JPS6326406B2 - - Google Patents

Description

[Detailed Description of the Invention] (a) Technical Field of the Invention The present invention relates to a data processing system, and particularly to a data processing system that writes the same data on a plurality of physically different volumes and logically treats these multiple volumes as one volume. The present invention relates to a data processing system having a multiplexed volume function that allows processing of multiple volumes.

(b) Prior art and problems The effects of failures in input/output devices or input/output paths have been reduced by multiplexing data on external storage, especially direct access storage devices (DASD), in units of volumes. There is. For this purpose, it is necessary to place the same data on a plurality of physically different volumes and to make it possible to treat these volumes logically as one volume.

FIG. 1 is a conceptual diagram of a multiplexed volume, in which volumes indicated by solid lines are physical volumes, and volumes indicated by broken lines are logical volumes. As shown in the figure, the same data A is stored in all physical volumes. The only volume that the user's general program is aware of is the logical volume shown by the broken line.
Only one piece of data A is visible.

In order for physical volumes 1 to n shown in FIG. 1 to be logically seen as one volume, all physical volumes must have the same data. In this way, "equivalence" means that all volumes have the same data.
is guaranteed.

This "equivalence" is lost due to a system failure during a data update (write operation) or an I/O failure during a data update (when this occurs for at least one physical volume).

If this equivalence is lost, it is necessary to perform equivalence guarantee processing between volumes at the next system startup or volume mount.

Conventionally, this method has been adopted by copying the entire volume, or by comparing the contents of the volumes and copying only the parts that differ. However, both of these methods require a large amount of time and overhead, and are not suitable for a system having a large volume or a large number of multiplexed volumes. In particular, there was a drawback that the system start-up time was significantly increased.

(c) Purpose of the Invention The present invention provides a data processing system having multiplexed volumes, in which the time required for equivalence guarantee processing,
The purpose is to reduce overhead and shorten system startup time.

(d) Structure of the Invention In order to achieve the above object, the present invention makes it possible to write the same data on a plurality of physically different volumes and to treat these plural volumes as one logical volume. In a data processing system having a multiplexed volume function, a journal record area is provided in each of the plurality of physical volumes, and when a write request is made to the physical volume, update range information indicating a predetermined update range in which write processing is performed; Create journal information that includes version number information indicating the update frequency, and write the journal information in each of the journal record areas of the plurality of physical volumes, so that when starting up after a system down, etc. The present invention is characterized by referring to the contents of the journal record area of the physical volume, identifying the data area that was in the updating state based on the update range information, and performing equivalence guarantee processing for the data contents of the data area. do.

(e) Examples of the invention Hereinafter, the present invention will be explained with reference to the drawings.

FIG. 2 is a block diagram of a data processing system according to an embodiment of the present invention, FIG. 3 is a diagram showing the format of a journal record in the embodiment, FIG. 4 is a diagram showing the operation flow of equivalence restoration processing, and FIG. FIG. 2 is a diagram showing a mode of data copying between volumes.

In Figure 2, 1 is the central processing unit (CPU)
and a data processing device consisting of main memory (MM), etc. 2 and 3 are direct access storage devices (DASD)
4 is an input/output request source program, 5 is an input/output control program, 6 is a system control program, 7 is a journal record area, 8
1 is a data area, 9 is a journal update display processing section, 10 is an input/output control section, 11 is a journal update display deletion processing section, and 12 is an equivalence recovery processing section.

In FIG. 3, reference numerals 20 and 21 are version number information holding units, and 22 is an update range information holding unit that shows the cylinder being updated in bitmap format.

The operation of the embodiment is as follows.

First, when the input/output requesting program 4 requests a write request to the input/output control program 5, the journal updating display processing section 9 displays "Updating" in the journal record area 7 on the main memory (MM). That is, in the update range information holding unit 22 shown in FIG. 3, the bit corresponding to the cylinder containing the update data is turned on ("1"). In addition, in FIG. 3, each bit of the update range information holding section 22 is stored corresponding to a specific cylinder, and the update range information holding section 22
A pair of version number information 20 with the same value before and after
Make sure to sandwich it at 21. After this, input/output control section 1
0 is an input/output device activation command (SIO) for each volume.
issue (instructions). As a result, the contents of the journal record area 7 and the contents of the data area 8 are respectively written to the corresponding areas of each volume. When the writing is completed, an I/O interrupt is sent from each volume, so the input/output control unit 10 notifies the input/output request source program 4 of the completion of writing. Then, when the data update is successfully completed on all physical volumes, the journal update display erasure processing section 11 clears the update range information holding section 22 in the journal record area 7. In other words, the "updating" display is deleted. Thereafter, the contents of this new journal record area 7 are written to each physical volume when a subsequent input/output request occurs or when an appropriate amount of time has elapsed.

When writing the contents of the journal record area 7 to the physical volume before and after updating the data of the physical volume, the update range is written between the pair of version information 20 and 21 as described above. The information 22 is written. This is to ensure that even if a failure occurs while writing the update range information 22 onto the physical volume, the failure will be known later. That is,
If a failure occurs during writing of the update range information and the writing operation is interrupted, the two version information 20 and 21
will be different values (version information 21 remains the old value), so if you compare both version information when reading them later, the update range information will be incorrect because they do not match. It turns out that there is.

Next, the operation when the system goes down while updating data is as follows.

If the system goes down while data is being updated, an "updating" message remains in the journal record on the physical volume. Therefore, after the system goes down, before using the logical volume, the synchronization recovery processing unit 12 of the system control program 6 checks the journal records on each physical volume and restores the synchronization for the part that is "updating". Process.

The recovery processing means is as shown in FIG.

First, a journal record is read from each physical volume.

Next, the latest journal for each physical volume is stored in the version number information holding unit 2 in the journal record.
Determined based on the contents of 0 and 21. In addition,
If the contents of the holding units 20 and 21 are equal, the record is considered to be correct.

After that, the journal as a logical volume is determined. This determination method differs depending on how the journal is taken.

Then, a portion that is "updating (bit of update range information holding unit 22 is on)" on the journal as a logical volume is copied to the same position (address) of another volume.

With the above steps, the equivalence restoration process is completed, and the current volume mount process is performed.

Figure 5 shows how data is copied from an arbitrary normal physical volume to another physical volume for the cylinder part corresponding to the bit position where "1" is set ("updating") in the journal record. This shows that.

Various methods can be used to determine the latest journal.

First, instead of only one journal record, a track can have a plurality of records in order to speed up processing. Furthermore, from the viewpoint of reliability, it is also possible to use a plurality of tracks cyclically, so that even if an error occurs during journal updating, the latest one remains. In this case, to determine the latest journal, it is sufficient to read all tracks containing journal records and find the record with the maximum version number.

Next, in determining the journal as a logical volume, the journal as a logical volume is determined from the latest journal of the latest volume of each physical volume.

In other words, (i) if the journal is recorded every time data is updated, the version number is the highest among the journals for each physical volume, and (ii) the updated part on the physical volume is the one that is updated the previous time. If a new journal is written when the parts differ, the logical volume's journal will be the merged result of the latest journals of each physical volume.

In the above embodiment, a cylinder is used as one update range unit, but the present invention is not limited to this, and a track or physical block, etc. can also be used as one update range unit.

(F) Effects of the Invention As explained above, according to the present invention, the range that was being updated when the system went down is clarified, and all journal records on each physical volume can be read when the system is started up or when a volume is mounted. Then, the latest journal information is obtained (if the version number is the maximum, it is the latest), and the equivalence guarantee process is performed for the range indicated on this journal, so the time required for the equivalence guarantee process is , overhead can be significantly reduced.

[Brief explanation of the drawing]

Fig. 1 is a conceptual diagram of a multiplexed volume, Fig. 2 is a block diagram of a data processing system according to an embodiment of the present invention, Fig. 3 is a diagram showing the format of a journal record in an embodiment, and Fig. 4 is an equivalence recovery. FIG. 5, which is a diagram showing the operational flow of processing, is a diagram showing a mode of data copying between volumes. In FIG. 2, 1 is a data processing device, 2, 3
is the volume, 7 is the journal record area, 8
9 is a data area, 9 is a journal update display processing section, and 11 is a journal update display erase processing section.

Claims (1)

[Claims] 1. In a data processing system that writes the same data on multiple physically different volumes and treats these multiple volumes as one logical volume, a journal record area is provided in each of the multiple physical volumes, and the above-mentioned When a write request is made to a physical volume, journal information is created in which update range information indicating a predetermined update range in which write processing is performed is sandwiched between a pair of version number information indicating the update frequency, and journal information for the plurality of physical volumes is created. The relevant journal information is written to each record area before the data of the relevant physical volume is updated.When starting up the system after a system failure, the content of the journal record area of the above-mentioned multiple physical volumes is referred to and the above-mentioned update is performed. A pair of version number information that sandwich the range information is compared, and if they are the same, the updated range information is considered correct, and the data area that was in the updating state is identified by the updated range information, and the data area that was in the updating state is identified. A multiplexed volume processing method characterized by performing equivalence recovery processing to guarantee the equivalence of data contents.