JP2004310163A - Mirror disk system and mirror disk controller - Google Patents

Abstract

An object of the present invention is to provide a mirror disk system and a mirror disk control device for shortening the data write coincidence (mirroring) time of two disk devices in a disk system of a RAID redundant configuration.
In a RAID 1 system, a disk controller (3) includes a cache memory (3d) for storing write / read data from the host computer and a non-volatile memory (3e) for storing a write request history management table. When there is a mirroring process request for matching the data to be written to one or two disk devices 4A and 4B, the mirroring process is omitted for data that has passed a predetermined write time by referring to the history management table. Control.
[Selection diagram] Fig. 1

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a mirror disk control device and a mirror disk system, and more particularly, to a mirror disk control device and a mirror disk system for reducing the time required for a mirroring operation for controlling writing and reading of data to and from a mirror disk device.
[0002]
[Prior art]
In a disk array device for the purpose of improving the reliability of a storage device, a redundant configuration in which data is multiplexed and stored in a plurality of disk devices in order to prevent writing and reading of data from being impossible due to a failure of the storage device. I have.
[0003]
Various configurations of this redundant configuration are known as RAID (Redundant Array of Inexpensive Disks). In a redundant configuration known as a RAID1 name among these several types of configurations, even if one of the two or two systems of disk devices fails, the same data is transferred to another disk device. Since the information is stored, it is possible to prevent the system from being stopped, and it is employed in a system requiring high reliability.
[0004]
For example, in a social system such as medical care and finance, there is a mirror disk system that uses a disk device with high reliability by mirroring between disk array devices.
[0005]
In such a conventional data duplication system, the processing time for writing an increased amount of data to the disk device increases, and the data bus between the host computer and the host computer is not opened for a long time, thereby deteriorating the performance of the system. was there.
[0006]
Further, when the system fails, when the power is turned on again after the failure and the system is restored, a mirroring process for matching the contents of a plurality of disk devices is required. In this case, it takes a long time to sequentially process the entire area of the disk device, and there is a problem that it takes too much time to recover from a system failure.
[0007]
In a disk device using a redundant configuration of RAID1, with regard to a method of shortening the writing time when writing the same content to two disk devices in data duplication, a method of writing to only one device without performing simultaneous writing is known. It is known (Patent Document 1).
[0008]
In this method, writing data to the first disk device serving as a master and the second disk device serving as a slave first stores the data in a buffer memory of the first disk device, and stores the data in the first disk device. After writing, the data is transferred from the buffer memory of the first disk device to the second disk device for writing, thereby shortening the writing time.
[0009]
However, in this method, there is a problem that the responsiveness of the system is deteriorated when a read request is issued from the host computer during the write operation. Therefore, a method of performing writing at a different cycle for each of two disk devices is known. Patent Document 2.).
[0010]
[Patent Document 1]
JP-A-10-74129
[Patent Document 2]
JP 2000-20250 A
[Problems to be solved by the invention]
The above-mentioned mirror disk system according to the prior art is effective for improving the throughput of the system during normal operation. However, when the system is recovered from an abnormal situation such as when the power is turned on again, further recovery from failure is required. It is necessary to reduce the time, but there is no known method of reducing the writing time in such a case.
[0013]
SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and there are cases where there is a request process for matching write data to two disk devices from a host computer, or for system power-on again. It is an object of the present invention to provide a mirror disk system and a mirror disk control device capable of processing request processing for matching data over the entire area of a disk storage area of a disk device in a short time as in the case.
[0014]
[Means for Solving the Problems]
In order to achieve the above object, according to the first aspect of the present invention, there is provided a mirror disk system comprising: two disk devices; and a disk control device which receives an input / output request from a host computer and controls input / output of the disk device. A mirror disk system comprising: a cache memory for storing write / read data from the host computer; and a non-volatile memory for storing a write request history management table. When there is a mirroring process request for matching the data to be written to the two disk devices, the disk controller refers to the history management table and skips the mirroring process for data that has passed a predetermined write time. Control as described above.
[0015]
Therefore, the mirroring processing time can be minimized, and the system recovery time can be shortened even when the power is turned on again.
[0016]
In the mirror disk control device provided between the host computer and the first and second disk devices each storing the same data, a cache memory for storing write data received from the host computer may be provided. A nonvolatile memory for storing a history management table of write requests from the host computer, and controlling writing to the disk device with reference to the history management table.
[0017]
Accordingly, the bus occupation time due to the access of the host computer to the disk control device can be reduced, so that the efficiency of the system can be improved. Further, since the number of times of disk access of the disk device can be reduced, the life of the disk device is also improved.
[0018]
BEST MODE FOR CARRYING OUT THE INVENTION
(First Embodiment)
Hereinafter, a first embodiment of the present invention will be described with reference to FIGS. FIG. 1 shows an example of the configuration of the present invention. The mirror disk system is connected to the host computer 1, the host computer 1, one of which is connected to the system bus 2, and the other is connected to two disk devices 4 A, 4 B, and controls a data write / read request from the host computer 1. A disk control device 3 is provided. The two disk control devices 4A and 4B include a disk device 4A as a master and a disk device 4B as a mirror.
[0019]
Further, the disk control unit 3 includes a bus interface 3a for controlling an interface with the host computer 1, an internal bus 3b of the disk control unit 3, an arithmetic control unit 3c including a microprocessor, and the host computer 1 and the disk control unit 3. A cache memory 3d for holding data transferred between the two, a nonvolatile memory 3e for storing a history management table for managing a data write request history, and data for transferring data in the cache memory 3d to two disk devices 4A and 4B. And a transfer control unit 3f.
[0020]
In the arithmetic control unit 3c of the disk control device 3 configured as described above, a control program for controlling writing and reading to and from the disk devices 4A and 4B is mounted.
[0021]
First, the write / read operation of the control operation by this control program will be described, and the mirroring processing operation for matching the data written in the two disk devices 4A and 4B using the write request history management table will be described.
[0022]
In FIG. 1, when a write request is transmitted from the host computer 1 to the disk controller 3 via the bus interface 3a, first, the write data is written into the cache memory 3d.
[0023]
The arithmetic control unit 3c of the disk control device 3 sends the data stored in the cache memory 3d to the data transfer control unit 3f in accordance with the content of the write request from the host computer 1, and sets the data as a master. In the disk device 4A or 4B.
[0024]
Upon receiving the write command, the data transfer control unit 3f transfers the data in parallel to the disk devices 4A and 4B or to one of the master disk devices 4A and 4B, and returns a write completion message from the disk devices 4A and 4B. Then, the fact that the writing has been completed is transmitted to the host computer 1 via the control operation unit 3c, and the writing process is completed.
[0025]
The mirror disk system writes data to two disk devices 4A and 4B at the same time, or writes data to one master device and then uses the idle time of the system to write data to the mirror disk device. There are some which perform writing, but in the present invention, this writing control is applied to any case. Further, the two disk devices are also applied to a case where each of the disk devices is composed of a plurality of disks in an array configuration.
[0026]
When a read request from the host computer 1 is transmitted to the disk control device 3 via the bus interface 3a, the arithmetic control unit 3c sends the read request to the data transfer control unit 3f in accordance with the content of the read request from the host computer 1. At the same time as transmitting the command content, the control unit instructs the cache memory 3d to transmit the data stored in the active disk device 4A or 4B serving as the master.
[0027]
The data transfer control unit 3f receiving the read command transfers the contents to the disk device 4A or 4B, receives the data from the disk device 4A or 4B, and the control operation unit 3c transmits the data to the cache memory 3d. The data is transmitted to the host computer 1 via the bus interface 3a to complete the reading process.
[0028]
The above operation is a normal operation of the mirror disk system that enables data redundancy corresponding to RAID1. In this case, when the host computer 1 issues a request for matching data called mirroring processing written in the disk devices 4A and 4B, a process for matching the write data over the entire area of the disks of the respective disk devices 4A and 4B. In order to perform the operation, a system having a large storage capacity may require several hours.
[0029]
Next, the mirroring processing operation will be described with reference to FIGS. FIG. 2 shows an example of a history management table in which a write request is issued from the disk control device 3 to the disk devices 4A and 4B. FIG. 3 shows an operation flow of the disk control device 3 using this history management table. This history management table is stored in the non-volatile memory 3e shown in FIG. 1, and is controlled by a control program mounted on the arithmetic control unit 3c (not shown).
[0030]
FIG. 2A is an example of a history management management table in which the occurrence of a write request is recorded from the disk control device 3 of the system shown in FIG. 1 to the disk devices 4A and 4B. Each time a write occurs, the history management table records the identification name of the data, the date and time when the write request occurred, and the address of the data.
[0031]
In the figure, for example, there are N, M, L, and K as data identification names, and access histories from a new one in which a write access has occurred until a predetermined time has elapsed are sequentially stored in this history table. . In this example, for the data identification name N, three histories of N + 0, N-1, and N-2 are recorded in the order of the newest write access.
[0032]
In this history management table, an update cycle is set in advance, and old data is automatically deleted assuming that the writing process has been completed for the two disk devices 4A and 4B. Therefore, this update cycle interval time is set by measuring the write processing time in advance when constructing the system.
[0033]
For example, assuming that the time for performing the all-area mirroring process is t, a predetermined time is set as the update interval time T of the history management table as T> t.
[0034]
The history management table recorded in this manner is shown in FIG. 2A, and FIG. 2B shows the history management at 13:30 after the update time 13:00 and the update cycle T time in FIG. 2A. Show the table. In the data identification names N-2, M + 0, L-2, and K + 0 in FIG. 3A, the update interval time T (for example, T = 30 minutes) from the occurrence time of each write access in FIG. This indicates that the data before 12:30 has been automatically deleted from the history management table since one hour or more, which is twice as long as the predetermined time, has elapsed.
[0035]
Furthermore, if the host computer 1 requests a mirroring process from the host computer 1 at 13:40 after the update time in FIG. 2B, or if a system failure occurs and the power is turned on again, in this case, Since the history management table stored in the non-volatile memory retains the state before the access from the host computer 1 or the power is turned on again, the mirroring process is executed by referring to the past access time recorded in the history management table. I do.
[0036]
Next, a control operation of the mirror disk system using the history management table stored on the nonvolatile memory controlled as described above will be described with reference to FIGS.
[0037]
In FIG. 3, in this control operation, when the disk controller 3 receives this access in response to a power-on of the system or a data write request from the host computer, a write control program for the disk devices 4A and 4B starts.
[0038]
Next, it is determined whether the write request command is a normal access or a mirroring access (step S01 in FIG. 3).
[0039]
Here, normal access indicates a normal write processing request, and mirroring access indicates a processing request for matching data stored in the two disk devices described above.
[0040]
If this request is a normal access, the history management table stores the contents of this access. As described with reference to FIG. 2, for this access, the identification name of the data, the date and time of access, and the address of the data are written (step S04 in FIG. 3).
[0041]
Further, the data at this time is written to the disk devices 4A and 4B (Step 05 in FIG. 3). This write operation is a normal processing flow.
[0042]
If the request from the host computer 1 is a mirroring process, a mirroring mode determination process is performed (step S02 in FIG. 3).
[0043]
In this process, it is determined whether the mirroring is the whole area mirroring or the history management area mirroring according to the mirroring mode set in advance, and the process proceeds to the next processing. If this setting is all-area mirroring, the request mirroring process processes all areas of the storage data of the disk devices 4A and 4B (step S03 in FIG. 3). This operation is performed when the system is first started.
[0044]
Thereafter, when a mirroring processing request is issued from the host computer 1, this mode is switched to the history management area mirroring processing mode. This mode switching method can be performed automatically or manually.
[0045]
In the case of the history management mirroring process, first, the content of the occurrence of this access is recorded in the history management table (step S04 in FIG. 3). At the same time, the request data is written to the disk devices 4A and 4B (step S05 in FIG. 3).
[0046]
The history management table updates the access content of the write request at a fixed interval time, and deletes the old access record data after a predetermined time has elapsed (step S06 in FIG. 3).
[0047]
In this way, the processing at the time of starting up the system is performed. Normally, each time a write request is made, the contents are recorded in the history management table and the write processing is executed on the disk devices 4A and 4B. Then, this write history management table is updated at regular intervals.
[0048]
If a failure occurs in the system operating in this state for some reason and the system is restarted, the history management area mirroring process is performed only on the area stored in the history management table.
[0049]
For example, the state of the history management management table at the current time 13:16 in FIG. 3A is changed to the history management table at the update time 13:30 at the current time 13:41 in FIG. The data has been updated, and the data before 12:30, that is, the data names N-2, L-2, and K + 0 have been deleted.
[0050]
Therefore, when the mirroring processing request is issued at 13:40, the history management area mirroring processing is performed not in the entire area mirroring processing but in the area limited to only the identification data recorded in the history management table.
[0051]
That is, when the system is restarted from a system failure or the like, since the history management of the write request before the failure is maintained, the mirroring process over the entire storage area of the disk as in the related art becomes unnecessary, and the system restarts. Raising time is reduced.
[0052]
【The invention's effect】
As described above, since the mirroring processing time is predicted in advance and the history management table requested to be written by the host computer is stored in the non-volatile memory, this data is used when the system is restarted. The mirroring process can be performed in a short time by referring to the data. Therefore, the recovery time from a system failure is reduced.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a system according to an embodiment of the present invention.
FIG. 2 is a flowchart of a mirroring processing operation of the present invention.
FIG. 3 is an example of a history management table according to the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Host computer 2 System bus 3 Disk controller 3a Bus interface 3b Internal bus 3c Operation controller 3d Cache memory 3e Non-volatile memory 3f Data transfer controller 4A, 4B Disk device

Claims (4)

In a mirror disk system including two disk devices and a disk control device that controls input / output of the disk device in response to an input / output request from a host computer,
A cache memory that stores write / read data from the host computer;
A nonvolatile memory for storing a history management table of write requests,
When there is a mirroring request from the host computer requesting matching of data to be written to the two disk devices, the disk control device refers to the history management table to check data having passed a predetermined write time. A mirror disk system characterized by performing control so as to omit a mirroring process. The history management table stores at least a time at which a write request is issued from the host computer and an address of the data, updates the data every time a predetermined time elapses from a previous write request, and updates a predetermined time since a previous update request. 2. The mirror disk system according to claim 1, wherein the data in said history management table is deleted. In a mirror disk control device provided between a host computer and first and second disk devices each storing the same data,
A cache memory for storing write data received from the host computer,
A nonvolatile memory for storing a history management table of a write request from the host computer,
A mirror disk control device, wherein writing to the disk device is controlled with reference to the history management table. The history management table stores at least the time of the write request from the host computer and the address of the data,
4. The mirror disk according to claim 3, wherein the mirror disk is updated every time a predetermined time elapses from a previous write request, and data in the history management table after a predetermined time elapses from the previous update request is deleted. Control device.

Images