JP2004094710A - Data reduplicating method for storage subsystem and system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a data reduplicating method for storage subsystems and its system for securing the compatibility of data among the plurality of storage subsystems on the remote center side even if the main center side is abnormally terminated by a calamity, or the like. <P>SOLUTION: A main center 101 outputs synchronization information (CP) to all the storage subsystems at a synchronizing point for transaction. Storage subsystems 105 (110) communicate the received CP to storage subsystems 125 (140) on the remote center side 121 by means of a copying function of the subsystems 105 (110). The subsystems 125 (140) having received the CP communicate with peripheral storage subsystems. When CP acquired at the same time reaches all the subsystems 125 (140) including own storage subsystem, renewal information having reached prior to the CP is reflected in storage media 129 (145). <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a storage subsystem for storing data to be referenced and updated by a computer, and more particularly to a method and system for duplicating data held by the storage subsystem.
[0002]
[Prior art]
In preparation for a disaster such as an earthquake, a remote center system is installed at a location several hundred kilometers away from the main center system that normally performs business, and data is duplicated between these systems. However, some are already in practical use.
[0003]
In this type of data duplication system, information on updating the master file of the main center system (hereinafter referred to as update information) is transmitted to the remote center system via a line or the like, and the contents of updating the master file in the main center system Generally, a method of updating the backup file of the remote center side system so as to have the same contents as described above is adopted.
[0004]
Recently, a disaster backup method using a copy function between storage subsystems included in a storage subsystem has been proposed, and Japanese Patent Application Laid-Open No. 2001-282628 discloses a technique as an example. According to the method disclosed in the publication, the backup file is updated on the remote center side in accordance with the order of the update requests generated on the main center side.
[0005]
[Problems to be solved by the invention]
In the case of a data duplication system that performs input / output to a plurality of storage subsystems, if each storage subsystem copies to a remote center at a remote location at an arbitrary timing, depending on the timing at which the main center is damaged, It is conceivable that the data between the plurality of storage subsystems on the remote center side becomes inconsistent and the data copied to the remote center side cannot be used.
[0006]
Regarding this problem, Japanese Patent Application Laid-Open No. 2001-282628 discloses that the same time is shared between storage subsystems, a serial number and a time are given to data to be transmitted to the remote center, and the remote center side A method of updating the disk storage device has been proposed. Similarly, Japanese Patent Application Laid-Open No. 6-290125 proposes a remote data backup method in which the data is reflected on the secondary business site in accordance with the update order on the main business site. In these methods, since updates are performed in the order of the generated update requests, for example, a mechanism for sharing the same time between storage subsystems is required. To reflect in the update order means that the main office main center and the secondary office remote center need a disk or the like to store a journal, etc., which is information indicating the update order. There is a problem that the cost of a line or the like is required to transfer the data.
[0007]
In the method of Japanese Patent Application Laid-Open No. 2001-282628, the in-flight transaction is settled on the remote center side (updates performed by transactions that have not reached the synchronization point are rolled back (returned to the state before the update) and synchronized). Since the update performed by the transaction that has reached the point needs to be rolled forward (validating the update), in addition to the master file, the history information (post-update information) acquired by the higher-level device is also It needs to be sent to the remote center. This is because it is impossible to know how far to roll back or roll forward without looking at the history information.
[0008]
The present invention has been made in view of the above-described problems in the related art, and can ensure data consistency among a plurality of storage subsystems on a remote center side even when the main center side is abnormally terminated due to a disaster or the like. In this case, an object of the present invention is to provide a method and system for data duplication of a storage subsystem in which it is not necessary to share the same time between storage subsystems or to send history information to a remote center.
[0009]
[Means for Solving the Problems]
To achieve the above object, in the present invention, the main center side system outputs synchronization information (CP) to all the storage subsystems at the transaction synchronization point. The storage subsystem notifies the received CP to the storage subsystem on the remote center side by the copy function of the storage subsystem. The storage subsystem on the remote center side that has received the CP contacts the surrounding storage subsystems, and the CP obtained at the same timing in the system on the main center side reaches all the storage subsystems on the remote center side. During this time, the update information of the file received after the previous CP reception is held in the buffer device, and is not reflected on the storage medium. When the CP acquired at the same timing reaches all the storage subsystems, the update information that has arrived before that is reflected on the storage medium. In this manner, consistency of the master file between the remote center-side storage subsystems is ensured.
[0010]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments according to the present invention will be described in detail.
[0011]
FIG. 1 shows the overall configuration of a system to which the present invention is applied. This system includes a main center 101 and a remote center 121. The main center 101 includes a host device 102 on which an online program or the like operates and a plurality of storage subsystems. In FIG. 1, two storage subsystems 105 and 110 will be described as an example.
[0012]
The host device 102 is a host computer in a main center having a central processing unit (CPU) that performs data processing, and includes a file update control 103 and a CP write control 104. The file update control 103 issues an input / output request to the storage subsystem 105 (110) in the main center. The CP write control 104 outputs synchronization information (CP) to all the storage subsystems 105 (110) in the main center at the transaction synchronization point.
[0013]
A plurality of storage subsystems 105 (110) in the main center 101 include an input / output control 106 (111), a transmission control 107 (112), and a storage medium 108 (113). Input / output processing is performed according to the output request. The input / output control 106 (111) performs input / output processing on the storage medium 108 (113). The transmission control 107 (112) performs data transmission (data copy) to the remote center 121 at a remote location.
[0014]
The remote center 121 located sufficiently away from the main center 101 is provided with a plurality of storage subsystems. FIG. 1 illustrates two storage subsystems 125 and 140 as an example. The remote center 121 reflects the update information transmitted from the storage subsystem 105 (110) of the main center on the storage medium.
[0015]
The storage subsystem 125 (140) includes a medium reflection control 130 (146) and a buffer device 128 (143) in addition to the input / output control 126 (141), the reception control 127 (142), and the storage medium 129 (145). . The input / output control 126 (141) performs control of storing data received from the storage subsystem 105 (110) of the main center 101 in each buffer device 128 (143). The reception control 127 (142) controls data reception from a remote place. The storage medium 129 (145) is for backing up (duplicating) data of the storage subsystem 105 (110) of the main center 101. The medium reflection control 130 (146) performs control to reflect the update information on the storage medium 129 (145). With this control, all the storage subsystems 125 (140) of the remote center 121 communicate with each other, and only when all the storage subsystems 125 (140) receive a CP having the same serial number, The update information is reflected on the storage medium 129 (145). Until CPs are collected, the update information is accumulated in the buffer device 128 (143).
[0016]
FIG. 2 is a diagram illustrating a data flow in the system of FIG.
[0017]
The host device 102 in the main center 101 sends update information (data for which the CPU has issued a write request to the storage subsystem) A1, update information A2, and update information B1 to the storage subsystems 105 and 110, respectively. It is assumed that the update information B2 and the update information B3 have been output ((1) in FIG. 2). This process is performed by the file update control 103. At this time, at the transaction synchronization point, the synchronization information CP1 and CP2 are output to all the storage subsystems 105 and 110. This process is performed by the CP writing control 104.
[0018]
Upon receiving the data, the storage subsystems 105 and 110 output the data portion of the update information excluding the CP to the storage media 108 and 113, respectively. This processing is performed by the input / output control 106 (111). Thereafter, the storage subsystems 105 and 110 transfer the data including the CP to the corresponding storage subsystems 125 and 140 on the remote center 121 side ((2) in FIG. 2). This processing is performed by the transmission control 107 (112).
[0019]
The storage subsystems 125 and 140 of the remote center 121 store the received data in the buffer devices 128 and 143, respectively. This processing is performed by the input / output control 126 (141). At this time, it is assumed that the CP 2 of the storage subsystem 110 has not reached the storage subsystem 140. Data transfer from the storage subsystem of the main center 101 to the storage subsystem of the remote center 121 is performed independently and asynchronously for each storage subsystem. Therefore, only the CP 2 of the storage subsystem 110 may not reach the remote center.
[0020]
The storage subsystems 125 and 140 in the remote center 121 communicate with each other, confirm that all the storage subsystems 125 and 140 have reached CP1, and output data before CP1 to the storage medium. ((3) in FIG. 2). This process is performed by the medium reflection control 130 (146).
[0021]
By such a flow, the state of the storage subsystem 125 (140) on the remote center 121 side can be a state where consistency at the time when the CP1 is output on the main center side 101 is ensured. This is because a CP is not issued in the middle of a transaction, and when the CP is issued, all transactions up to that point must have been processed.
[0022]
If the main center 101 is damaged at this time, the remote center 121 can continue the system in the state at the time when the main center 101 outputs CP1 without performing the recovery process using the history information. it can.
[0023]
A specific example will be described with reference to FIG. Now, the storage subsystems 108 and 113 are both databases storing information on bank accounts. It is assumed that a process of withdrawing 10,000 yen from the account in the storage subsystem 108 and transferring the 10,000 yen to an account in the storage subsystem 113 is executed. Immediately thereafter, the main center 101 is damaged and only the update information of the storage subsystem 108 (information that 10,000 yen has been drawn) reaches the remote center 121, and the update information of the storage subsystem 113 (10,000 yen is transferred). If the information does not reach the remote center 121, 10,000 yen will be lost in the conventional method. Alternatively, since the order of communication cannot be guaranteed, the update information of the storage subsystem 113 may arrive first and the update information of the storage subsystem 108 may not arrive. In this case, it is assumed that 10,000 yen has not been withdrawn. Become.
[0024]
In the system according to the present embodiment, in order to prevent the above-mentioned problem from occurring, the state of the DB at the moment when there is consistency on the main center 101 side is created on the remote center 121 side as it is. This is realized by checking the synchronization information CP. That is, the data before the CP received by all the storage subsystems in the remote center 121 is output to the storage medium. Since the CP is output at the transaction synchronization point, in the above example, the update information of the process of withdrawing 10,000 yen reaches the remote center 121, and the update information of the process of transferring 10,000 yen reaches the remote center 121. If it has not arrived, the CP corresponding to those processes has not reached the remote center 121. Therefore, the update information of the process of extracting 10,000 yen on the remote center 121 side may be written in the buffer, but is not written in the DB, so that the consistency of the DB is maintained. In this case, it is necessary to restore the withdrawal and the transfer from the history information. If the CP arrives at the remote center 121, it means that the update information of both the process of withdrawing 10,000 yen and the process of depositing 10,000 yen is reflected on the DB which has reached the remote center 121. At this time, the update information of the process of withdrawing 10,000 yen and the update information of the process of transferring 10,000 yen are sent to the remote center 121 side asynchronously, and whichever has arrived first. When the CP arrives, the update information up to that point is reflected in the DB, so that the consistency of the DB of the remote center 121 is always maintained.
[0025]
FIG. 3 shows a processing procedure of the file update control 103 and the CP write control 104 on the main center 101 side. The process of the file update control 103 is executed for each storage subsystem, and the process of the CP write control 104 is executed once for all the storage subsystems.
[0026]
In step 301, the file update control 103 receives an update request from the CPU and issues the update request to the storage subsystem 105 (110). In step 302, it is determined whether it is the CP acquisition timing. If it is not the CP acquisition timing, the process returns to step 301. When the notification of the CP acquisition timing is received from the CP writing control 104, the process proceeds from step 302 to step 303, and the completion of preparation for CP acquisition is notified to the CP writing control 104. Next, in steps 304 and 305, the process waits until acquisition of CPs of all storage subsystems is completed. When the CP writing control 104 notifies that the CP acquisition has been completed, the process returns to step 301.
[0027]
In step 311, the CP writing control 104 waits for a predetermined time until the CP acquisition timing comes. When the CP acquisition timing comes, in step 312, the CP acquisition timing is notified to the file update control 103. In steps 313 and 314, the process waits until the file update control 103 corresponding to all the storage subsystems 105 (110) completes preparation for CP acquisition. When the preparation of the CP acquisition is completed in all the file update controls 103, the same serial number CP is output to all the storage subsystems 105 (110) in step 315, and the CP is acquired. In step 316, the completion of CP acquisition is notified to the file update controls 103 of all the storage subsystems 105 (110), and the process returns to step 311.
[0028]
FIG. 4 shows a flowchart example of the processing of the medium reflection control 130 (146) of each storage subsystem of the remote center 121. This processing starts when data reception from the main center 101 is started.
[0029]
In step 401, the serial numbers of the CPs in the buffer device that reach all the storage subsystems including the self storage subsystem are confirmed (communication is established). In step 402, it is determined whether or not there is a serial number that has not been output (not yet output to a storage medium) and is common to all storage subsystems including its own storage subsystem. There is no unoutput CP in the buffer device, or even if there is an unoutput CP, a serial number common to all the storage subsystems (confirm the serial number of the unoutput CP in the buffer devices of all the storage subsystems) If not, go to step 405. If there is a corresponding CP, in step 403, data before the CP is output to the storage medium. Then, in step 404, the data before that including the CP is deleted from the buffer device.
[0030]
In step 405, it is determined whether the reception processing in all the storage subsystems has been completed or there is an end instruction (interruption of the reflection processing due to the damage of the main center 101) by the operator on the remote center 121 side. I do. Otherwise, proceed to step 406 and after an appropriate waiting time, return to step 401.
[0031]
FIG. 5 is a diagram showing an example of the data format of the CP. The CP includes a CP identifier 501 for identifying that the data is a CP, and a serial number 502 assigned by the CP writing control on the main center side. Instead of the serial number, the time at which the CP was issued may be used.
[0032]
【The invention's effect】
As described above, according to the present invention, by modifying the host device, a plurality of storage subsystems on the remote center side can be shared without sharing the same time among the plurality of storage subsystems on the main center side. Data consistency can be ensured between them. Further, since it is not necessary to determine the transaction in progress, it is not necessary to transmit the history information acquired by the higher-level device to the remote center. Furthermore, in the conventional method in which data is reflected on the remote center side according to the update order on the main center side, a disk or the like for storing a journal or the like is required, and a line cost for transferring the journal or the like is also required. In the present invention, there is no need to transfer the update order, so there is no need to transfer journals and the like, and line costs and the like can be reduced.
[Brief description of the drawings]
FIG. 1 is a diagram showing an overall configuration of a data duplication system of a storage subsystem to which the present invention is applied.
FIG. 2 illustrates the flow of data according to the present invention.
FIG. 3 is a diagram showing a processing procedure on the main center side.
FIG. 4 is a diagram illustrating an example of a flowchart of a medium reflection control process according to the embodiment;
FIG. 5 is a diagram showing an example of a data format of a CP according to the present invention.
[Explanation of symbols]
101: Main center 102: Host device 103: File update control 104: CP writing control 105, 110 ... Storage subsystems 106, 111 ... I / O control 107, 112 ... Transmission control 108, 113 Storage medium 121 Remote center 125, 140 Storage subsystem 126, 141 Input / output control 127, 142 reception control 128, 143 Buffer device 129, 145 ... storage media 130, 146 ... medium reflection control

Claims (2)

A data duplication method for a storage subsystem that realizes data duplication between a plurality of storage subsystems of a main center and a plurality of storage subsystems of a remote center,
On the main center side, reflecting the received file update information on a storage medium of a predetermined storage subsystem on the main center side;
Outputting the synchronization information (hereinafter, referred to as CP) to all the storage subsystems on the main center side at the transaction synchronization point on the main center side;
Transmitting the file update information and the CP asynchronously from each storage subsystem on the main center side to each storage subsystem on the remote center side using the copy function of each storage subsystem; ,
In each of the storage subsystems on the remote center side, a step of storing the received file update information and the CP in a buffer device;
In each storage subsystem on the remote center side, a step of determining whether or not the CP has been prepared in all the storage subsystems on the remote center side including the own storage subsystem;
Reflecting the update information of the file before the CP stored in the buffer device in a storage medium when the CPs are prepared in all the storage subsystems on the remote center side. Subsystem data duplication method. A data duplication system of a storage subsystem that realizes data duplication between a plurality of storage subsystems of a main center and a plurality of storage subsystems of a remote center,
Means for reflecting, on the main center side, the received file update information in a storage medium of a predetermined storage subsystem on the main center side;
Means for outputting synchronization information (hereinafter, referred to as CP) to all storage subsystems on the main center side at a synchronization synchronization point of the transaction on the main center side;
Means for transmitting the file update information and the CP asynchronously from each of the storage subsystems on the main center side to each of the storage subsystems on the remote center side using the copy function of each storage subsystem. ,
Means for holding the received file update information and the CP in a buffer device in each storage subsystem on the remote center side;
Means for determining whether or not the CP has been prepared in all the storage subsystems on the remote center side, including the self storage subsystem, in each storage subsystem on the remote center side;
Means for reflecting, on a storage medium, update information of a file before the CP held in the buffer device when the CPs are prepared in all storage subsystems on the remote center side. Subsystem data duplication method.

Images