JPH01266640A - Equivalent property guarantee processing system for duplex volume - Google Patents

Abstract

PURPOSE:To curtail an exclusive overhead to an updating request by constituting the respective journals which are connected to a journal setting processing part and an updating end processing part, of an entry having information of 'the blocks which are being updated are continuous something blocks from a relative something block diagram in a duplex volume'. CONSTITUTION:An entry of journals 14X, 14Y is made usable cyclically in advance, and a size of the journals 14X, 14Y is allowed to correspond to a load of a duplex volume and determined suitably. That is, the size of the journal is determined by the number of entries, and such a problem as the size of the journal is different at every kind of a device is obviated. Therefore, the equivalent property guarantee processing system for the duplex volume is constituted of a processor 10 containing an I/O issue processing part 11 for an updating request containing a journal setting processing part 12, and an updating end processing part 13 for an I/O end interruption, a physical volume 15X and 15Y are connected to the processing part 12, and the journal 14X and 14Y are connected in advance to the processing part 12 and 13.

Description

[Detailed Description of the Invention] [Summary] A journal for managing update request execution information is provided for each volume of a duplex volume, and the equivalence of the duplex volume is guaranteed in the event of a system failure based on the information in the journal. Regarding a processing method for ensuring equivalence of duplexed volumes in a processing system.

The purpose of this invention is to provide a means to reduce the overhead for exclusive use of duplex volume update requests.

a journal having an entry having a structure in which information on the starting block position being updated and information on the number of updated blocks are stored;
When executing an update request, a journal setting processing means sets the update execution information to the latest entry of the journal, and when the update for one physical volume is completed,
and update completion processing means for emptying entries in each journal corresponding to the request, and if there is a secondary request when the update in one physical volume is completed.

Configure the primary request to execute without waiting for the other update to complete.

[Industrial application field]

The present invention provides a processing system for duplexed volumes in which a journal for managing update request execution information is provided for each volume of duplexed volumes, and the equivalence of duplexed volumes is guaranteed based on the information in the journal in the event of a system failure. Concerning equality guarantee processing method.

In computer systems, to prevent operation from stopping in the event of a storage medium failure.

A duplex volume is used in which the same information is set twice in different physical volumes. In such a duplex volume, it must be guaranteed that both physical volumes always have the same content. Therefore, journals are used to guarantee the equivalence of duplexed volumes.

Because it affects the response to access requests.

It is necessary to efficiently carry out processing related to equivalence guarantee.

[Conventional technology]

FIG. 6 shows an example of the conventional method, and FIG. 7 shows the processing flow of the conventional method.

Conventionally, journals for ensuring the equivalence of duplexed volumes have been collected as shown in Part 6.

Each physical volume 15X, 15Y is divided into a certain unit (for example, cylinder unit), and the entire area within the physical volume is divided into two, and each unit is represented by a bitmap 30X, 30Y corresponding to an I bit. Then, the bit corresponding to the unit including the area being updated is turned on, and the bits other than the unit are turned off.

That is, when executing an update request to the duplex volume, the bit of the bitmap 30X or 30Y corresponding to the unit including the updated portion is turned on. and,
When the update request is completed in both physical volumes 15X and 15Y that make up the duplex volume, the bits that were turned on are turned off.

By acquiring journals from Bitmano 130X and 30Y, for example, in the event of a failure such as a system down, when restarting the system, check the area where the bit is turned on in the focus map and update that area. Since there is a possibility that the contents of both physical volumes have not been lost, the contents of the two physical volumes are restored by copying from one to the other, thereby ensuring the equivalence of the duplicated volumes.

The processing flow according to the conventional method is shown in FIG.

If there is a request to update the duplicated polyneme, the seventh
Execute the processes ① to ② shown in Figure (a).

■ First, it is determined whether the physical volume 15X is free. If it is not empty, move on to process ■.

(2) If the physical volume 15X is free, check whether the bit corresponding to the unit including the area to be updated is on in the journal for the physical volume 15X (focus map 30X shown in FIG. 6). If it is on, a request for physical volume 15X is placed in a waiting state.

Move on to processing ■.

■ If the corresponding bit is off, turn on the corresponding bit in bitmap 30X.

(2) Then, an I10 command for updating is issued to the physical volume 15X.

■ Next, check whether the physical volume 15Y is free.

(2) If the physical volume 15Y is free, processing similar to process (2) and process (2) for the physical volume 15X is executed, and the process ends.

For example, when there is an I10 end interrupt from the physical volume 15X, processes 1 to 0 shown in FIG. 7(b) are executed.

■ Check whether the update on the physical volume 15Y side, which is the other side, has been completed. If not completed, the process moves to process 0.

■ If the update on the other side has been completed, one journal (Bitmap 30X and Bitmap 30 shown in Figure 6)
Turn off the corresponding bit of Y).

■ Check whether there is a waiting request in the physical volume 15X, and if there is, execute the secondary process 0.

[Phase] Turn on the corresponding bit of bitmap 30X.

Issue the ■10 command for updating to the physical volume 15X.

■ Check whether there is a waiting request on the physical volume 15Y, and if there is, execute the first process ■.

0 Turn on the corresponding bit of bitmap 30Y.

The 10 command for updating is issued to the physical volume 15Y.

0 If the update on the other side has not been completed yet, check whether I10 can be issued to the physical volume 15X. That is, it is checked whether there is a waiting request for the physical volume 15X.

If there is a waiting request, the same processes as shown in FIG. 7(a) are performed.

[Problem to be solved by the invention]

The above conventional method has two problems of first order (al and (bl).

(a) In the update process for a certain area, the update in one physical volume 15X is completed, and the update in the other physical volume 15Y is not completed. In this state, if a new update request comes for an area that is turned on on the bitmap, the request must wait until the update of the physical volume 15Y is completed.

This is because if Ilo is issued immediately in response to a new update request, a conflict will occur on the bitmap.

In this way, it is necessary to apply exclusion to the area to be updated,
The overhead for this exclusion becomes a problem. In particular, if there are multiple update requests for the area corresponding to the unit divided into three, the primary update request is executed under exclusive control until Ilo is completed in both physical volumes that make up the duplex volume in response to one update request. Because you can't.

There is a problem that the response as a physical volume deteriorates.

(bl Also, the size of the bitmap shown in FIG. 6 is.

Since it depends on the size of the physical volumes that make up the duplex volume, the size of the journal varies depending on the type of device, and there is a problem in that when changing devices, the processing to deal with it becomes complicated.

SUMMARY OF THE INVENTION The present invention aims to solve the above-mentioned problems by providing a means for reducing the overhead for exclusive use of duplexed volume update requests, and also by providing a journal management means independent of the type of device.

[Means to solve the problem]

FIG. 1 is a diagram explaining the principle of the present invention.

In FIG. 1, 10 is a processing unit consisting of a CPU, memory, etc., and 11 is for an update request for a duplex volume! 12 is a journal setting processing unit that collects journal information of information during update execution; 13 is an update completion processing unit that processes I10 completion interrupts from the duplex volume; 14X; 14Y represents a journal provided corresponding to each physical volume of the duplex volume, and 15X and 15Y represent physical volumes forming the duplex volume.

In the present invention, the journals 14X and 14Y for ensuring equidominance of duplexed volumes have entries with a structure in which information on the starting block position being updated and information on the number of consecutive updated blocks from there are stored. configured.

When the journal setting processing unit 12 executes a duplex volume update request in the ■10 publication processing unit 11,
The update execution information, that is, the start block position information to be updated and the number information of update blocks, is stored in the journal 14.
The process of setting the latest entry in X or journal 14Y is performed.

When the update to both physical volumes 15X and 15Y constituting the duplex volume is completed, the update completion processing unit 13 updates each journal 14X and 14 corresponding to the request.
Perform processing to empty the entry of Y.

If there is a 5th request when the update on one physical volume 15X or 15Y is completed.

by the I10 issue processing unit 11 without waiting for the update in the other physical volume to be completed.

Execute the next request.

[Effect]

Journals 14X, 14Y [The block being updated is
Since the structure is made up of entries that have information on the relative number of blocks in the duplex volume and the number of contiguous blocks, there is no need to apply exclusion to the area to be updated for each update request. The problem of control overhead is eliminated. Furthermore, since there is no need to apply exclusion, the problem of conventional responses is also solved.

Entries in journals 14X and 14Y can be used cyclically.
The size of Y can be appropriately determined depending on the load on the duplex volume. That is, since the size of the journal can be determined by the number of entries, the conventional problem of journal sizes differing depending on the type of device is solved, and the journal size can be made unique in the system.

If the response of one physical volume decreases for some reason and there are no free entries in the journal, you can make subsequent update requests wait until there is free space in the journal entry. good.

〔Example〕

FIG. 2 shows an example of the structure of a journal according to an embodiment of the present invention.
FIG. 3 shows an example of the use of a journal according to an embodiment of the present invention.
FIG. 4 is an explanatory diagram of management of request events for duplexed volumes, and FIG. 5 is an explanatory diagram of processing in an embodiment of the present invention.

In the present invention, the journal 14X corresponding to each physical volume is
, 14Y are not provided as bitmaps, but have entries having the structure shown in FIG. 2, for example. The number of entries (n) is.

It may be determined appropriately, taking into consideration the load on the duplex volume. The leading entry number in use and the unused leading entry number are stored in the leading part of the journal.

Each entry is used cyclically, and after the last nth entry is used, the first entry is used for the ninth request.

Each entry includes the number of the first block being updated.

From there, the consecutive block numbers being updated are set.

Figure 3 shows a duplex volume.

(a) An update request to update three consecutive blocks starting from the second block.

(b) An update request to update two consecutive blocks starting from the fourth block.

An example of how the journals 14X and 14Y are used in the case where the journals 14X and 14Y are used is shown.

In the initial state, the first used entry number and the first unused entry number for each journal are:

As shown in FIG. 3(a), both are "1".

When the update request (a) is executed for both physical volumes, as shown in Figure 3, the first block number being updated is "2" for the first entry of each journal. "3" is set as the number of consecutive blocks being updated.The unused leading entry number is updated to "2".

Next, when the update on the physical volume on the X side is completed, the update request (b) is executed on the physical volume on the X side even if the update on the Y side is not completed. At this time, the journal 14X is updated as shown in FIG. 3(c).

In the physical volume on the Y side, when the update request in (a) is completed, the first entry in the journals 14X and 14Y is made a free entry, and the journal 14Y is filled with the following:
Set the information to execute the update request for the physical volume on the Y side.

(mountain) update request. Journal 14 at this time
X and 14Y are as shown in FIG. 3(d).

Management of request events for duplexed volumes is performed as shown in FIG. 4, for example. In addition.

This management method is the same as before.

Corresponding to the physical volumes 15X and 15Y, there are device control tables 20X and 20Y that manage them.

When there are update request events) Wl and W2, the respective device control tables 20X and 20Y are updated as shown in FIG.
will be queued.

Now, update request event W for physical volume 15X
Assuming that Ilo 1 is completed, the update request event W1 is removed from the queue of the device control table 20X, and the device control table 20X processes the primary update request event 1-W2 as shown in FIG. point.

FIG. 5(A) shows an example of the processing of the I10 issue processing unit 11 shown in FIG. 1, and FIG. 5(B) shows an example of the processing of the update termination processing unit 13 shown in FIG.

■~@ in the following explanation refers to the process shown in FIG.
corresponds to

■ If there is an update request for a duplexed volume,
First, whether physical volume 15X is free or not, that is, physical volume 15X is! Determine whether it is in the 101st line. If it is not empty, move on to process ■.

■ If physical volume 15X is free, the first free entry in journal 14X.

Set update execution information as shown in FIG.

■ I10 for updating physical volume 15X
Issue orders.

■ Next, check whether the physical volume 15Y is free. If it is not empty, the process ends.

■ If physical volume 15Y is free, it becomes the first free entry in journal 14Y.

Set update information.

■ 110 for update for physical volume 15Y
Issue the command and end the process.

If there is an I10 end interrupt from the physical volume 15X, the ninth process ① to 0 is executed.

■ Check whether the update of the physical volume 15Y on the other side has been completed. If not completed, process
Move to.

■ If the update on the other side is completed, journal 14X
, 14Y are made empty.

■ Check whether there is a waiting request for physical volume 15X. If not, jump to the primary process [phase] and process ■.

■ In the first free entry in journal 14X,
Set update information.

■ I10 for updating physical volume 15X
Issue orders.

■ Perform other termination processing, such as notifying the update requester.

Similar processing is executed when there is an I10 end interrupt from the physical volume 15Y side.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to reduce the number of overloads for exclusive use with respect to update requests for duplexed volumes, and improve the response to requests for physical volumes. Also.

Journal size can be determined regardless of device type.

[Brief explanation of the drawing]

FIG. 1 is a diagram explaining the principle of the present invention. FIG. 2 shows an example of the structure of a journal according to an embodiment of the present invention. FIG. 3 is an example of the use of a journal according to an embodiment of the present invention. FIG. 4 is an explanatory diagram of management of request events for a duplex volume. FIG. 5 is an explanatory diagram of processing in an embodiment of the present invention. FIG. 6 is an example of a conventional method. FIG. 7 shows the processing flow of the conventional method. In the figure, 10 is a processing device, 11 is an I10 issue processing unit, 12
13 is a journal setting processing section, 13 is an update termination processing section, and 14 is a journal setting processing section.
X, 14Y are journals, 15X. 15Y represents a physical volume.

Claims (1)

[Scope of Claims] A processing system for guaranteeing the equivalence of duplexed volumes in a processing system that includes a journal for managing information during update request execution for each volume of the duplexed volume, and guarantees equivalence of the duplexed volumes by the journal in the event of a system failure. In the processing method, regarding update execution information in each volume, there is a journal (14X, 14Y) having an entry structure in which information on the starting block position being updated and information on the number of consecutive updated blocks from there are stored, and a duplex volume. a journal setting processing means (12) that sets the update execution information to the latest entry of the journal when executing an update request; update completion processing means (13) for emptying entries in each journal corresponding to the physical volume;
A dual volume equivalence guarantee processing method characterized in that if there is a next request, the next request is executed without waiting for the update in the other physical volume to be completed.