the third storage device, and an update included in a data set managed last by the second storage subsystem and stored in the fourth storage device Based on the number, the data set to be read is selected from within the third storage device, the selected data set is read from the three storage devices in the order of update numbers, and the write data in the read data set is Stored in the fourth storage device,
The third storage subsystem is
A fifth storage device associated with at least one of the one or more second storage devices and assigned with an attribute as the second replication destination;
A sixth storage device associated with the fifth storage device and assigned with the attribute as the first replication destination,
Receiving the data set from the first storage subsystem, storing the received data set in the fifth storage device;
An update number included in one or more of the data sets in the fifth storage device, and an update included in the data set last managed in the sixth storage device, managed by the third storage subsystem A data set to be read is selected from the fifth storage device based on the number, the selected data set is read from the fifth storage device, and the write data in the read data set is read from the sixth storage device. Store it in a storage device,
When a failure occurs in the first host terminal,
The first storage subsystem is managed by the second storage subsystem and the update number included in each of the one or more data sets in the second storage device, and is finally stored in the third storage device. Read out an unsent data set to at least the second storage subsystem out of the one or more data sets in the one or more second storage devices based on the update number included in the set data set; Transmitting the read untransmitted data set to the second storage subsystem;
The second storage subsystem is
Receiving the untransmitted data set from the first storage subsystem; storing the received data set in the third storage device;
One or more data sets that have not yet been read in the third storage device are read in the order of update numbers, and the write data in the read data sets are stored in the four storage devices,
The data processing system replaces the first replication source that is the attribute of the first storage device and the first replication destination that is the attribute of the fourth storage device, and the one or more second storage devices Replacing the second replication source that is the attribute of the second and the second replication destination that is the attribute of the third storage device,
The second storage subsystem is
Receiving write data from the second host terminal, storing the received write data in the fourth storage device;
When the write data is stored in the fourth storage device, an update number indicating the order in which the fourth storage device is updated is issued, and a data set including the issued update number and the write data is generated. And storing the generated data set in the third storage device,
Reading a data set from the third storage device, sending the read data set to a first storage subsystem;
The first storage subsystem is
Receiving the data set from the second storage subsystem, storing the received data set in the one or more second storage devices;
Based on update numbers included in one or more of the data sets in the one or more second storage devices, a data set to be read is selected from the one or more second storage devices, and the selected Reading a data set from the one or more second storage devices, and storing write data in the read data set in the first storage device;
Data processing system. The data processing system of claim 1, wherein
The third storage subsystem is connected to the second storage subsystem;
The first storage subsystem or the second storage subsystem reads a data set that has not been transmitted to at least the third storage subsystem among the one or more data sets in the one or more second storage devices, Sending the read unsent data set to the third storage subsystem;
The third storage subsystem is
Receiving the untransmitted data set from the first storage subsystem, storing the received data set in the fifth storage device;
One or more data sets that have not yet been read in the fifth storage device are read in the order of update numbers, the write data in the read data set is stored in the sixth storage device,
In the data processing system, the fifth storage in which the second replication destination is allocated to the third storage device to which the second replication source is allocated by exchanging the second replication destination and the second replication source. Map devices,
The second storage subsystem is
Reading a data set from the third storage device, sending the read data set to a third storage subsystem;
The third storage subsystem is
Receiving the data set from the second storage subsystem and storing the received data set in the fifth storage device;
Data processing system. The data processing system of claim 1, wherein
The third storage subsystem is connected to the second storage subsystem;
The second storage subsystem is
Receiving write data from the second host terminal and storing the generated data set in the third storage device;
The third storage subsystem is
Based on an update number included in each of the one or more data sets in the fifth storage device, and an update number included in each of the one or more data sets in the fourth storage device, Receiving a data set not yet transmitted to the third storage subsystem from the second storage subsystem, storing the received data set in the fifth storage device;
Based on an update number included in each of the one or more data sets in the fifth storage device, a data set to be read is selected from within the fifth storage device, and the selected data set is selected from the fifth storage device. Read from the storage device, and store the write data in the read data set in the sixth storage device,
Data processing system. The data processing system of claim 1, wherein
The first storage subsystem and the third storage subsystem are independent of the first storage subsystem even if communication between the first storage subsystem and the second storage subsystem is interrupted. Maintaining communication with the third storage subsystem;
Data processing system. The data processing system of claim 1, wherein
At least one of the second storage subsystem and the third storage subsystem sends a read command to the first storage subsystem;
The first storage subsystem receives the read command and, in response to the read command, transmits the read data set to at least one of the second storage subsystem and the third storage subsystem;
Data processing system. The data processing system of claim 1, wherein
The one or more second storage devices include a first second storage device associated with the third storage device and a second second storage device associated with the fifth storage device. And
The first storage subsystem stores the generated data set in both the first and second second storage devices;
Data processing system. The data processing system of claim 1, wherein
If at least one of the first to third storage subsystems has no free storage area in the storage device to which the second replication source or the second replication destination is assigned, the storage device Delete one of the stored datasets with the oldest update number,
Data processing system. The data processing system of claim 1, wherein
The second storage subsystem is connected to the third storage subsystem and a failure occurs in the first storage subsystem;
At least one of the second storage subsystem and the third storage subsystem is the third storage device to which the attribute as the second replication source is assigned, and the attribute to which the attribute as the second replication destination is assigned. Associate with the fifth storage device,
The second storage subsystem sends the data set read from the third storage device to the third storage subsystem;
The third storage subsystem receives a data set from the second storage subsystem and stores the received data set in the fifth storage device;
Data processing system. The storage subsystem of claim 1, wherein
The one or more second storage devices are one.
Storage subsystem. A first storage subsystem that receives the write data from a first host terminal that transmits write data that is data to be written and stores the received write data; and a second storage connected to the first storage subsystem In a data processing method realized by a data processing system comprising a storage subsystem and a third storage subsystem connected to the first storage subsystem,
The first storage subsystem comprises a first storage device and one or more second storage devices;
The second storage subsystem comprises a third storage device and a fourth storage device;
The third storage subsystem comprises a fifth storage device and a sixth storage device;
The first storage subsystem is
Storing the write data transmitted from the first host terminal in a first storage device;
Issuing an update number representing an order in which the first storage device is updated by storing the write data in the first storage device;
Generating a data set including the issued update number and the write data, and storing the generated data set in one or more second storage devices;
Reading a data set from the one or more second storage devices;
The second storage subsystem is
Storing the read data set in the third storage device;
An update number included in each of the one or more data sets in the third storage device, and an update included in a data set managed last by the second storage subsystem and stored in the fourth storage device And reading the data set from the third storage device based on the number, and storing the write data in the read data set in the fourth storage device;
The third storage subsystem is
Storing the read data set in the fifth storage device;
An update number included in one or more of the data sets in the fifth storage device, and an update included in the data set last managed in the sixth storage device, managed by the third storage subsystem And reading the data set from the fifth storage device based on the number, and storing the write data in the read data set in the sixth storage device;
When a failure occurs in the first host terminal,
The first storage subsystem is managed by the second storage subsystem and the update number included in each of the one or more data sets in the second storage device, and is finally stored in the third storage device. Read out an unsent data set to at least the second storage subsystem out of the one or more data sets in the one or more second storage devices based on the update number included in the set data set; Transmitting the read untransmitted data set to the second storage subsystem;
The second storage subsystem is
Receiving the untransmitted data set from the first storage subsystem and storing the received data set in the third storage device;
Reading one or more data sets that have not yet been read in the third storage device in the order of update numbers, and storing the write data in the read data sets in the four storage devices;
The data processing system replaces the first replication source that is the attribute of the first storage device and the first replication destination that is the attribute of the fourth storage device, and the one or more second storage devices Replacing the second replication source that is the attribute of the second replication destination and the second replication destination that is the attribute of the third storage device;
The second storage subsystem is
Receiving write data from the second host terminal and storing the received write data in the fourth storage device;
When the write data is stored in the fourth storage device, an update number indicating the order in which the fourth storage device is updated is issued, and a data set including the issued update number and the write data is generated. Storing the generated data set in the third storage device;
Reading a data set from the third storage device and transmitting the read data set to a first storage subsystem;
The first storage subsystem is
Receiving the data set from the second storage subsystem and storing the received data set in the one or more second storage devices;
Based on update numbers included in one or more of the data sets in the one or more second storage devices, a data set to be read is selected from the one or more second storage devices, and the selected A data processing method comprising: reading a data set from the one or more second storage devices; and storing write data in the read data set in the first storage device.