JP2013210919A - Replication device, replication method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a replication device, a replication method, and a program, capable of executing roll forward without waiting until transfer of a journal is completed.SOLUTION: A replication device 110 includes: replication information reception means 118 for receiving a journal block; a journal volume 120 including a storage area for storing a fixed number of journal blocks; and roll forward means 111 for updating a duplicate volume by reading out a journal block from the storage area. A storage area has a circulation structure in which when the number of the journal blocks exceeds a predetermined number, stored journal blocks are overwritten, in chronological order from the oldest one, by a new journal block, and the journal block includes a circulation number indicative of the number of rounds of the overwriting. The roll forward means 111 discriminates, when a journal block is read out, a latest journal block on the basis of the number of circulation to stop the roll forward.

Description

  The present invention relates to a replication apparatus, a replication method, and a program. For example, the present invention relates to a replication technique capable of performing roll forward without waiting for completion of journal transfer.

  As one of measures against BCP (Business Continuity Plan), a method is known in which a remote site having the same data as the main site is prepared in advance, and the business is continued using the remote site when a disaster occurs. ing. As one of the methods for reflecting data changes occurring at the main site to the remote site, a technique called replication is often used (see Patent Document 1 as an example).

  Replication refers to the transfer of MV update data from the master volume at the main site (hereinafter referred to as MV) to the replication volume at the remote site (hereinafter referred to as RV), and the remote site transfers the replication volume based on the update data. It is a technology to update. One replication method is sequential replication in which data is written in RV while maintaining the data write order in MV.

  A feature of sequential replication is that RPO (Recovery Point Objective) is short and operations at a remote site can be resumed early.

  As one of the methods used with sequential replication, the storage device at the main site transfers only the MV update log (hereinafter referred to as journal) to the remote site, and the storage device at the remote site is based on the update log. There is a method for performing roll forward processing and updating RV.

  This method has an advantage that the transfer amount can be reduced by narrowing the transfer target to a journal.

Japanese Patent No. 4731975

  On the other hand, in the above-described method, journal transfer is performed between storage apparatuses. For this reason, the host device in charge of database (hereinafter referred to as DB) update processing at the main site and the remote site cannot grasp how far the journal transfer has been completed while the journal transfer is in progress. For this reason, when writing to the storage area of the received journal and reading of the journal for roll forward collide on the remote site side, the journal in the middle of writing is read on the roll forward execution side. Incorrect reflection sometimes occurred.

  Therefore, when performing roll forward, it is necessary to wait for the completion of journal transfer, which causes an increase in RTO (Recovery Time Objective).

  The present invention has been made to solve such problems, and a replication apparatus, a replication method, and a program capable of performing roll forward at high speed and maintaining consistency without waiting for completion of journal transfer The purpose is to provide.

  The replication apparatus according to the present invention includes a replication information receiving means for receiving a journal block including update information of a replication source volume, a journal volume including a storage area for storing a fixed number of the journal blocks in the order received, and the storage area Roll journal means for sequentially reading the journal blocks from each other and executing roll forward for updating the replication volume based on the read journal blocks, and the storage area has the fixed number of the journal blocks stored therein. If it exceeds, the stored journal block is a circular structure overwritten by the new journal block in order from the oldest one, and each journal block indicates the number of times the overwriting has made one round in the circular structure. The roll forward means, when reading the stored journal block, determines the latest journal block based on the circulation count, and uses the latest journal block to roll the roll forward. Is to stop.

  The replication method according to the present invention includes a replication information reception step of receiving a journal block including update information of a replication source volume, a storage step of storing a fixed number of the journal blocks in the storage area of the journal volume in the order received, A roll-forward step of sequentially reading the journal blocks from the storage area and executing a roll-forward for updating the replication volume based on the read journal blocks, wherein the stored journal blocks are the fixed If the number exceeds the number, the stored journal blocks are overwritten by new journal blocks in order from the oldest, and each of the journal blocks is overwritten in the circular structure. In the roll forward step, when reading the stored journal block, the latest journal block is determined based on the circulation number, and the latest The roll forward is stopped with a journal block.

  A program according to the present invention includes a replication information receiving step for receiving a journal block including update information of a replication source volume in a computer, and a storage step for storing a fixed number of the journal blocks in the storage area of the journal volume in the order received. And a roll forward step of executing a roll forward step of executing a roll forward for updating a replication volume based on the read journal block, and sequentially reading the journal blocks from the storage area. When the number of journal blocks that have been exceeded exceeds the certain number, the stored journal blocks are overwritten by new journal blocks in order from the oldest one, and each journal block is , Including a circulation number indicating the number of times the overwriting has been completed in the circulation structure. In the roll forward step, when the stored journal block is read, the latest journal block is based on the circulation number. And the roll forward is stopped with the latest journal block.

  According to the present invention, it is possible to provide a replication apparatus, a replication method, and a program capable of performing roll forward at a high speed and maintaining consistency without waiting for completion of journal transfer.

It is a figure which shows the structure of embodiment of this invention. It is a figure which shows the structure of embodiment of this invention. It is a figure which shows the structure of embodiment of this invention. It is a figure which shows the structure of embodiment of this invention. It is a figure which shows the structure of embodiment of this invention. It is a figure which shows the structure of embodiment of this invention. It is a figure which shows the process of embodiment of this invention. It is a figure which shows the process of embodiment of this invention. It is a figure which shows the process of embodiment of this invention. It is a figure which shows the process of embodiment of this invention. It is a figure which shows the process of embodiment of this invention. It is a figure which shows the process of embodiment of this invention.

  Hereinafter, specific embodiments to which the present invention is applied will be described in detail with reference to the drawings.

  First, the configuration of the replication apparatus 110 according to the embodiment of the present invention will be described with reference to FIG.

  The replication device 110 is a device that configures a remote site that reflects MV updates at a main site (not shown). The replication apparatus 110 is typically configured by a server computer or the like, and includes hardware such as a central processing unit (CPU), a volatile memory, a nonvolatile memory, a communication device, and an input / output device. The CPU executes various processes described later based on a control program (software) stored in a volatile memory or a nonvolatile memory. The replication apparatus 110 does not have to be a physically single apparatus, and may have a configuration in which functions described below are realized in a distributed manner by a plurality of apparatuses.

  The replication apparatus 110 typically includes a host and a storage apparatus. The storage device is a device that mainly manages journals, and includes a replication information receiving unit 118 and a journal volume 120. The host is a device that mainly executes RV update processing, and includes roll-forward means 111.

  The replication information receiving unit 118 is communicably connected to a replication source volume, that is, a main site having an MV, and sequentially receives a journal that is an MV update log. Here, the journal is transmitted / received in the form of a journal block including a plurality of update logs (journal records).

  The journal volume 120 has a plurality of storage areas for storing journal blocks received by the replication information receiving means 118. Journal blocks are stored in these storage areas sequentially from the received one. If all the storage areas are already used at the time of receiving the journal block, the received journal block is stored in the storage area where the oldest, that is, the journal block received at the earliest time is stored. Shall be. This is hereinafter referred to as circular writing.

  Here, in order to manage the cumulative number of times this circular writing has been performed in a certain storage area, in other words, the generation, in this embodiment, in order to record the number of circulations indicating the cumulative number of this circular writing in the journal block. A sector is established.

  The roll forward unit 111 sequentially reads journal blocks from the journal volume 120 and executes a process of updating the replication volume, that is, RV based on the read journal block, that is, roll forward.

  Here, the roll forward unit 111 determines which journal block is the latest journal block in the journal volume 120 based on the number of circulations recorded in the journal block when sequentially reading the journal blocks from the journal volume 120. To do. If the latest journal block can be determined, the roll forward unit 111 reflects the journal block in the RV and then stops the roll forward.

  Note that the host, storage device, various means, and the like are not necessarily physically independent devices, and are logically realized by the cooperation of the hardware and software included in the replication device 110 described above. It can be a thing.

  A more specific implementation example of the replication apparatus 110 will be described with reference to FIGS.

  The replication apparatus 110 as a remote site is connected to the main site 100 via communication means 117 and 105.

  The main site 100 includes a DB (MV) 109 as a replication source volume, a host, and a storage device. The host includes DB update means 101 that performs update processing on the DB (MV) 109. The storage apparatus includes an update log of the DB (MV) 109, that is, a journal volume (MV) 108 that stores a journal, and a replication information transmission unit 104 that transmits the journal to the replication apparatus 110 as a remote site.

  The replication device 110 as a remote site includes a DB (RV) 121 as a replication volume, a host, and a storage device. The host includes roll forward means 111 that performs roll forward processing of the DB (RV) 121 based on a journal received from the main site 100. The storage apparatus includes a replication information receiving unit 119 that receives a journal from the main site 100, a journal volume (RV) 120 that stores a journal, and an exclusive unit 114 that performs exclusive control of access to the journal volume (RV) 120.

  Hereinafter, these components and the data structure used in this embodiment will be described in detail.

<Journal block>
The journal block is a data structure that is used when the DB update unit 101 writes a log describing the update contents of the DB (MV) 109, that is, a journal record, to the journal volume (MV) 108.

  First, the output mechanism of the journal block will be described with reference to FIG.

  FIG. 3 shows a journal volume 201 in which the journal volume (MV) 108 and the journal volume (RV) 120 are schematically shown. The journal blocks 203 to 207 are output to an address area 202 (hereinafter referred to as journal area 202) reserved in advance in the journal volume 201. This structure is common to the journal volume (MV) 108 and the journal volume (RV) 120.

  In order to cope with a situation in which a large number of journal blocks are output and storage is required, the journal area 202 is cyclically used. That is, when the journal blocks 203, 204,... 207 are sequentially stored in the journal area 202 and there is no longer enough area to store the next journal block, the next journal block is the largest of these journal blocks. The old, that is, the earliest stored journal block 203 is overwritten.

  In this embodiment, the journal block has a fixed length. That is, the journal blocks 203 to 207 are all the same size. Thereby, even when the journal area 202 is used cyclically, the storage position of the journal block is uniquely determined. In other words, the storage position of the journal block in the journal area 202 does not change every time it circulates. As a result, the main site 100 and the replication apparatus 110 can easily and quickly specify the address at which the journal block is to be written and read.

  Next, the internal structure of the journal block will be described with reference to FIG.

  The journal block has one numerical value indicating how many times the output of the journal block has circulated in the journal area (the number of circulations), that is, a circulation serial number, and one or more journal records in which the update contents of the DB (MV) 109 are recorded. including. In the example of FIG. 4, the journal block includes journal records 302 to 305 following the first circulation serial number 301. Here, the journal records 302 to 305 may have a variable length. However, it is assumed that the journal record is stored in one journal block, that is, is not output across a plurality of journal blocks. Therefore, an unused area 306 may be added at the end of the journal block.

  The circulation sequence number 301 is added when a predetermined number of journal blocks are output, reach the end of the journal area (hereinafter referred to as the end of the area), and the output address of the new journal block returns to the beginning of the journal area. The circulation serial number 301 is used to specify the latest, that is, the most recently output journal block (hereinafter referred to as the journal end) when the journal area is cyclically used in this way.

  The journal record includes a journal containing the DB update contents (hereinafter referred to as an update journal record), a journal record indicating the start of a transaction (hereinafter referred to as a start journal record), and a journal record indicating the end of a transaction (hereinafter referred to as a journal record). Called end journal record). One transaction is represented by a set of a start journal record, zero or more update journal records, and an end journal record. The absence of an end journal record corresponding to the start journal record means that the transaction related to the start journal record has not been completed. Therefore, the journal record related to this transaction should not be reflected in the DB at the time of roll forward. Note that the start journal record and the end journal record may be included in different journal blocks.

  By giving such a structure to the journal block, the end of the journal can be easily identified on the remote site 110 side, and roll forward can be controlled appropriately. That is, according to the present embodiment, it is possible to easily keep the DB (RV) 121 on the remote site 110 side in the latest state by transferring only the journal block.

<Storage controllers 103 and 113>
The storage controllers 103 and 113 are means that the storage device 102 or 112 of the main site 100 or the remote site 110 has. The storage controller 103 or 113 receives a write request from the DB update unit 101 or a read request from the rollforward unit 111 and controls access to the journal volume (MV) 108 or journal volume (RV) 120 included in the storage apparatus 103 or 113. It has the function to do.

  The storage controller 103 of the main site 100 includes replication information transmission means 104 having a replication state management function and a journal transmission function. The storage controller 113 of the remote site 110 includes replication information receiving means 118 having a replication status and replication speed management function and a journal reception function. The storage controller 113 of the remote site 110 further includes an exclusive unit 114 that performs exclusive control of access to the journal volume (RV) 120.

  With reference to FIG. 6, the access control to the journal volume (MV) 108 or the journal volume (RV) 120 performed by the storage controllers 103 and 113 based on a write request by the DB update unit 101 or a read request by the roll forward unit 111. Explain the mechanism.

  The write request by the DB update unit 101 is composed of a write destination volume 501, a write start point address 502, a write end point address 503, and a write image 504. The write image 504 includes one or more journal blocks. The read request by the roll forward unit 111 is composed of a read destination volume 508, a read start point address 509, and a read end point address 510. The address here is information for uniquely specifying the position on the volume. For example, a number indicating how many journal blocks the target journal block is when viewed from the head journal block can be used as a relative address in this storage area.

<Replication information transmission means 104>
The replication information transmission unit 104 is a unit included in the storage controller 103 of the main site 100. The replication information transmission unit 104 has a function of transferring a journal block write request to the journal volume (MV) 108 received from the host by the storage controller 103 to the remote site 110.

<Replication information receiving means 118>
The replication information receiving unit 118 is a unit that the storage controller 113 of the remote site 110 has. The replication information receiving unit 118 has a function of receiving a write request transferred from the replication information transmitting unit 104 of the main site 100 and writing a journal block to the journal volume (RV) 120 using the function of the storage controller 113.

  The replication information receiving unit 118 obtains write exclusion using the exclusion unit 114 described later so that the later-described roll forward unit 111 does not read the journal block being written to the journal volume (RV) 120. The exclusive unit is a journal block which is a write unit by the host side.

  In addition, the replication information receiving unit 118 and the above-described replication information transmitting unit 104 hold the current replication state as replication state information 119 in order to prevent unauthorized update of the DB by replication whose order is not guaranteed. The replication status information 119 includes a state in which sequential replication is being performed (hereinafter referred to as a SYNC state), a state in which replication is being performed but the write order on the main host is not known and sequential replication cannot be performed (hereinafter, It is assumed that data that can distinguish two states (referred to as REPLICATE states) is included.

  The replication information receiving means 118 refers to this replication status information 119, and when it is in the REPLICATE status, the exclusion means 114 described later is used to prevent unauthorized update of the DB by rolling forward journal blocks whose writing order is unknown. Thus, the entire journal area of the journal volume (RV) 120 is excluded, and the progress of roll forward is suppressed.

  Further, the replication information receiving means 118 holds information indicating the replication speed with the main site 100, that is, information indicating the journal transfer speed, as replication speed information 126. The replication information receiving unit 118 uses the replication speed information 126 as information for determining the range of exclusive control requested to the exclusive unit 116. The replication information receiving unit 118 and the roll forward unit 111 exclude as many journal blocks as possible in accordance with the replication rate information 126 at a time. Thereby, it is possible to reduce the frequency of use of the exclusion unit 114. Specifically, when the current replication speed is higher than the past replication speed, since many journal blocks can be read and written at a time, the number of journal blocks to be exclusive at a time is increased. On the other hand, when the current replication rate is lower than the past replication rate, the number of journal blocks to be excluded at a time is reduced in order to prevent the roll forward unit 111 from proceeding with roll forward due to exclusion waiting.

<Exclusion means 114>
The exclusion unit 114 is a unit included in the storage controller 114 of the remote site 110. The exclusion unit 114 prevents the collision between the writing of the journal block to the journal volume (RV) 120 by the replication information receiving unit 118 and the reading of the journal block from the journal volume (RV) 120 by the roll forward unit 111. It has a function to perform exclusive control in journal block units.

  When the exclusion unit 114 receives an exclusion request from the replication information reception unit 118, it simultaneously receives the write volume, the write start point address, and the write end point address as information indicating the exclusion range, and registers these information in the exclusion information 116.

  Further, when the exclusion unit 114 receives the exclusion request from the roll forward unit 111, it simultaneously receives the reading volume, the reading start point address and the reading end point address as information indicating the exclusion range, and these pieces of information are registered in the exclusion information 116. The roll-forward means 111 confirms whether there is an overlap with the write volume, the write start point address and the write end point address, that is, whether there is a conflict. Notice.

<Communication means 105 and 117>
The communication means 105 and 117 are means that the storage device 102 or 112 of the main site 100 or the remote site 110 has. The communication units 105 and 117 connect the storage apparatuses 102 and 112 and transmit / receive various data related to replication.

<DB update unit 101>
This is a means used by the host of the main site 100 to update the DB (MV) 109. When updating the DB, the DB update unit 101 stores a journal record describing the update contents in a journal block and outputs it to the journal volume (MV) 108. The DB update unit 101 has a function of circulating and outputting journal blocks within the journal area in accordance with the data structure of the journal block described above.

  The operation in which the DB update unit 101 circulates and outputs journal blocks will be described with reference to FIG.

  FIG. 5 shows a journal area having an address area for five journal blocks as an example.

  It is assumed that the circulation sequence numbers of the journal blocks 402 to 406 are “2”, “2”, “2”, “1”, and “1” in order at time t. In this case, the journal block 404 is the journal block output last at the time t, and the journal end is between the journal blocks 404 and 405.

  The DB update means outputs the latest journal block at the position next to the end of the journal, that is, the position of the journal block 405 at time t.

  Time (t + 1) shows a state in which the DB update unit 101 outputs a new journal block to the journal blocks 404 and 405 and outputs a new journal block after time t. However, since the journal end has reached the end of the area at this time, the DB update unit 101 adds the circulation serial number of the new journal block from “2” to “3” in accordance with the above-described circulation use rule, The output position of the correct journal block is determined as the journal block 402 at the head of the journal area.

  At time (t + 2), a new journal block is written in the journal block 402 located at the head of the journal area.

  Here, it is assumed that the DB update unit 101 holds the last output position of the journal block as the journal block final write position 122.

<Roll forward means 111>
The roll-forward means 111 is a means possessed by the remote site 110 host. The roll-forward means 111 reads the journal block from the journal volume (RV) 120 and reflects it in the DB (RV) 121, thereby keeping the DB (RV) 121 in the latest state replicated with the DB (MV) 109. The roll-forward means 111 has a function of circularly inputting journal blocks from the journal area corresponding to the journal block data structure described above.

  With reference to FIG. 5, an operation in which the roll-forward unit 111 circularly inputs journal blocks will be described.

  At time t, the roll forward unit 111 reflects to the DB (RV) 121 up to the journal block 404 that is the end of the journal. On the other hand, the circulation sequence number “1” of the next journal block 405 is smaller than the circulation sequence number “2” of the journal block 404. At this time, the roll forward unit 111 determines that the journal end has been exceeded and stops reflecting the journal block in the DB (RV) 121.

  When the time advances to (t + 1), a new journal block is written in the journal blocks 405 and 406, and the end of the journal is moved, the roll forward unit 111 reads the journal block and reflects it in the DB (RV) 121. Resume.

  At time (t + 2), when the roll-forward unit 111 reaches the end of the area as a result of reading the journal block, it adds the circulation serial number to “3” and the journal at the head of the journal area, as with the DB update unit 101. Reading of the journal block is started from block 402. The roll forward means 111 performs subsequent detection of the journal end based on the added circulation serial number.

  Here, it is assumed that the roll forward unit 111 holds the journal block last read position 123, the unreflected journal record position information 124, and the roll forward target circulation serial number 125 as information for managing the progress of the roll forward.

  The journal block final read position 123 is information indicating the final position of the journal block read during the previous operation of the roll forward unit 111. The unreflected journal record position information 124 is information indicating the position of the journal block having the oldest unreflected journal record. The roll forward target circulation sequence number 125 is information indicating the circulation sequence number of the journal block including the oldest unreflected journal record.

  In the journal area 401, in addition to reading by the roll forward unit 111, writing by the replication information receiving unit 118 is performed. Therefore, there is a possibility that both reading and writing collide. In order to prevent this read / write collision, the roll-forward means 111 uses the exclusion means 114 before performing the read and performs exclusive control of the journal block to be read. If the journal block to be read is exclusive by the replication information receiving means 118, the roll-forward means 111 waits for the journal block to be read until the replication information receiving means 118 releases the exclusion.

  The journal block that has not yet been read by the roll forward unit 111 is overwritten by the new journal block by the replication information receiving unit 118, and the roll forward unit 111 is not in a state where the roll forward cannot be normally performed. In the embodiment, it is assumed that the roll forward unit 111 operates sufficiently faster than the replication information receiving unit 118 and the roll forward unit 111 is always in a state of “following up” the replication information receiving unit 118. This assumption is sufficiently realistic when comparing the processing contents of the replication means 118 that requires exchange between sites and the roll forward means 111 that completes all the processes in the remote site 110.

<DB (MV) 109>
This is a DB volume used in the main site 100. In this embodiment, since the DB (MV) 109 itself is not a replication target, it can be placed in any storage device (not shown).

<Journal volume (MV) 108>
This is a volume in which the journal of the DB (MV) 109 is recorded, and needs to be arranged in the storage apparatus 102 having a replication function. As described in the section <Journal block>, an area for outputting a journal is secured in advance in the volume. Writing to the journal volume (MV) 108 is transferred and reflected to the journal volume (RV) 120 by a series of replication processes.

<DB (RV) 121>
This is a DB volume used at the remote site 110. For example, when a business operation is performed using the remote site 110 when a disaster or the like occurs, the DB (RV) 121 is used instead of the DB (MV) 109. Like the DB (MV) 109, the DB (RV) 121 is not a replication target, and can be placed in an arbitrary storage device (not shown). The DB (RV) 121 is kept in the latest state by using the journal stored in the journal volume (RV) 120 by the roll forward unit 111.

<Journal volume (RV) 120>
This is a volume in which a journal for keeping the DB (RV) 121 in the latest state is stored. The journal volume (RV) 120 needs to be arranged in the storage apparatus 112 that can perform replication processing with the storage apparatus 102 including the journal volume (MV) 108.

  Next, the operation of this embodiment will be described with reference to FIGS. 2 and 7 to 12.

  First, the operation of outputting a journal block to the journal volume (MV) 108 when the DB update unit 101 updates the DB (MV) 109 will be described with reference to FIGS.

  First, the DB update unit 101 acquires the last write position of the previous journal block in order to determine the output position of the journal block (step 601). When the DB update unit 101 is immediately after startup, this is determined by the method for specifying the journal end point described in the section <Journal block>. On the other hand, if the DB update unit 101 is already operating and outputting a journal block, this is acquired from the journal block last write position 122.

  When the output position of the previous journal block acquired in step 601 has reached the end of the area, the DB update unit 101 returns the output position to the beginning of the area (step 602) and adds the circulation sequence number (step) 603). On the other hand, if the output position of the journal block has not reached the end of the area, the DB update unit 101 sets the output position to the journal block next to the last write position of the previous journal block (step 605).

  When the DB block 101 determines the output position of the journal block, the DB update unit 101 writes the journal block to the output position of the journal volume (MV) 109 (step 606), and uses the output position as the journal block final write position 122. (Step 607) so that it can be acquired in step 601 at the next operation.

  Here, with respect to writing to the journal volume (MV) 109 in step 606, the operation of the storage apparatus 102 of the main host will be described in detail with reference to FIGS.

  The storage controller 103 included in the storage apparatus 102 receives a journal block write request (see FIG. 6) from the DB update unit 101 (step 701), and directly uses the received write request as a remote host using the replication information transmission unit 104. (Step 702). Thereafter, the journal block is written into the journal volume (MV) 108 (step 703), and a response to the write request is returned to the DB update unit 101 (step 704).

  Through the above operation, the journal output by the DB update unit 101 is transferred to the remote site 110.

  Next, an operation in which the replication information receiving unit 118 writes the journal block transferred from the main site 100 to the journal volume (RV) 120 will be described with reference to FIGS.

  First, when receiving a write request transferred from the main site 100 (step 801), the replication information receiving means 118 checks the current replication status held as the replication status information 119 (step 802). As described in the section of <Replication Information Receiving Unit>, the replication status includes a SYNC status in which sequential replication is operating and a REPLICATE status in which sequential replication is not operating, and writing is performed under the REPLICATE status. When roll forward is performed with the journal block thus obtained, the DB (RV) 121 becomes in an illegal state. In order to prevent this, the replication information receiving unit 118 uses the exclusion unit 114 in the REPLICATE state to determine the entire journal area (from the address of the journal block located at the head of the journal area to the journal located at the end of the journal area). After the block address is excluded (step 811), the journal block is written to the journal volume (RV) 120 using the function of the storage controller 113 (step 812). As a result, reading of the journal volume (RV) 120 by the roll forward unit 111 is suppressed, and illegal reflection of the journal to the DB (RV) 121 can be prevented.

  On the other hand, when the replication state is the SYNC state, the replication information receiving unit 118 acquires the replication speed information 126 (step 803), and the current replication speed and the previous replication speed stored in a memory (not shown) or the like. (Step 804). As a result of this comparison, if the current replication speed is higher than the previously acquired replication speed, the replication information receiving means 118 increases the number of journals to be excluded at one time (step 805). On the other hand, if the current replication speed is lower than the previous replication speed, the replication information receiving unit 118 reduces the number of journal blocks to be excluded at one time (step 806). As described in the section <Journal block>, since the position of the journal block is unique, that is, fixed, the number of exclusive journal blocks can be increased or decreased by increasing or decreasing the write end point address requested to the exclusion means 114. . The replication speed acquired in step 803 is stored in a memory or the like (not shown) for the next replication speed comparison (step 807).

  When the number of journal blocks to be excluded is determined, the replication information receiving unit 118 gives the exclusion unit 114 a write destination volume, a write start point address, and a write end point address as exclusion ranges, and requests exclusive control of the journal block. (Step 808). If the entire journal area is excluded in step 811, the exclusion range is left and the exclusion of other ranges is released.

  Although the detailed operation of the exclusion unit 114 will be described later, the exclusion unit 114 records the exclusion range requested in step 808 as the exclusion information 116. Then, the exclusion unit 114 checks whether the exclusion information 116 has already been registered at the time of the exclusion request by the roll forward unit 111, that is, whether the replication information reception unit 118 is acquiring the exclusion, and according to the confirmation result A read permission or rejection is returned to the roll forward means 111. With these operations, the exclusion unit 114 controls the writing of the journal block by the replication information receiving unit 118 and the reading of the journal block by the roll-forward unit 111 so as not to collide with each other.

When the exclusion in step 808 is completed, the replication information receiving unit 118 issues a write command using the function of the storage controller 113, writes a journal block to the journal volume (RV) 120 (step 809), and the exclusion unit 114. Is requested to release the exclusive (step 810).
If the journal block writing by the replication information receiving means 118 and the journal block reading by the roll forward means 111 collide, the roll forward means 111 moves to the journal block requesting exclusion at the end of this step 809. Will be allowed access.
Through the above operation, the journal block write request transferred from the main site 100 is reflected in the journal volume (RV) 120.

  Next, using FIG. 2 and FIG. 10, the means by which the roll forward unit 111 reads a journal block from the journal volume (RV) 120 and reflects it in the DB (RV) 121 will be described in detail.

  The roll forward unit 111 first acquires information at the time of the previous operation of the roll forward unit 111 in order to specify the reading position of the journal record (step 901). The information acquired here includes the end position of the last read journal block, the position of the start journal record of the unreflected transaction (the number of the record in which journal block), and the cyclic serial number of the journal block with the unreflected journal record It is. These pieces of information are acquired from information stored on a disk or the like when the roll forward unit 111 is activated. If the roll forward unit 111 is already operating, the journal block final read position 123, unread It can be acquired from the reflection journal record position information 124 and the roll forward target circulation serial number 125.

  Next, the roll-forward means 111 acquires the replication speed information 126 (step 902), compares the current replication speed with the previous replication speed, and if the replication speed has increased, journals that are exclusive at once. Is increased (step 904), and if the replication speed is reduced, the number of journals to be excluded at a time is decreased (step 905). Then, the replication speed acquired in step 902 for the next comparison is stored in a memory or the like (not shown) (step 906).

  The roll-forward means 111 continues with the exclusion means 114 included in the storage controller 113 as an exclusion range in order to prevent collision with the journal block write by the replication information reception means 118 as a read destination volume read start address and read end address. And request exclusion (step 907). The read start point address requested to the exclusion unit 114 is the address of the journal block in which there is an unreflected journal record acquired in step 901. Further, the read end point address is an address advanced by the number of exclusive journal blocks determined in steps 904 and 905 as seen from the read start point address. Here, if the read end address exceeds the end of the area, the read end address is aligned with the end of the area, and a request to exclude the remaining number of journal blocks from the address of the journal block located at the head of the journal area may be made. .

  When the exclusion unit 114 rejects the exclusion request with respect to the exclusion request in step 907, that is, when the acquisition of the exclusion fails, the journal block to be read is being written by the replication information receiving unit 118, so roll forward The means 111 waits (step 908) and then makes an exclusive request again at step 907. Waiting in step 908 includes a method in which the roll forward unit 111 is in a sleep state for a certain period of time, a method in which the exclusion unit 114 notifies the exclusion unit 114 that the replication information reception unit 118 has requested the exclusion unit 114 to release the exclusion. Conceivable. In the present embodiment, a method is adopted in which the roll-forward means 111 is in a sleep state for a certain time.

  On the other hand, when the exclusion unit 114 permits the exclusion request, the roll forward unit 111 reads one journal block (step 909) and determines the end of the journal by comparison with the circulation serial number acquired in step 901. (Step 910). If it is detected that the number of circulating serial numbers has decreased, that is, the journal end has been exceeded, the last read journal block should not be reflected in the DB (RV) 121, and no more journals Since the roll forward cannot proceed even if the block is read, the position of the last read journal block is stored in the journal block last read position 123 (step 918), and the process is terminated.

  On the other hand, if the journal end is not exceeded in step 910, the roll forward unit 111 determines whether the journal record included in the journal block can be reflected in the DB (RV) 121 (step 911).

  The determination in step 911 will be described in detail with reference to FIG.

  FIG. 11 shows the journal block 1000 read by the roll forward unit 111 during the previous operation and the journal block 1010 read during the current operation.

  In journal block 1000, transaction 2 (TR 2) starting with journal record 1002 ends with end journal record 1004. Therefore, the roll forward unit 111 reflects the journal records 1002, 1003, and 1004 related to TR2 in the DB (RV) 121 (step 912).

  On the other hand, the transaction 1 (TR1) starting from the journal record 1001 and the transaction 3 (TR3) starting from the journal record 1005 are not completed in the journal block 1000. Therefore, the roll forward means 111 cannot reflect the journal record 1001 related to TR1 and the journal record 1105 related to TR3 to the DB (RV) 121. Therefore, the roll forward unit 111 needs to read the journal block 1000 again at the next operation. As will be described later, the roll forward unit 111 stores the position of the start journal record of the transaction that could not be reflected in the DB (RV) 121 for the next operation in step 913, and manages it in step 918. The position of the most recent journal record among the recorded journal records is stored in the unreflected journal record position information 124, and the circulation sequence number of the journal block to which the journal record belongs is stored in the roll-forward target circulation sequence number 125. In the case of the example shown in FIG. 11, the roll forward unit 111 stores the position of the journal record 1001 in the unreflected journal record position information 124 and the circulation sequence number of the journal block 1000 in the roll forward target circulation sequence number 125.

  In the next operation, the roll-forward unit 111 refers to the unreflected journal record position 124, acquires the position of the previously unreflected journal record 1001 (step 901), and rereads the journal block 1000 (step 909). At this time, the roll forward unit 111 also reads the journal records 1002, 1003, and 1004 again, but these journal records are related to the transaction completed in the previously read journal block. Therefore, it can be determined that the data has already been reflected in the DB (RV) 121. Therefore, for these journal records 1002, 1003, and 1004, the roll forward unit 111 does not reflect in the DB (RV) 121.

  Next, the roll forward unit 111 reads the next journal block 1010. Here, when the roll forward means 111 finds the journal record 1014 indicating the end of TR1, the roll forward means 111 reflects the journal records 1001, 1011 and 1014 related to TR1 in the DB (RV) 121 (step 912). Further, the roll forward unit 111 deletes the position of the start journal 1001 from the unreflected journal record position information 124 (step 913).

  Similarly, when the roll forward unit 111 processes all journal records of the journal block 1010, the position of the start journal record managed by the unreflected journal record position information 124 becomes the position of the start journal record 1005 of TR3. The roll forward unit 111 further holds the position of the journal record 1005 at the unreflected journal record position 124 in preparation for the next operation of the roll forward unit (step 918).

  When step 912 is completed, the roll forward unit 111 updates the unreflected journal record position information 124 (step 913). As described above, if there is a transaction start journal that does not have a corresponding transaction end journal in the current journal block read in step 909, the position (which journal block number) is stored. If there is, the corresponding transaction start journal is searched for, and the position information of the transaction start journal is deleted.

  When the processing of the journal block read in step 909 is completed, the roll forward unit 111 confirms whether all journal blocks in the range for which exclusion is requested in step 907 have been processed (step 914). If the processing of all journal blocks in the exclusive range has been completed, the roll-forward unit 111 determines the information related to the current operation, that is, the journal block last read position 123 which is the end of the journal block read this time, the unreflected journal The record position 124 and the roll forward target circulation sequence number 125, which is the circulation sequence number of the journal block to which the journal record belongs, are saved, and the process ends (step 918).

  On the other hand, if all the journal blocks in the exclusive range have not been processed, the roll forward unit 111 advances the reading position of the journal block first (step 915), and checks whether the reading position has reached the end of the area (step 915). 916). If the end of the area has been reached, the next reading position is returned to the position of the journal block located at the head of the journal area, the circulation sequence number is incremented by 1, and the processing is restarted from step 909.

  Finally, the operation of the exclusion unit 114 will be described with reference to FIGS.

  When the exclusion unit 114 first receives an exclusion request from the replication information reception unit 118 or the roll-forward unit 111 (step 1101), the exclusion unit 114 determines which unit the request is from (step 1102). If it is a request by the replication information receiving means 118, the exclusion means 114 determines the request type (step 1103), and if it is exclusive acquisition for writing, the write volume and write notified by the replication information receiving means 118 The start point address and the write end point address are registered as exclusive information 116 (step 1104). At this time, if the entire journal area is exclusive in step 811 described above, the exclusive information for the entire journal area is updated with the content requested by the replication information receiving means 118. If the request type is exclusive release after completion of writing, the matching information is deleted from those registered in the exclusive information 116 (step 1105).

  On the other hand, if the exclusion request is from the roll forward unit 111, the exclusion unit 114 determines the exclusion information based on the requested read destination volume, the read start point address and the read end point address, and the exclusion request from the replication information receiving unit 118. The write destination volume, write start point address, and write end point address registered in 116 are collated (step 1106). If there is an overlapping part in both exclusion request areas, the exclusion unit 114 notifies the roll forward unit 111 of access refusal (step 1107). On the other hand, if there is no overlapping part, the access permission is notified to the roll forward means 111 (step 1108).

  In the present embodiment, the journal volume (RV) 120 adopts a circular journal structure in which the roll forward unit 111 can quickly identify the position of the latest journal block written by the replication information receiving unit 118. Yes. As a result, the roll forward unit 111 promptly determines how far replication has been completed, prevents the journal blocks after the latest journal block from being illegally reflected in the DB (RV) 121, and always keeps the latest journal. Can be reflected in DB.

  In the present embodiment, the exclusion unit 114 performs access exclusion control in units of journal blocks. Thereby, it is possible to prevent the roll-forward unit 111 from reading the journal that is being written by the replication information receiving unit 118, to prevent illegal reflection of the DB (RV) 121 due to access collision between the two, and to maintain consistency. .

  In this embodiment, the replication information receiving unit 118 has a function of managing the replication speed, and the replication information receiving unit 118 and the rollforward unit 111 exclude the number of journal blocks that are exclusive at a time according to the replication rate. Can be changed. As a result, when the replication speed is high, overhead due to exclusive control can be minimized and high-speed replication can be realized. On the other hand, when the replication speed decreases, the replication information receiving unit 118 and the roll forward unit 111 reduce the number of journals that are exclusive at a time. As a result, the roll forward unit 111 can proceed with replication up to the latest journal block. Therefore, it is possible to avoid a situation in which roll forward does not proceed due to exclusion waiting and RPO is extended.

  In this embodiment, the replication information receiving means 118 is provided with a function for managing the operation state of sequential replication. When the sequential replication is not operating, the replication information receiving means 118 has a journal block (RV). ) By reading the entire 120, journal reading by the roll forward unit 111 is suppressed. As a result, in a state where the replication is not sequential, such as at the start of replication or a line failure, illegal reflection of the DB (RV) 121 by a journal whose writing order is unknown can be prevented and consistency can be maintained. At this time, even if the roll-forward means 111 is kept operating regardless of the replication status, it does not adversely affect the DB (RV) 121, so that the operation of the remote site 110 can be simplified. .

  Furthermore, the present embodiment can be easily implemented by replacing a program (firmware) possessed by the storage controller of the current storage apparatus and changing its function.

  That is, according to the present invention, in a method of performing replication by transferring only journals between storage apparatuses, it is possible to achieve both reduction in RPO and low running cost.

  It should be noted that the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the present invention.

  For example, in the above-described embodiment, the present invention has been described as being mainly configured by hardware. However, the present invention is not limited to this, and any processing can be performed on a CPU (Central Processing Unit) with a computer program. It can also be realized logically by executing. In this case, the computer program can be stored and provided to the computer using various types of non-transitory computer readable media. Non-transitory computer readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (for example, flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (for example, magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / W, semiconductor memory (for example, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)). The program may also be supplied to the computer by various types of transitory computer readable media. Examples of transitory computer readable media include electrical signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can supply the program to the computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

DESCRIPTION OF SYMBOLS 100 Main site 101 DB update means 102 Storage apparatus 103 Storage controller 104 Replication information transmission means 105 Communication means 106 Replication status information 107 Journal block 108 Journal volume (MV)
109 DB (MV)
110 Remote site 111 Roll forward means 112 Storage device 113 Storage controller 114 Exclusive means 116 Exclusive information 118 Replication information receiving means 119 Replication status information 120 Journal volume (RV)
121 DB (RV)
123 Journal block last read position 124 Unreflected journal record position 125 Circulation number for roll-forward target 126 Replication speed information

Claims (10)

Replication information receiving means for receiving a journal block including update information of the replication source volume;
A journal volume including a storage area for storing a fixed number of the journal blocks in the order received;
Roll forward means for sequentially reading the journal blocks from the storage area and executing roll forward for updating the replication volume based on the read journal blocks;
The storage area is a circular structure in which, when the number of stored journal blocks exceeds the certain number, the stored journal blocks are overwritten by new journal blocks in order from the oldest,
Each of the journal blocks includes a circulation number indicating the number of times the overwriting has made a round in the circulation structure;
The roll forward means, when reading the stored journal block, determines the latest journal block based on the circulation count, and stops the roll forward with the latest journal block. The replication apparatus according to claim 1, further comprising an exclusive unit that performs exclusive control on the journal block when the journal block is stored in the journal volume. The replication apparatus according to claim 2, wherein the exclusion unit further performs exclusion control on the journal block when the roll forward unit reads the journal block. 4. The replication apparatus according to claim 1, wherein the roll forward unit changes the number of the journal blocks that execute the exclusive control at a time according to the reception speed of the journal blocks. 5. 5. The replication apparatus according to claim 1, wherein the replication information receiving unit performs exclusive control on the entire journal volume when the reception state of the journal block is in a state in which a replication order is not guaranteed. A replication information receiving step for receiving a journal block including update information of the replication source volume;
A storage step of storing a fixed number of the journal blocks in the storage area of the journal volume in the order received;
A roll forward step of sequentially reading the journal blocks from the storage area and executing a roll forward for updating a replication volume based on the read journal blocks;
The storage area is a circular structure in which, when the number of stored journal blocks exceeds the certain number, the stored journal blocks are overwritten by new journal blocks in order from the oldest,
Each of the journal blocks includes a circulation number indicating the number of times the overwriting has made a round in the circulation structure;
In the roll forward step, when the stored journal block is read, the latest journal block is determined based on the circulation count, and the roll forward is stopped using the latest journal block. The replication method according to claim 6, further comprising a first exclusive step of performing exclusive control on the journal block when the journal block is stored in the journal volume. The replication method according to claim 7, further comprising a second exclusive step of performing exclusive control on the journal block when reading the journal block in the roll forward step. The replication method according to any one of claims 6 to 8, wherein, in the roll forward step, the number of the journal blocks that execute the exclusive control at a time is changed according to the reception speed of the journal blocks. On the computer,
A replication information receiving step for receiving a journal block including update information of the replication source volume;
A storage step of storing a fixed number of the journal blocks in the storage area of the journal volume in the order received;
A roll forward step for reading the journal blocks from the storage area in order and executing a roll forward for updating a replication volume based on the read journal blocks,
The storage area is a circular structure in which, when the number of stored journal blocks exceeds the certain number, the stored journal blocks are overwritten by new journal blocks in order from the oldest,
Each of the journal blocks includes a circulation number indicating the number of times the overwriting has made a round in the circulation structure;
In the roll forward step, when the stored journal block is read, the latest journal block is determined based on the number of circulations, and the roll forward is stopped with the latest journal block.

Images