JP2001188690A - Computer system and checkpoint information preserving method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a computer system capable of coping with generation of faults and being recovered without preventing the lowering of system performance. SOLUTION: Checkpoint information of a process is managed for every unit of page, an updated page is recorded, whether or not the updated page is stored in the case of the previous collection of the checkpoint information as well is compared, when the updated page is stored in the case of the previous collection of the checkpoint information, only the updated part is transmitted from a primary computer to a backup computer and in other cases, the entire page is transmitted from the primary computer to the backup computer.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer system having a backup computer in addition to a primary computer for executing a process, and more particularly, to executing a process by a backup computer when a failure occurs in the primary computer. The present invention relates to a computer system and a checkpoint information storage method as described above.

[0002]

2. Description of the Related Art FIG. 14 illustrates the operation of a computer system of this kind. The system comprises a primary computer 1 and a backup computer 2. And the primary computer 1
The checkpoint information 4 of the process 3 operating in the step 3 is periodically collected, transferred to the backup computer 2 and stored as checkpoint information 5. In this case, if the failure causes the primary computer 1 to go down and the process 3 cannot continue, the backup computer 2 reproduces the process 6 based on the checkpoint information 5 and continues the processing.

[0003] A computer system composed of two computers of this kind, a primary computer and a backup computer, is disclosed in US Pat.
Issue 673, `` Checkpointing Computer System Having Duplic
ated Files forExecuting Process and Method for Man
aging the Duplicated Files for Restoring the Proces
s "and a paper" Distributed Replication Method Realizing Fault-Tolerant System by Combining with Distributed Checkpoint Method "(Transactions of the Institute of Electronics, Information and Communication Engineers, D-1 Vol.
J82-D-1 No3 pp.496-507 March 1999).

In these computer systems, in order to prevent the performance of the entire system from deteriorating, when transferring checkpoint information collected by a primary computer to a backup computer, it is necessary to minimize the amount of data to be transferred.

Conventionally, in order to reduce the amount of data in an address space that occupies most of the checkpoint information of a process,
Address space is managed in page units, and when a page is updated, the information of that page is recorded, and when checkpoint information is collected, only the contents of the recorded page are saved, a method called differential checkpoint. Was.

In this method, since only the contents of the page updated from the previous checkpoint are stored, it is possible to greatly reduce the data amount of the checkpoint as compared with storing the entire address space. Become.

FIG. 15 shows a conventional difference checkpoint method, and explains a page stored at a checkpoint x. That is, the information of the pages A and D updated after the checkpoint (x-1) is recorded, and the recorded pages A and D are recorded at the checkpoint x.
Indicates that you want to save.

[0008]

SUMMARY OF THE INVENTION As described in the prior art,
In the differential checkpoint method that manages the address space in page units and saves only the updated pages at the time of checkpoint, since the pages updated by the process have locality, when performing processing that reduces the number of updated pages, It works effectively to reduce the amount of data, and can greatly reduce the amount of stored data.

However, when a process executes a process that frequently updates a small amount of data of a large number of pages, as in a relational database, the advantage of storing the differences is lost, and a large number of checkpoints are lost for each checkpoint. You will save the page. As a result, the data amount becomes large, a large overhead is generated at the time of transfer from the primary computer to the backup computer, and the performance of the entire system is significantly reduced.

FIG. 16A shows a case 1 in which this method works effectively when an updated portion indicated by oblique lines is stored by the conventional differential checkpoint method, and FIG. , The case where there is almost no effect is shown as Case 2. In case 1 shown in FIG. 16A, since only page E is updated in a wide range, only page E needs to be stored when checkpoint information is collected. On the other hand, case 2 shown in FIG.
, Pages A, B, C, D, E, and F must be collected when checkpoint information is collected because each of the narrow ranges of pages A, B, C, D, E, and F is updated.

An object of the present invention is to reduce the amount of data to be stored even in a process for frequently updating small data of a large number of pages, such as a relational database, and to reduce the amount of data to be stored from a primary computer to a backup computer. It is an object of the present invention to provide a computer system and a checkpoint information storage method capable of preventing a decrease in system performance by suppressing a transfer amount to a checkpoint information.

[0012]

According to a first aspect of the present invention, there is provided a computer system comprising a primary computer and a backup computer, wherein checkpoint information of a process executed on the primary computer is stored in the backup computer. In a computer system that executes the process on the backup computer based on the checkpoint information when a failure occurs on the primary computer, the primary computer collects the checkpoint information of the process, Checkpoint information management means for managing information in page units, updated page recording means for recording updated pages, and updated pages are stored even when previous checkpoint information was collected. An updated page comparing means for comparing whether or not an updated page has been saved at the time of previous checkpoint information collection, and save only the updated portion, otherwise save the entire page It is characterized by comprising an information storage means and means for transmitting the checkpoint information stored in the checkpoint information storage means to the backup computer.

According to a second aspect of the present invention, there is provided the backup computer according to the first aspect, wherein the page information stored in the checkpoint storage means is information of the entire page or the page stored at the time of previous checkpoint information collection. Page information discriminating means for discriminating whether or not the information is difference information, and when the saved page information is difference information from the page information saved when the previous checkpoint information was collected, the page information is saved when the previous checkpoint information was collected Page information restoring means for restoring information of the entire page by merging with the page information.

According to a third aspect of the present invention, in the first or second aspect, the primary computer divides a page stored in the checkpoint storage unit into one or more blocks and manages the intra-page block management unit; And an intra-page block storing means for storing one or more blocks including the updated part as update data of the page.

According to a fourth aspect of the present invention, in a computer system including a primary computer and a backup computer, a checkpoint information storing method for causing a backup computer to take over and execute a process when a failure occurs in the primary computer. Managing the checkpoint information of the process on a page-by-page basis, recording the updated page, and comparing whether the updated page is saved at the time of the previous checkpoint information collection. If the updated page is stored at the time of previous checkpoint information collection, only the updated portion is transmitted to the backup computer, otherwise, the entire page is transmitted to the backup computer.

According to the first and fourth aspects of the present invention, the primary computer records the page updated from the previous checkpoint information collection time as managing the checkpoint information of the process in page units. Then, compare whether the pages updated since the previous checkpoint information collection were saved when the previous checkpoint information was collected, and compare the pages updated since the previous checkpoint information collection with the previous checkpoint information collection. Sometimes only save the updated part if it has been saved, otherwise save the entire page. This reduces the amount of checkpoint information to be saved when a failure occurs on the primary computer,
The amount of data from the primary computer to the backup computer is reduced.

Therefore, even in a process of frequently updating small data of a large number of pages, such as a relational database, the amount of data to be saved is small, and the data is transferred from the primary computer to the backup computer. By suppressing the amount, it is possible to prevent a decrease in system performance.

According to the second aspect of the present invention, based on the checkpoint information in which the transfer amount from the primary computer to the backup computer is suppressed, whether the page information is the information of the entire page or saved at the time of the previous checkpoint information collection Page information saved at the time of previous checkpoint information collection if the saved page information is difference information from the page information saved at the time of the previous checkpoint information collection By restoring the information of the entire page by merging with the backup computer, even when a failure occurs in the primary computer, the process can be reproduced by the backup computer and the processing can be continued.

According to the third aspect of the present invention, the primary computer performs block management, determines a portion updated for each block, and stores the updated data. The amount of stored data can be reduced.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of a computer system according to the present invention will be described below with reference to the drawings. FIG.
1 shows a computer system of the present embodiment, in which a primary computer 10 and a backup computer 20 are connected to a cable 30. Usually,
The process 40 is processed by the primary computer 10. The checkpoint information of the process 40 executed by the primary computer 10 is appropriately transferred to the backup computer 20. When a failure occurs in the primary computer 10, the process 40 'is executed by the backup computer 20 based on the checkpoint information.

The primary computer 10 includes a checkpoint information management unit 11, an updated page comparison unit 12,
A checkpoint information storage unit 13 and an updated page recording unit 14 are provided, and the presence or absence of an update is compared and determined on a page basis or on a block basis within the page related to the checkpoint information,
The data to be stored is determined based on the determination.

Further, the backup computer 20
A page information determining unit 21 and a page information restoring unit 22;
The process 40 'is reproduced based on the checkpoint information stored and transferred to the primary computer 10, and the process is continued.

Each part of the primary computer 10 will be described in detail with reference to FIG. That is, the checkpoint information management unit 11 collects information for each page or for each block in a page at each checkpoint of the process 40, and manages the stored checkpoint information. It manages checkpoint information related to a page to be compared and stored or a block in a page, which is required in the comparison unit 12 and the checkpoint storage unit 13. Update page comparison part 1
2 indicates that the pages updated after the checkpoint (x-1) based on the checkpoint information from the checkpoint information management unit 11 and the update page recording unit 14 are updated to the checkpoint (x-1) (or earlier, that is, x-
2, x-3,...) Are compared. The checkpoint information storage unit 13 determines that the page compared by the updated page comparison unit 12 is the checkpoint (x-1) (or earlier, that is, x-2, x
-3,...), The difference information from the information is stored. Otherwise, the entire page is stored. The updated page recording unit 14 records a page changed after the check point (x-1).

Hereinafter, a flow of creating page information to be stored at the checkpoint x will be described in chronological order. still,
In the following description, pages updated before checkpoint (x-1) will be appended with (x-1) at the end of the notation,
Pages updated after the checkpoint (x-1) are marked with x at the end of the description.

In FIG. 2, initially, at checkpoint (x-1), data of pages A, B, E, and G are converted into pages A (x-1), B (x-1), and E (x-1). , G (x−
1) and save them as checkpoint management unit 1
1 to manage. Next, checkpoint (x-1)
Thereafter, for example, the data of the pages B, C, and F updated to the checkpoint x are stored as pages Bx, Cx, and Fx, and are recorded by the checkpoint updated page recording unit 14.

The pages A (x-1), B (x-1), E stored by the checkpoint update page recording unit 14
(X-1), G (x-1) and pages Bx, Cx, Fx
, The updated page comparison unit 12 compares the page information stored at the checkpoint (x-1) with the page information stored at the checkpoint x, which is a subsequent process.

Next, for the page B stored at the checkpoint (x-1) by the checkpoint storage unit 13, only the difference from the page B (x-1) is stored as the page Bx. Pages C and F that were not saved at checkpoint (x-1) are replaced by pages Cx and Fx.
Save as

Next, the previously stored data is managed by the checkpoint information management unit 11.

According to the above processing, the checkpoint information to be saved when a failure occurs in the primary computer 10 can be updated part or all of the pages, and the information as a whole becomes small, and the primary computer 10 From the backup computer 20 to the backup computer 20 can be reduced.

Therefore, even in a process such as a relational database that frequently updates a small amount of data of a large number of pages, the amount of data to be stored in the primary computer 10 is small, and the primary computer 10 By reducing the amount of data transferred to the backup computer 20 from
It is possible to prevent a decrease in system performance.

The collection and storage of the checkpoint information described above. More specific examples will be described with reference to FIGS. FIG. 3 is a diagram showing an example of collection and storage of checkpoint information managed by the checkpoint information management unit 11 according to claim 3, and the flow of collection is shown below. The pages protected by write protection are indicated by diagonal lines. FIG. 3 is common to both the methods shown in FIGS. 4 and 5 in that the presence / absence of update is determined by undoing write protection for each page. 4 compares the page difference between the current checkpoint and the previous checkpoint, while FIG. 5 illustrates the page difference between the current checkpoint and the checkpoint two or more times before. Are different from each other.

In FIG. 3, a checkpoint module 111 for storing page data when checkpoint information is collected, a handle 112 for recording a page in which a page fault has occurred and removing write protection for the page, and an updated page An update management tag 113 for recording information and a storage format recording tag 114 for setting "1" when saving the entire page at the time of a checkpoint and setting "0" when saving only the difference are included.

At the time of checkpoint information collection, the following processing is performed.

As shown in FIG. 3A, each page is write-protected. This is a premise for managing the process space information in page units. And a handler 1 for performing processing when a page fault occurs.
Checkpoint module 111 for setting 12 and saving the updated page when checkpoint information is collected
Create Also, an update management tag 113 for recording that the page has been updated is added to each page. As a result of these series of measures, when a page is updated, a fault has occurred in the page.

Here, if the page B is updated as an example, a page fault occurs in the page B, and the control is transferred to the handler 112. FIG. 3 (b)
As shown in the figure, the handler 112 updates the page B
Is recorded in the update management tag 113, and the write protection of the page is released.

Next, as shown in FIG. 3C, when the first checkpoint information is collected, the checkpoint module 111 stores the updated data of the entire page B in the page B.
1, and “1” is set in the storage format recording tag 114. Further, the write protection of page B is re-applied.

Next, a method for comparing the difference between the current time and the previous time for each page will be described with reference to FIG. As shown in FIG. 4, at the time of the x-th checkpoint information collection,
The pages B and D updated after the checkpoint (x-1) are compared with the page data managed by the checkpoint module 111. The checkpoint module 111 sets the data of the entire page B as the page B1 when the first checkpoint CP1 is collected.
When the (x−1) -th checkpoint CP (x−1) is collected, the data of the entire pages A and D are respectively transferred to the page A (x−
1) and D (x-1).

In this case, since page B is not stored at checkpoint CP (x-1), page D is stored at checkpoint CP (x-1) with the data of the entire page as page Bx. Therefore, page D (x
Only the difference from -1) is stored as page Dx.

At this time, the storage format recording tag 1 of the page Bx
14 is set to “1”, the storage format recording tag 114 of the page Dx is set to “0”, and the page Dx is further provided with an offset within the page of the difference data necessary for the later-described restoration. And a difference data size.

Next, a method for comparing the difference between the current page and the page twice or more for each page will be described with reference to FIGS. This example corresponds to claim 5 and is another example of collection and storage of checkpoint information managed by the checkpoint information management unit 11. Since the functions of the illustrated modules and the flow of FIG. 3 are the same as those in the previous example shown in FIG. 4, here, only the x-th checkpoint processing different from the previous example will be described.

As shown in FIG. 5, when the x-th checkpoint information is collected, the pages B and D updated after the checkpoint (x-1) and the page data managed by the checkpoint module 111 are stored. Compare.

The checkpoint module 111 has 1
The data of the entire page B is stored as the page B1 when the checkpoint CP1 is collected for the first time, and the data of the entire pages A and D are respectively stored on the page A when the checkpoint CP (x-1) is collected for the (x-1) th time. (X-1), D (x
-1).

Since page B is stored at checkpoint CP1, only the difference from page B1 is stored in page Bx
And save page D as checkpoint CP
Since it is stored as (x-1), page D (x-1)
Is stored as page Dx.

In both the pages Bx and Dx, the storage format recording tag is set to "0", and the offset in the page and the difference data size of the difference data required at the time of restoration are added as in the example of FIG. .

The collecting and preserving method shown in FIG.
The difference from the collection and storage method shown in Figure 2 is that the former compares the difference between the current checkpoint page and the previous checkpoint page, while the latter differs from the current checkpoint page two or more times. The difference is that the method is to compare the difference with the page at the previous checkpoint, and both methods are common except for the checkpoint to be compared.

In the above-described embodiment, a part or all of the pages to be stored are determined by the difference comparison of each page. However, the page is divided into a plurality of blocks, and the pages to be stored are compared by the difference comparison of each block. A method of determining a part or all of the blocks can be adopted.

Such an embodiment will be described with reference to FIGS. 6 and 7, in which the same parts as those in FIGS. That is, the primary computer 10
5 is provided with an in-page block management unit 31 and an in-page block storage unit 32 corresponding to the checkpoint information management unit 11 and the checkpoint information storage unit 13 in FIG.

In FIG. 2, FIG. 6, and FIG. 7, when the x-th checkpoint information is collected, the checkpoint (x-
1) It is checked by the checkpoint information management unit 11 and the in-page block management unit 31 whether or not the page D updated thereafter is stored at the checkpoint up to that time. Next, since the data of the page D is also stored as the page D (x-1) at the checkpoint (x-1), the page D and the page D (x-1) are prepared in order to prepare the difference between the page D and the page D (x-1). Divide page D (x-1) into N equally sized blocks each.

Next, as shown in FIG. 7, page D (x-
1) and the corresponding blocks of the page D are compared with each other, and the blocks a and b (the hatched portions in FIG. 6) that do not coincide with each other are set as the page Dx, and the in-page block storage unit 32
To save. At that time, each block number is added, and “0” is set to the storage format recording tag.

As described above, the page is divided into a plurality of blocks instead of the page unit, and a part or all of the blocks in the page to be stored can be determined by the difference comparison for each block. It is effective when it is limited to a part of.

The description of the above embodiment shows the storage form of the checkpoint information in the primary computer 10, and the data relating to the stored checkpoint information is transmitted to the backup computer 20 via the cable 30 in FIG. Where it is restored.

An embodiment for performing such restoration is shown in FIGS.
The same parts as those in FIG. 7 will be described with reference to FIGS. 8 to 10 correspond to claim 2,
FIG. 9 is a diagram illustrating a method for restoring difference page data. The function of each illustrated module is the same as in the previous embodiment. The restoring method shown in FIG. 9 shows the restoring method of the data saved by the saving method shown in FIGS. 4 and 5, and the restoring method shown in FIG. 10 shows the restoring method of the data saved by the saving method shown in FIG. FIG. 8 shows a restoring method, and the restoring method shown in FIG. 8 shows a restoring method of data stored by the saving method shown in FIGS.

In FIG. 8, each of the checkpoint information transmitted from the primary computer 10 is discriminated by the page information discriminating section 21 and the page information restoring section 22 to restore a part of the page information. The information of each page is restored.

In FIG. 9, "0" is set in the storage format recording tag of page Dx, which is the data of page D stored at checkpoint x, and page D stored at checkpoint (x-1). Page D which is the data of
Since “1” is set in the storage format recording tag of (x−1), it is determined that page Dx is difference data of page D (x−1).

Next, the page D (x-1) is checked by overwriting and merging the difference data of the size designated by the difference data size from the position designated by the intra-page offset of the page Dx. The data of page D used when rolling back at point x is restored.

In FIG. 10, "0" is set in the storage format recording tag of page Dx, which is the data of page D stored at checkpoint x, and the data of page D stored at checkpoint (x-1) is stored. Page D which is data
Since “1” is set in the storage format tag of (x−1), it is determined that the page Dx is the difference data of the page D (x−1).

Next, the page D (x-1) is used when rolling back at checkpoint x by overwriting the data of blocks a and b contained in page Dx with the corresponding blocks and merging them. The page D data to be restored is restored.

A flowchart of the checkpoint information saving and restoring method in each of the above embodiments will be described with reference to FIGS.

Step S1 Write protection is set for a page in the process space.

Step S2: Set a handler for processing a page fault that occurs when data is written to the page for which write protection has been set in step S1.

Step S3 A checkpoint module for storing the page updated and recorded by the handler set in step S2 is created.

Step S41 The first to (x-1) th checkpoint information collection ends.

Step S5 The page P is written.

Step S6 The handler set in step S2 is started.

Step S7: The handler records that the page P has been updated, and
Cancel write protection on.

The x-th checkpoint process is started in step S8 shown in FIG.

Step S9 The checkpoint module set in step S3 checks pages updated since the (x-1) th checkpoint.

Step S10 In step S10, either the comparison of the current and previous checkpoint information shown in FIG. 4 or the comparison of the current and the checkpoint information more than twice before shown in FIG.
11, and S12.

Step S11 When the method of FIG. 4 is selected, the page of step S9 is compared with the page stored at the (x-1) -th checkpoint.

Step S12 When the method of FIG. 5 is selected, the page of step S9 is compared with the page saved at the (x-1) th or earlier checkpoint.

Step S13 In step S13, one of the data storage methods shown in FIGS. 4 and 5 or the data storage method shown in FIG.
15, S16.

Step S14 In step S11 or S12, if the same page as in step S9 is not stored at the previous checkpoint, the entire page P is stored.

Step S15 When the same page as in step S9 is stored at the previous checkpoint in step S11 or S12, only the difference from the page data is stored.

In step S16, when the same page as in step S9 is stored in the previous checkpoint in step S11 or step S12, the page P is divided into a plurality of blocks, Save only updated blocks.

Step S17: The update record of the page P is cleared, and the write protection of the saved page P is reset.

Step S1 shown in FIG.
8. The x-th checkpoint process ends.

Step S19 Rollback occurs.

Step S20: It is checked whether the page P stored at the checkpoint x stores the entire page or only the difference from the previously stored page.

Step S21 If the entire page P is stored in step S20, the data is used as it is for rollback.

Step S22 In step S22, the merging method shown in FIG. 9 or FIG. 10 can be selected. That is, when only the difference from the previously stored page data is stored in step S20, the data of the previously stored page P is overwritten with the data stored in the difference and merged.

In step S23, if only the difference from the previously stored page data is stored in step S20, the block of the previously stored page P is overwritten with the data of the previously stored page P and merged.

Step S25 The data merged in step S24 or S23 is used for rollback.

[0083]

As described above, in the present invention,
Process checkpoint information is managed on a page-by-page basis, updated pages are recorded, and whether the updated pages have been saved at the time of previous checkpoint information collection is compared to determine whether the updated pages have been updated. By sending only the updated part if it was saved at the time of checkpoint information collection, otherwise the entire page from the primary computer to the backup computer, so that a small amount of many pages, such as a relational database, The amount of data to be stored is small even in the process of executing the process that frequently updates the data of the system, and it is possible to prevent the system performance from being lowered by suppressing the transfer amount from the primary computer to the backup computer. Become.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of a computer system according to the present invention.

FIG. 2 is an exemplary view showing a checkpoint information storage method in the computer system shown in FIG. 1;

FIG. 3 is a view for explaining preparation for difference check points, recording of an updated page, and first check point processing in FIG. 2;

FIG. 4 is a view for explaining an example of an x-th checkpoint process in FIG. 2;

FIG. 5 is a view for explaining another example of the x-th checkpoint process in FIG. 2;

FIG. 6 is a view for explaining a checkpoint information storage method for each block according to the present invention.

FIG. 7 is a diagram showing a method for storing difference data in an x-th checkpoint process according to the present invention.

FIG. 8 is a view showing a backup computer according to the present invention, and showing a checkpoint information restoring method in the backup computer.

FIG. 9 is an exemplary view for explaining an example of restoring the saved difference page data in FIG. 8;

FIG. 10 is a view for explaining another example of restoring the saved difference page data in FIG. 8;

FIG. 11 is a flowchart illustrating a method for saving and restoring checkpoint information according to the present invention.

FIG. 12 is a flowchart illustrating a method for saving and restoring checkpoint information according to the present invention.

FIG. 13 is a flowchart illustrating a method of saving and restoring checkpoint information according to the present invention.

FIG. 14 is a block diagram showing a conventional computer system having a primary computer and a backup computer.

FIG. 15 is a view for explaining a difference storage checkpoint method according to the related art.

FIG. 16 is a diagram for explaining a problem of the difference storage checkpoint method in the related art.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Primary computer, 20 ... Backup computer, 30 ... Cable, 11 ... Checkpoint information management part, 12 ... Update page comparison part, 13 ... Checkpoint information storage part, 14 ... Update page recording part, 21 ...
Page information restoring unit, 22: Page information discriminating unit, 31: In-page block managing unit, 32: In-page block storing unit.

Claims (4)

[Claims] 1. A backup system comprising a primary computer and a backup computer, wherein checkpoint information of a process executed on the primary computer is transferred to the backup computer, and when a failure occurs in the primary computer, the checkpoint information is transmitted. A computer system that executes the process on the backup computer based on
A checkpoint information management unit that collects checkpoint information of the process and manages the checkpoint information on a page basis; an updated page recording unit that records an updated page; An updated page comparing means for comparing whether or not the updated page is also stored at the time of collection of the checkpoint information, and, when the updated page is stored at the time of the collection of the previous checkpoint information, storing only the updated portion;
Otherwise, a computer system comprising: checkpoint information storage means for storing the entire page; and means for transmitting the checkpoint information stored in the checkpoint information storage means to the backup computer. 2. The backup computer determines whether the page information stored in the checkpoint storage means is information of the entire page or difference information from the page information stored when the previous checkpoint information was collected. Page information discriminating means, and when the stored page information is difference information from the page information stored at the time of the previous checkpoint information collection, merging with the page information stored at the time of the previous checkpoint information collection to merge the entire page 2. The computer system according to claim 1, further comprising a page information restoring unit for restoring information. 3. The in-page block management unit that divides a page stored in the checkpoint storage unit into one or more blocks and manages the page, and one or more blocks including an updated part. And an in-page block storage unit for storing the data as update data of the page.
Or the computer system according to 2. 4. A checkpoint information storing method for causing a backup computer to take over and execute a process when a failure occurs in the primary computer in a computer system including a primary computer and a backup computer, wherein Manages the checkpoint information of each page, records the updated page, compares whether the updated page is saved even when the previous checkpoint information was collected, and compares the updated page with the previous checkpoint. A method of storing checkpoint information in a computer system, comprising: transmitting only an updated portion if point information is stored at the time of collection, and transmitting an entire page to the backup computer otherwise.