JP2000047891A - Device and method for controlling input/output for computer system and storage medium storing the programmed method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily realize the capacity efficiency improvement of a secondary storage device. SOLUTION: This device is provided with an address conversion map 111 assigning at least one logical lock to the logic block of a disk device 12 and assigning a pointer to a physical block with a logic block image at the time point of the newest checking point and a pointer to a physical block with a newest block image to each block, and an ineffective block managing module 113 managing a physical block without effective data on the disk device. Then, in a check point processing, the physical block with the newest logic block image in each logic block on the conversion map is set to be a physical block with a check point image with respect to the logical block and a newest image, and the physical block with the logic block image at the time point of a checking point in each logic block on the conversion map at the time of recovery is set to be a physical block with a check point image with respect to the logic block and the newest image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an input / output control method in a computer system, particularly an input / output control apparatus suitable for use in a highly reliable computer system, an input / output control method in the system, and a method in which the method is programmed and stored. The present invention relates to a computer-readable storage medium.

[0002]

2. Description of the Related Art In a high-reliability computer system, processing is performed while periodically collecting checkpoints, and when a failure occurs in the system, the state of the system is returned to the immediately preceding checkpoint (roll). back),
A device having a checkpoint / rollback mechanism that avoids a temporary failure of the system by re-executing the process has been considered. In a computer system that incorporates this checkpoint / rollback mechanism and stores and reads data to and from a secondary storage device, even after a rollback due to a failure, the main storage and the processor, etc. It is necessary to ensure consistency between the state and the input / output device such as the contents of the disk device.

For this reason, in the normal checkpoint / rollback method, the operating system (O
The input / output request requested by S) is issued to the disk device after the issuance is determined by the next checkpoint.

[0004]

However, in the input / output control method described above, issuance of at least a request to change (write) the state of the input / output device is delayed after the next checkpoint collection process. Had become. Furthermore, since requests delayed until the time of the checkpoint are issued collectively after the next checkpoint processing, this is one cause of deteriorating the performance of the computer system that achieves high reliability by the checkpoint rollback method. .

In order to solve this problem, two physical blocks are always assigned to one logical block on the secondary storage device. A method of performing a write process newly generated in a physical block has been considered. According to this method, the write processing to the secondary storage device can be performed without waiting for the checkpoint. However, in this case, a data area of 50% of the actual physical disk capacity cannot be used as an effective disk capacity, and there is a problem that the capacity efficiency of the physical disk is poor.

On the other hand, when a write request is issued to the secondary storage device, data to be written to the secondary storage device is temporarily stored in a non-volatile memory, and after the next checkpoint, the data in the non-volatile memory is rewritten. A method of writing data to a secondary storage device and performing a write process to the secondary storage device without waiting for a checkpoint is also disclosed in Japanese Patent Application No. 7-1 701.
No. 51737. However, in order to realize this, there is a problem that it is not easy to realize because special hardware is required in the peripheral part including the nonvolatile memory.

Accordingly, the present invention has been made in view of the above circumstances, and has been made in a computer system which is easy to realize, has small performance degradation, and can improve the capacity efficiency of a secondary storage device and which solves the above-mentioned problems. I / O control unit,
It is another object of the present invention to provide an input / output control method in the system and a computer-readable storage medium on which the method is programmed and recorded.

[0008]

According to an input / output control method in a computer system of the present invention, a process is performed while periodically collecting checkpoints. Checkpoints that return the system status to
In a computer system having a rollback mechanism and an input / output module for writing and reading data to and from a secondary storage device, at least one logical block constituting the secondary storage device is provided. An address translation map for allocating physical blocks and assigning a pointer to a physical block having a logical block image at the latest checkpoint, a pointer to a physical block having a latest logical block image to each logical block, and a secondary storage device When the I / O module receives a write request to the next storage device, the latest checkpoint image is assigned to the logical block when the I / O module receives a write request to the next storage device. For different physical blocks Data, and registers this physical block as the latest logical block image in the address conversion map. When the input / output module receives a read request for the secondary storage device, it has the latest logical block image for that logical block. Data is read from the physical block, and in the checkpoint processing, for each logical block on the address translation map, the physical block having the latest logical block image is replaced with the physical block having the checkpoint image and the latest image for this logical block. At the time of recovery, for each logical block on the address translation map, the physical block having the logical block image at the checkpoint is replaced with the checkpoint image for this logical block and the latest image. Characterized in that the physical block with.

The address conversion map further records the time of the last write to each logical block, and when the input / output module receives a write request to the secondary storage device, the logical block is written to the logical block. On the other hand, data is written to a physical block different from the latest checkpoint image, this physical block is registered as the latest logical block image, and the time at which the write processing was performed on this physical block is registered in the address translation map. When the output module receives a read request from the secondary storage device, data is read from the physical block having the latest logical block image for the logical block, and the current time is stored as a checkpoint time stamp in the checkpoint processing. ,recovery If a pointer to the latest physical block with a time stamp newer than the checkpoint time stamp is registered for each logical block on the address translation map, the latest physical block is sent to the invalid block management module. It is also characterized in that it is passed and released, and the pointer to the physical block at the time of the checkpoint is registered in the address translation map as a pointer to the latest physical block.

Further, for a logical block that has been subjected to a write process after the latest checkpoint, a difference recording mechanism for recording a set of the logical block and a physical block assigned to the logical block before the write process is performed is provided. When the input / output module receives a write request to the secondary storage device, the mapping of the logical block and the physical block on the address translation map at this time is recorded in the difference recording mechanism, and the invalid block management module The newly allocated physical block is registered in the address conversion map as a physical block for this logical block, a write process is performed on this physical block, and in the checkpoint process, all the physical blocks recorded in the difference recording mechanism are checked. Solution to invalid block management module Then, the registration of all pairs of logical blocks and physical blocks recorded in the above-mentioned difference recording mechanism is deleted, and at the time of recovery, all the logical blocks recorded in the above-mentioned difference recording mechanism are subjected to the logical conversion in the address conversion map. Pass the physical block mapped to the block to the invalid block management module to release it, and write back the set of the logical block and the physical block registered in the difference recording mechanism to the address translation map, so that the latest checkpoint is obtained. It is also characterized by restoring the address conversion map at the time.

When the input / output module receives a write request to the secondary storage device, if the specified logical block is not registered in the differential recording mechanism, the addresses of the logical block and the physical block at this time are After recording the mapping on the translation map in the difference recording mechanism, registering the newly secured physical block from the invalid block management module as a physical block for this logical block in the address translation map, and performing a write process on this physical block If the specified logical block is registered in the difference recording mechanism, data is written to the physical block mapped to this logical block in the address conversion map, and in the checkpoint processing, all data recorded in the difference recording mechanism are written. The physical block is converted to the invalid block management module. Release, delete the registration of all the pairs of logical blocks and physical blocks recorded in the differential recording mechanism, and at the time of recovery, for all the logical blocks recorded in the differential recording mechanism, Pass the physical block mapped to the block to the invalid block management module to release it, and write back the set of logical block and physical block registered in the difference recording mechanism to the address translation map, so that the latest checkpoint It is also characterized by restoring the contents of the address conversion map.

According to the above configuration, in a checkpoint / rollback computer system, input / output to / from the secondary storage device can be issued without waiting for a checkpoint, thereby improving the performance of the system. Furthermore, since the number of redundant physical blocks on the secondary storage device can be reduced, the capacity efficiency of the secondary storage device can be improved.

[0013]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) FIG. 1 is a block diagram showing a schematic configuration of a high-reliability computer system according to a first embodiment of the present invention. In the figure, reference numeral 11 denotes a computer main body, which is related to the present invention.
Then, the invalid block management module 113 is allocated and stored in the memory. 12 is a large-capacity disk device serving as a secondary storage device, and 13 is a nonvolatile memory (NVRAM)
It is.

Usually, the disk input / output module 112
Receives a disk I / O request from the file system and issues an I / O request to the disk device 12. According to the present embodiment, the disk input / output module 112 manages the logical address and the physical address of the data block of the disk device 12.

FIG. 2 shows the structure of the address conversion map 111. The address conversion map 111 is stored in the disk device 12
Has an entry of two physical block addresses of valid and old valid for all logical blocks.

Here, the effective physical block is a physical block having the latest disk image of the logical block. The old valid physical block is the disk image at the latest checkpoint of this logical block, and another physical block is assigned to this logical block because a write process has occurred on the corresponding logical block after the latest checkpoint. Is a physical block that is allocated as an effective physical block to

One valid physical block exists for all logical blocks, and the address of this physical block is stored in the valid physical block address entry of the address translation map 111. For a logical block having an old valid physical block, the address of the old valid physical block is stored in the old valid physical address entry. For a logical block in which an old valid physical block does not exist, NULL is recorded in the old valid physical block entry.

FIG. 3 shows an invalid block management module 11.
3 shows the internal structure. Invalid block management module 11
3 manages a free list of physical blocks. That is,
This module manages physical blocks (invalid block address lists B, C,..., Etc.) that are not valid physical blocks of any logical block or old valid physical blocks among all physical blocks on the disk device 12.

The operation of the present embodiment will be described below with reference to the flowchart shown in FIG. 4 showing the operation when the disk input / output module 112 receives a write request.

When the disk input / output module 112 receives a write request, the disk input / output module 1
12 is the address conversion map 111 in step S41.
Is used to check whether an old valid address exists for the logical address. Address conversion map 111
If the old valid physical block address is registered for the specified logical block (step S4
1; Y), write processing is performed on a physical block registered as a currently valid physical block (step S4)
2). On the other hand, in the address conversion map 111, the old valid logical block is NULL for the specified logical block.
(N in step S41), the address of the currently valid physical block is registered as the old valid physical block address for the specified logical block (step S43), and the invalid block management module 113 A new physical block is secured, and this physical block is registered in the address translation map 111 as an effective physical block of the specified logical block (step S44).
Then, a write process (step S42) is performed on this physical block.

In the checkpoint process, all the old valid physical blocks registered in the address translation map 111 are released by passing them to the invalid block management module 113, and the values of all old valid physical block entries are set to NULL. I do. In the recovery processing, in the address conversion map 111, the valid physical block is released by passing the valid physical block to the invalid block management module 113 with respect to all the logical blocks in which the value of the old valid physical address entry is not NULL. The value of the block address entry is moved to the valid physical block address entry, and the value of the old valid physical block address entry is set to NULL. Furthermore, at the time of system shutdown,
The address conversion map 111 is stored in a nonvolatile memory (NVRA).
M) 13, the address conversion map 111 is restored to the state at the time of the shutdown completion at the next boot. (Second Embodiment) FIG. 5 is a block diagram showing a schematic configuration of a highly reliable computer system according to a second embodiment of the present invention. In the figure, reference numeral 21 denotes a computer main body, which has an address conversion map 211, a disk input / output module 212 (disk I / O module), an invalid block management module 213, and a time stamp function in relation to the present embodiment. A time stamp 214 is allocated and stored in the memory. 22 is a large-capacity disk device serving as a secondary storage device, and 23 is a nonvolatile memory (NVRAM)
It is.

In the embodiment shown here, the disk input / output module 212 manages the logical address and physical address of the data block of the disk device 22, and the time at which the write processing was performed on the physical block.

FIG. 6 is a diagram showing an example of the address conversion map 211 used in the disk input / output module 212 of this embodiment. The address translation map 211 has entries of two physical block addresses, valid and old valid, for all the logical blocks of the disk device 22, and further writes the physical block address entry which is not NULL to the physical block at the end. Has a time stamp 215 indicating the time at which the operation was performed.

Here, the effective physical block is a physical block having the latest disk image of the logical block. The old valid physical block is a logical block that has a disk image at the latest checkpoint of this logical block and another physical block has been written to the corresponding logical block since the latest checkpoint. Is a physical block that is allocated as an effective physical block to

One valid physical block exists for all logical blocks, and the address of this physical block is stored in the valid physical block address entry of the address conversion map 211. For a logical block having an old valid physical block, the address of the old valid physical block is stored in the old valid physical address entry. For a logical block in which an old valid physical block does not exist, NULL is recorded in the old valid physical block entry.

FIG. 7 shows an invalid block management module 213.
The structure of is shown. The invalid block management module 213 includes:
Like the invalid block management module 113 shown in FIG. 1, it manages a free list of physical blocks. In other words, the module manages physical blocks that are neither valid physical blocks of any logical block nor old valid physical blocks among all physical blocks on the disk device 22. Here, further, the invalid block management module 213 periodically scans the address conversion map 211, and when the time stamp 215 of the valid physical block address entry is older than the time stamp 214 at the latest check point (check point time stamp) 214. Releases the old valid physical block allocated to the same logical block as the valid physical block, and sets the value of the old valid physical block entry to NULL.

The operation of this embodiment will be described below with reference to the flowchart of the operation when the disk input / output module 212 shown in FIG. 8 receives a write request.

When the disk input / output module 212 receives a write request, the disk input / output module 212 first checks whether or not the valid block time stamp is older than the latest check point time stamp 214 (step S81). ). If the time stamp 215 of the physical block valid for the specified logical block is not older than the checkpoint time stamp 214 in the address conversion map 211 (N in step S81), the process proceeds to step S82, and the process proceeds to step S82. A write process is performed on a physical block registered as a valid physical block. Address conversion map 211
If the time stamp 215 of the physical block valid for the specified logical block is older than the checkpoint time stamp 214 (Y in S81), it is further determined in step S83 whether an old valid address exists. To check. If the old valid physical block address of the specified logical block is not NULL (step S
83, the old valid physical block is released by passing it to the invalid block management module 213 (step S8).
5) With respect to the specified logical block, the address of the currently valid physical block is registered as the old valid physical block address (step S84), and a new physical block is secured from the invalid block management module 213. The physical block is registered in the address conversion map 211 as a valid physical block of the specified logical block (step S86). Then, a write process is performed on this physical block (step S82).

In the checkpoint process, the time of the latest checkpoint is registered as a checkpoint time stamp 214. In the recovery process, for all logical blocks on the address translation map 211, if the time stamp 215 of the valid physical block corresponding to the logical block is newer than the checkpoint time stamp 214, the valid physical block is invalidated. The value is released by passing it to the block management module 213, the value of the old valid physical block address entry is moved to the valid physical block address, and the value of the old valid physical block address entry is set to NULL. At the time of system shutdown, the address conversion map 211 is stored in the nonvolatile memory (NVRAM) 23, so that the address conversion map 211 is restored to the state at the time of the shutdown completion at the next boot. (Third Embodiment) FIG. 9 is a block diagram showing a schematic configuration of a highly reliable computer system according to a third embodiment of the present invention. In the figure, reference numeral 31 denotes a computer main body, which is related to the present embodiment.
12, the invalid block management module 313, and the difference recording mechanism 315 are allocated and stored in the memory. Reference numeral 32 denotes a large-capacity disk device serving as a secondary storage device, and reference numeral 33 denotes a nonvolatile memory (NVRAM).

In the present embodiment, for a logical block newly written after the latest checkpoint, a physical block having a logical block image at the time of the checkpoint is managed by the difference recording mechanism 315.

FIG. 10 shows the disk input / output module 31.
2 shows a difference recording mechanism 315 used in FIG. Difference recording mechanism 3
Reference numeral 15 records a set of a logical block address and an address of a physical block assigned at the time of the check point, with respect to the logical block of the disk device 32 which has newly performed write processing after the latest check point.

FIG. 11 shows an invalid block management module 3
13 is a diagram illustrating a structure of a thirteenth embodiment. The invalid block management module 313 manages a free list of physical blocks.
In other words, the module manages physical blocks that are not allocated to any logical block among all physical blocks on the disk storage device 32.

Hereinafter, the operation of this embodiment will be described with reference to a flowchart showing the operation when the disk input / output module 312 of FIG. 12 receives a write request.

First, when the disk input / output module 312 receives a write request, the disk input / output module 312 stores a set of a current logical block and a physical block with respect to a specified logical block.
15 (step S121), a new physical block is secured from the invalid block management module 313, and this physical block is registered in the address translation map 311 as a mapping for the specified logical block (step S122). Thereafter, write processing (step S123: data write to the physical block specified by the address conversion map 311) is performed on this physical block.

In the checkpoint process, all physical blocks registered in the difference recording mechanism 315 are released by passing them to the invalid block management module 313, and all the logical block addresses registered in the difference recording mechanism 315 are released. And the registration of the set of the physical block address is deleted. In the recovery processing, for all the logical blocks registered in the difference recording mechanism 315, the physical blocks mapped to the logical blocks on the address translation map 311 are invalidated by the invalid block management module 31.
3 to be released by passing it to the address conversion map 311.
By overwriting the mapping between the logical block and the physical block registered in the difference recording mechanism 315, the address conversion map 3 at the latest checkpoint
11 is restored. Note that the address conversion map 311, the invalid block management module 313, and the difference recording mechanism 315
Is configured on a memory to be rolled back, the state of each structure can be automatically restored by rollback to the state at the time of the checkpoint, so that the recovery processing here is not necessary.

When the system is shut down, the address conversion map 311 is stored in the nonvolatile memory 33 in the same manner as in a normal system having a one-to-one address conversion map, so that the shutdown of the address conversion map is completed at the next boot. Restore to the point in time. (Fourth Embodiment) FIG. 13 is a block diagram showing a schematic configuration of a high-reliability computer system according to a fourth embodiment of the present invention. In the figure, reference numeral 41 denotes a computer main body, which is related to the embodiment of the present invention.
, A disk input / output module (disk I / O module) 412, and an invalid block management module 413
Then, the difference recording mechanism 415 is allocated and stored in the memory. Reference numeral 42 denotes a large-capacity disk device serving as a secondary storage device, and reference numeral 43 denotes a nonvolatile memory (NVRAM).

In the embodiment shown here, a physical block having a logical block image at the time of the checkpoint is managed by the difference recording mechanism 415 for a logical block newly written after the latest checkpoint. The difference recording mechanism 415 records a set of the logical block address and the address of the physical block assigned at the time of the check point, for the logical block of the disk device 42 that has newly written after the latest check point (FIG. 14).

FIG. 15 shows an invalid block management module 41.
3 shows the structure of No. 3. Invalid block management module 413
Manages a free list of physical blocks. That is,
This is a module that manages physical blocks that are not allocated to any logical block among all physical blocks on the disk device 42.

Hereinafter, the operation of this embodiment will be described with reference to the flowchart of FIG. 16 showing the operation when the disk input / output module 412 receives a write request.

When the disk input / output module 412 receives a write request, the disk input / output module 412 first stores the logical address in the difference recording mechanism 41.
5 is checked (step S16).
1). If the designated logical block is the difference recording mechanism 4
15 (N in S161), for the specified logical block, the current pair of the logical block and the physical block is registered in the difference recording mechanism 415 (step S163), and the invalid block management module 413 newly registers the set. A physical block is secured, and this physical block is registered in the address translation map 411 as a mapping for the specified logical block (step S164). Thereafter, write processing is performed on the physical block mapped to this logical block (step S16).
2). On the other hand, if it exists in the difference recording mechanism 415 (Y in step S161), the process proceeds to step S162 to perform a write process on the physical block specified by the address conversion map 411.

In the checkpoint process, all physical blocks registered in the difference recording mechanism 415 are released by passing them to the invalid block management module 413, and all logical block addresses registered in the difference recording mechanism 415 are released. And the registration of the set of the physical block address is deleted. In the recovery processing, all the logical blocks registered in the difference recording mechanism 415 are released by passing the physical block mapped to the logical block on the address conversion map 411 to the invalid block management module 413, and the address conversion is performed. By overwriting the map 411 with the mapping of the logical block and the physical block registered in the difference recording mechanism 415,
Address conversion map 411 at the time of the latest checkpoint
Restore the contents of As in the embodiment shown in FIG. 9, the address conversion map 411, the invalid block management module 4
13. When the difference recording mechanism 415 is configured on a roll-back memory, the state of each structure can be automatically restored by rollback to the state at the time of the checkpoint, so that the recovery process becomes unnecessary. At the time of system shutdown, the address conversion map 411 is stored in the non-volatile memory 43 in the same manner as in a normal system having a one-to-one address conversion map, so that the next time the boot is performed, the address conversion map 411 is set to the state at the time of shutdown completion. Restore. As a result, in the computer system of the checkpoint / rollback method, the performance of the system is improved and the capacity efficiency of the secondary storage device is improved by issuing the I / O to the secondary storage device without delay. . (Fifth Embodiment) FIG. 17 is a block diagram showing a schematic configuration of a high-reliability computer system according to a fifth embodiment of the present invention. In the figure, reference numeral 51 denotes a computer main body, which is related to the embodiment of the present invention.
And a disk input / output module (disk I / O module) 512, an invalid block management module 513, and an address conversion map paging mechanism 516 are allocated and stored in the memory. A large-capacity disk device 52 serving as a secondary storage device has an address conversion map 521. 53 is a nonvolatile memory (NVRAM).

In the embodiment shown here, the disk input / output module 512 normally receives a disk input / output request from the file system and issues an input / output request to the disk device 52. The disk input / output module 512 manages the logical address and the physical address of the data block of the disk device 52.

FIG. 18 shows the disk input / output module 512.
2 shows an example of an address conversion map used in the example. The address translation map 511 has entries of two physical block addresses, valid and old valid, for all logical blocks of the disk device 52. That is, the address conversion map 511
Out of the entries, only a finite number of entries exist on the main memory, and the other entries are
2 on

Here, in the address translation map 511, a valid physical block is a physical block having the latest disk image of this logical block. The old valid physical block is a logical block that has a disk image at the latest checkpoint of this logical block and another physical block has been written to the corresponding logical block since the latest checkpoint. Is a physical block that is allocated as an effective physical block to One valid physical block exists for all logical blocks, and the address of this physical block is stored in the valid physical block address entry of the address conversion map 511. For a logical block having an old valid physical block, the address of the old valid physical block is stored in the old valid physical address entry. For a logical block having no old valid physical block, NULL is recorded in the old valid physical block entry.

FIG. 19 shows an invalid block management module 51.
3 shows the structure of No. 3. Invalid block management module 513
Manages a free list of physical blocks. That is,
It is a module that manages physical blocks that are neither valid physical blocks of any logical block nor old valid physical blocks among all physical blocks on the disk device 52.

Hereinafter, the operation of this embodiment will be described with reference to a flowchart showing the operation when the disk input / output module 512 of FIG. 20 receives a write request.

When the disk input / output module 512 receives a write request, the disk input / output module 512 issues a write request in the following procedure. First, when the entry corresponding to the designated logical block does not exist in the address translation map 511 on the memory (N in step S502), the address translation map paging mechanism 516 performs paging of the address translation map 511, and executes the designated logic. The entry of the address translation map 511 corresponding to the block is placed on the memory (steps S508 and S510). An entry corresponding to the specified logical block exists in the address conversion map 511 on the memory (Y in step S502), and an old valid physical block address is registered for the logical block specified in the address conversion map 511. If yes (Y in S504), a write process is performed on a physical block registered as a currently valid physical block (step S506). In the address translation map 511,
If the old valid physical block is NULL for the specified logical block (N in step S504), the address of the currently valid physical block for the specified logical block is changed to the old valid physical block. It registers as an address (step S512), secures a new physical block from the invalid block management module 513, and registers this physical block in the address conversion map 511 as a valid physical block of the specified logical block (step S514). Then, a write process is performed on this physical block (step S506).

At the time of system shutdown, the address conversion map 511 is stored in the nonvolatile memory 53, so that the address conversion map is restored to the state at the time of shutdown completion at the next boot.

The function of the address translation map paging mechanism 516 will be described. The address translation map 511 in the memory is to return to the state at the latest checkpoint by rollback.

When it is necessary to rewrite the address translation map 511 for a certain logical block on the disk device 52, if an entry of the address translation map 511 for the logical block exists in the memory,
It is only necessary to appropriately rewrite each valid / old valid value of the entry. On the other hand, when the entry does not exist in the memory, that is, when it is paged out on the disk, one or more entries of the address translation map 511 on the memory are selected by an appropriate algorithm (such as LRU), and the selected Page out the entry to disk. At this time, the entry of the address conversion map 511 for the page-out destination logical block is also set correctly. In such a procedure, the address translation map 51 on the memory
1 is replaced. Sixth Embodiment FIG. 21 is a block diagram showing a schematic configuration of a high-reliability computer system according to a sixth embodiment of the present invention. In the figure, reference numeral 61 denotes a computer main body, which is related to the embodiment of the present invention.
Then, a disk input / output module (disk I / O module) 612, an invalid block management module 613, an address conversion map paging mechanism 616, and an address difference table 617 are allocated and stored in the memory. 6
Reference numeral 2 denotes a large-capacity disk device serving as a secondary storage device,
It has an address translation map 621. 63 is a nonvolatile memory (NVRAM).

In the embodiment shown here, the disk input / output module 612 normally receives a disk input / output request from the file system and issues an input / output request to the disk device 62. In this embodiment, the disk input / output module 612 manages the logical address and the physical address of the data block of the disk device 62.

FIG. 22 shows the disk input / output module 612.
4 shows an example of the address conversion map 611 used in the example. The address translation map 611 has entries of two physical block addresses of valid and old valid for all logical blocks of the disk device 62. Address conversion map 61
1 is a storage area that does not return to the latest checkpoint state by rollback (for example, PC
Only a finite number of entries exist on the memory on the I device), and the other entries exist on the disk device 62.

In the address translation map 611, a valid physical block is a physical block having the latest disk image of this logical block. The old valid physical block is a logical block that has a disk image at the latest checkpoint of this logical block and another physical block has been written to the corresponding logical block since the latest checkpoint. Is a physical block that is allocated as an effective physical block to There is one valid physical block for all logical blocks,
The address of this physical block is stored in the valid physical block address entry of the address conversion map 611.

For a logical block having an old valid physical block, the address of the old valid physical block is stored in the old valid physical address entry. For a logical block in which an old valid physical block does not exist, NULL is recorded in the old valid physical block entry.

FIG. 23 shows an invalid block management module 61.
3 is shown. Invalid block management module 613
Manages a free list of physical blocks. That is,
This module manages physical blocks that are neither valid physical blocks of any logical block nor old valid physical blocks among all physical blocks on the disk device 62.

FIG. 24 shows an address map difference table 61.
7 is shown. The address map difference table 617 stores the change history of the address conversion map 611 on the memory after the latest check point. The contents of the area assigned to the address conversion map 611 on the memory are
When writing to the disk device 62 by paging the address translation map 611, the contents of the entry to be paged out are stored in the address map difference table 617. After the contents of the entry of the address translation map 611 on the memory are stored in the address map difference table 617 and written to the disk device 62, the entry of the address translation map 611 on this memory becomes invalid. When another entry is newly read from the address conversion map 612 on the disk device 62 to the address conversion map 611 on the memory, the entry can be read only into an invalid area of the address conversion map 611 on the memory.

Hereinafter, the operation of this embodiment will be described with reference to the flowchart shown in FIG. 25 which shows the operation when the disk input / output module 612 receives a write request.

When the disk input / output module 612 receives a write request, the disk input / output module 612 issues a write request in the following procedure. First, when the entry corresponding to the specified logical block does not exist in the address conversion map 611 on the memory (N in step S602), the address conversion map paging mechanism 616 performs paging of the address conversion map 611, and executes the specified logical block. The entry of the address conversion map 611 corresponding to the block is placed on the memory (steps S610 and S612), and the process proceeds to step S604. An entry corresponding to the specified logical block exists in the address conversion map 611 on the memory (Y in step S602), and an old valid physical block address is registered for the logical block specified in the address conversion map 611. If yes (Y in step S604), write processing is performed on a physical block registered as a currently valid physical block (step S60).
8). On the other hand, in the address translation map 611, the old valid physical block is NULL for the specified logical block.
(N in step S604), the address of the currently valid physical block for the specified logical block is registered as the old valid physical block address (step S614), and the invalid block management module 613
, A new physical block is secured, and this physical block is registered in the address translation map 611 as an effective physical block of the specified logical block (step S6).
14), write processing is performed on this physical block (step S616).

In the checkpoint process, all old valid physical blocks registered in the address translation map 611 are released by passing them to the invalid block management module, and the values of all old valid physical block entries are set to NULL.
L. In the recovery processing, in the address conversion map 611, the valid physical block is released by passing the valid physical block to the invalid block management module for all the logical blocks in which the value of the old valid physical address entry is not NULL. The value of the block address entry is moved to the valid physical block address entry, and the value of the old valid physical block address entry is set to NULL.
At the time of system shutdown, the address conversion map 61
By saving 1 in the nonvolatile memory 63, the address conversion map 611 is restored to the state at the time of the shutdown completion at the next boot.

The address conversion map paging mechanism 616 will be described. The address conversion map 611 in the memory does not return to the state at the time of the latest checkpoint by rollback.

When it is necessary to rewrite the address translation map 611 for a certain logical block on the disk device 62, if an entry of the address translation map for the logical block exists in the memory, each valid entry of that entry is It is only necessary to rewrite the old valid value appropriately.

If the entry does not exist in the memory (it is paged out on the disk), one or more entries of the address translation map 611 on the memory are selected by an appropriate algorithm (such as LRU), and The selected entry is copied to the address map difference table 617 on the memory. Further, the selected entry is paged out to the disk device 62. At this time, the address translation map entry for the logical block at the page-out destination is also set correctly. In such a procedure, a part of the address conversion map 611 on the memory is replaced.

As part of the checkpoint processing, the address map difference table 617 is cleared. As part of the recovery processing, each entry recorded in the address map difference table 617 is written back to the address conversion map 611.

The methods described in the above embodiments can be provided as a computer-readable storage medium programmed so that the methods can be executed by a computer.

[0065]

As described above in detail, according to the present invention, a checkpoint / rollback computer system can issue an input / output to / from a secondary storage device without waiting for a checkpoint. Improves system performance. Furthermore, since the number of redundant physical blocks on the secondary storage device can be reduced, the capacity efficiency of the secondary storage device can be improved. Further, by using the time stamp, the checkpoint processing is lightened, and the performance of the system is further improved.

The embodiments shown in FIGS. 1 and 5 can be applied to one-to-many special mapping of logical blocks and physical blocks, whereas the embodiments shown in FIGS. One general mapping,
In the computer system of the checkpoint / rollback system, the input / output to / from the secondary storage device can be issued without waiting for the checkpoint, thereby improving the performance of the system. Furthermore, since the number of redundant physical blocks on the secondary storage device can be reduced, the capacity efficiency of the secondary storage device can be improved.

In the embodiment shown in FIG. 13, the processing at the time of the checkpoint is lighter than that in the embodiment shown in FIG. 9, so that the performance of the system is further improved.

Further, according to the present invention, the checkpoint
In a computer system of the rollback system, input / output to / from a secondary storage device can be issued without waiting for a checkpoint, so that system performance can be improved. In addition, regardless of the capacity of the secondary storage device, the required storage area on the main memory or in the address conversion map is constant, and the main memory is not pressed. The load of checkpoint processing is also reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a highly reliable computer system according to a first embodiment of the present invention.

FIG. 2 is an exemplary view showing the structure of an address conversion map according to the embodiment;

FIG. 3 is an exemplary view showing the structure of an invalid block management module according to the embodiment;

FIG. 4 is an exemplary flowchart showing the operation procedure of the disk input / output module according to the embodiment;

FIG. 5 is a block diagram showing a schematic configuration of a highly reliable computer system according to a second embodiment of the present invention.

FIG. 6 is a view showing the structure of an address translation map according to the embodiment;

FIG. 7 is an exemplary view showing the structure of an invalid block management module according to the embodiment;

FIG. 8 is an exemplary flowchart showing the operation procedure of the disk input / output module according to the embodiment;

FIG. 9 is a block diagram showing a schematic configuration of a high-reliability computer system according to a third embodiment of the present invention.

FIG. 10 is an exemplary view showing the structure of a difference recording mechanism according to the embodiment.

FIG. 11 is an exemplary view showing the structure of an invalid block management module according to the embodiment;

FIG. 12 is an exemplary flowchart showing the operation procedure of the disk input / output module according to the embodiment;

FIG. 13 is a block diagram showing a schematic configuration of a high-reliability computer system according to a fourth embodiment of the present invention.

FIG. 14 is an exemplary view showing the structure of a difference recording mechanism according to the embodiment.

FIG. 15 is an exemplary view showing the structure of an invalid block management module according to the embodiment;

FIG. 16 is an exemplary flowchart showing the operation procedure of the disk input / output module according to the embodiment.

FIG. 17 is a block diagram showing a schematic configuration of a high-reliability computer system according to a fifth embodiment of the present invention.

FIG. 18 is an exemplary view showing the structure of a dress conversion map according to the embodiment;

FIG. 19 is an exemplary view showing the structure of an invalid block management module according to the embodiment;

FIG. 20 is an exemplary flowchart showing the operation procedure of the disk input / output module according to the embodiment;

FIG. 21 is a block diagram showing a schematic configuration of a high-reliability computer system according to a sixth embodiment of the present invention.

FIG. 22 is an exemplary view showing the structure of an address translation map according to the embodiment;

FIG. 23 is an exemplary view showing the structure of an invalid block management module according to the embodiment;

FIG. 24 is a view showing the structure of an address difference table according to the embodiment.

FIG. 25 is an exemplary flowchart showing the operation procedure of the disk input / output module according to the embodiment;

[Explanation of symbols]

11 (22, 32, 42, 51, 61) ... computer body,
12 (22, 32, 42, 52, 62): Disk device, 13 (23, 33, 43, 53, 63): Non-volatile memory, 111 (211, 311, 411, 511, 6)
11) Address conversion map 112 (212, 31)
2, 412, 512, 612) Disk input / output module, 113 (213, 313, 413, 513, 61)
3) Invalid block management module 214 Time stamp 315 (415) Difference recording mechanism 516 (6)
16) Address translation map paging mechanism 617
Address difference table.

Claims (22)

[Claims] 1. A checkpoint / roll function for performing a process while periodically collecting checkpoints and, when a failure occurs in the system, returning the state of the system to the immediately preceding checkpoint and re-executing the process. In a computer system having a back mechanism and an input / output module for writing and reading data to and from a secondary storage device, at least one physical block for each logical block constituting the secondary storage device An address translation map for assigning a pointer to a physical block having a logical block image at the latest checkpoint, a pointer to a logical block having the latest logical block image for each logical block, and Manages physical blocks without valid data When the I / O module receives a write request to the secondary storage device, the I / O module writes data to a physical block different from the latest checkpoint image for the logical block. When the input / output module receives a read request for the secondary storage device, the physical block having the latest logical block image for the logical block is registered as the latest logical block image in the address conversion map. Data is read from the block, and in the checkpoint process, for each logical block on the address translation map, a physical block having the latest logical block image is replaced with a physical block having the latest checkpoint image and the latest image for this logical block. At the time of recovery, for each logical block on the address translation map, the physical block having the logical block image at the time of the checkpoint shall be the physical block having the checkpoint image for this logical block and the latest image. Characteristic input / output control method in a computer system. 2. A checkpoint / role for performing a process while periodically collecting checkpoints, and when a failure occurs in the system, returning the state of the system to the point of the checkpoint collected immediately before and re-executing the process. In a computer system having a back mechanism and an input / output module for writing and reading data to and from a secondary storage device, one or more physical blocks are assigned to each logical block of the secondary storage device. Allocate a pointer to a physical block having a logical block image at the latest checkpoint, a pointer to a physical block having a latest logical block image for each logical block,
Further, it has an address conversion map for recording the time of the last write processing to each physical block, and an invalid block management module for managing physical blocks that do not have valid data on the secondary storage device. When the output module receives a write request to the secondary storage device, data is written to the logical block in a physical block different from the latest checkpoint image, and this physical block is registered as the latest logical block image. At the same time, the time at which the writing process was performed on this physical block is recorded in the address conversion map, and when the input / output module receives a read request from the secondary storage device, it has the latest logical block image for the logical block. Read data from physical block In the checkpoint process, the current time is stored as a checkpoint timestamp. At the time of recovery, each logical block on the address translation map has the latest physical timestamp with a timestamp newer than the checkpoint timestamp. If the pointer to the block is registered, pass the latest physical block to the invalid block management module and release it,
An input / output control method in a computer system, wherein a pointer to a physical block at a check point is registered in the address conversion map as a pointer to the latest physical block. 3. A checkpoint / role for performing a process while periodically collecting checkpoints, and when a failure occurs in the system, returning the state of the system to the immediately preceding checkpoint and re-executing the process. An address translation map that has a back mechanism and maps logical blocks and physical blocks of the secondary storage device;
In a computer system having an input / output module for writing and reading data to and from the secondary storage device, a logical block that has been subjected to write processing after the latest checkpoint, and the logical block and the A difference recording mechanism for recording a set of physical blocks allocated to the logical block, and an invalid block management module for managing a physical block having no valid data on the secondary storage device. When the output module receives a write request to the secondary storage device, the mapping of the logical block and the physical block on the address translation map at this time is recorded in the difference recording mechanism, and a new one is sent from the invalid block management module. Assign the allocated physical block to this logical block. Registered in the address translation map as a physical block, write processing is performed on this physical block, and in the checkpoint processing, all physical blocks recorded in the difference recording mechanism are released to the invalid block management module, and the difference recording mechanism is released. In the recovery, all the logical blocks recorded in the difference recording mechanism are mapped to the logical blocks by the address conversion map. Passing the physical block to the invalid block management module and releasing it, and writing back the set of the logical block and the physical block registered in the difference recording mechanism to the address translation map, thereby obtaining the Computer system characterized by restoring an address translation map Input and output control method in the arm. 4. A checkpoint / roller for performing a process while periodically collecting checkpoints and, when a failure occurs in the system, restoring the state of the system to the immediately preceding checkpoint and re-executing the process. An address translation map that has a back mechanism and maps logical blocks and physical blocks of the secondary storage device;
In a computer system having an input / output module for writing and reading data to and from the secondary storage device, a logical block that has been subjected to write processing after the latest checkpoint, and the logical block and the A difference recording mechanism for recording a set of physical blocks assigned to the logical block, and an invalid block management module for managing a physical block having no valid data on the secondary storage device. If the specified logical block is not registered in the differential recording mechanism when the module receives the write request to the secondary storage device, the mapping of the logical block and the physical block at this time on the address translation map Is recorded in the difference recording mechanism, and the invalid block management mode is recorded. After a physical block newly secured from the logical block is registered as a physical block for this logical block in the address translation map, a write process is performed on this physical block, and the specified logical block is registered in the above-mentioned differential recording mechanism. For example, data is written to the physical block mapped to this logical block in the address conversion map. In the checkpoint process, all physical blocks recorded in the difference recording mechanism are released to the invalid block management module, and the difference recording is performed. The registration of all pairs of logical blocks and physical blocks recorded in the mechanism is deleted, and at the time of recovery, all the logical blocks recorded in the difference recording mechanism are mapped to the logical blocks in the address translation map. The physical block that is To the address conversion map, and write back the set of the logical block and the physical block registered in the difference recording mechanism to the address conversion map, thereby restoring the contents of the address conversion map at the latest checkpoint. An input / output control method in a computer system. 5. A checkpoint / role for performing a process while periodically collecting checkpoints, and when a failure occurs in the system, returning the state of the system to the point of the checkpoint collected immediately before and re-executing the process. In a computer system having an input / output module for writing and reading data to and from a secondary storage device having a back mechanism, one or more physical blocks are provided for each logical block of the secondary storage device. And assign a pointer to the physical block that has the logical block image at the latest checkpoint and a pointer to the physical block that has the latest logical block image for each logical block. Return memory area and located on the secondary storage device An address translation map, an address translation map paging mechanism for paging the address translation map between the secondary storage device and the memory, and an invalid block for managing a physical block having no valid data on the secondary storage device When the input / output module receives a write request to the secondary storage device, if the address translation map corresponding to the logical block does not exist in the memory, the address map paging mechanism After writing a partial area of the address conversion map on the memory on the secondary storage device after the next checkpoint, the entry of the address conversion map corresponding to the logical block corresponding to this area is written to the Read from secondary storage and Writes data to a physical block different from the new checkpoint image, registers this physical block as the latest logical block image in the address translation map, and accepts a read request from the secondary storage device by the input / output module If the address translation map corresponding to the corresponding logical block does not exist in the memory, the address map paging mechanism causes a partial area of the address translation map in the memory to be moved to the next checkpoint or later. After writing on the secondary storage device, an entry of the address translation map corresponding to the logical block corresponding to this area is read from the secondary storage device, and data is read from the physical block having the latest logical block image for the logical block. Read In the checkpoint processing, for each logical block on the address translation map, the physical block having the latest logical block image is used as the physical block having the checkpoint image and the latest image for this logical block. An input / output control method in a computer system, characterized in that: 6. A checkpoint / roller for performing a process while periodically collecting checkpoints and, when a failure occurs in the system, returning the state of the system to the point of the checkpoint collected immediately before and re-executing the process. In a computer system having a back mechanism and an input / output module for writing and reading data to and from a secondary storage device, one or more physical blocks are assigned to each logical block of the secondary storage device. Allocation, assigning a pointer to a physical block having a logical block image at the latest checkpoint, a pointer to a physical block having a latest logical block image to each logical block,
A memory area that does not return to the state at the time of the latest checkpoint due to recovery, an address translation map arranged on the secondary storage device, and an address translation map for paging the address translation map between the secondary storage device and the memory A paging mechanism, an address map difference table for storing a change history of the address translation map on the memory after the latest checkpoint, and an invalid block management for managing a physical block having no valid data on the secondary storage device When the input / output module receives a write request to the secondary storage device, if the address conversion map corresponding to the logical block does not exist in the memory, the address map paging mechanism By the above address on the memory A part of the area of the conversion map is registered in the address map difference table, and after writing to the secondary storage device after the next checkpoint, the area of the address conversion map corresponding to the logical block corresponding to this area is written. The entry is read from the secondary storage device, data is written to a physical block different from the latest checkpoint image of the logical block, and this physical block is registered in the address conversion map as the latest logical block image, and When the output module receives a read request from the secondary storage device, if the address translation map corresponding to the corresponding logical block does not exist in the memory, the address map paging mechanism causes the address map paging mechanism to execute the read operation on the memory. Some of the address translation maps The area is registered in the address map difference table, and after writing to the secondary storage device after the next checkpoint, the entry of the address conversion map corresponding to the logical block corresponding to this area is stored in the secondary storage. Read from the device, read data from the physical block having the latest logical block image for the logical block, and in the checkpoint process, for each logical block on the address translation map, A physical block having a checkpoint image and the latest image for this logical block is deleted, and the entries registered in the address map difference table are deleted. Output control method in a computer system, wherein a write back to the conversion table. 7. A checkpoint / roller that periodically proceeds with processing while collecting checkpoints, and when a failure occurs in the system, returns the state of the system to the point of the checkpoint that was collected immediately before and re-executes processing. In a computer system having a back mechanism and an input / output module for writing and reading data to and from a secondary storage device, one or more physical blocks are assigned to each logical block of the secondary storage device. Allocation, assigning a pointer to a physical block having a logical block image at the latest checkpoint, a pointer to a physical block having a latest logical block image to each logical block,
A memory area for recording the time of the last write processing in each physical block, and returning to the latest checkpoint by recovery, an address translation map arranged on the secondary storage device, the secondary storage device and the memory An address translation map paging mechanism that performs paging of the address translation map, and an invalid block management module that manages a physical block that does not have valid data on the secondary storage device. When a write request to the secondary storage device is received, if the address translation map corresponding to the relevant logical block does not exist in the memory, the address translation map paging mechanism causes a part of the address translation map in the memory to be read. Area after the next checkpoint After writing to the secondary storage device, the entry of the address translation map corresponding to the logical block corresponding to this area is read from the secondary storage device, and the logical block is stored in a physical block different from the latest checkpoint image. Write the data to the physical block to make the physical block the latest logical block image, and register the time at which the write processing was performed on the physical block in the address conversion map. When the read request is received, if the address translation map corresponding to the relevant logical block does not exist in the memory, the address map paging mechanism causes a partial area of the address translation map on the memory to be updated in the next time. After the checkpoint After writing on the storage device, read the entry of the address translation map corresponding to the logical block corresponding to this area from the secondary storage device, read data from the physical block having the latest logical block image for the logical block, In the checkpoint processing, an input / output control method in a computer system, wherein a current time is stored as a checkpoint time stamp. 8. A checkpoint / roll which performs processing while periodically collecting checkpoints, and when a failure occurs in the system, returns the state of the system to the time of the immediately preceding checkpoint and re-executes the processing. In a computer system having a back mechanism and an input / output module for writing and reading data to and from a secondary storage device, one or more physical blocks are assigned to each logical block of the secondary storage device. Allocate, for each logical block, assign a pointer to the physical block having the logical block image at the latest checkpoint, assign a pointer to the physical block having the latest logical block image,
A memory area that records the time of the last write processing in each physical block, does not return to the state at the time of the latest checkpoint by recovery, and an address translation map arranged on the secondary storage device; An address translation map paging mechanism for paging the address translation map between the memory and the memory; an address map difference table for storing a change history of the address translation map on the memory after the latest checkpoint; An invalid block management module that manages a physical block having no valid data, wherein when the input / output module receives a write request to the secondary storage device, the address conversion map corresponding to a corresponding logical block is provided. Is not in the memory, The address conversion map corresponding to the logical block corresponding to this area after writing a part of the address conversion map on the memory to the secondary storage device after the next checkpoint by the paging mechanism. Is read from the secondary storage device, data is written to a physical block different from the latest checkpoint image for the logical block, and this physical block is used as the latest logical block image. The time at which the write processing was performed is registered in the address conversion map. When the input / output module receives a read request from the secondary storage device, the address conversion map corresponding to the corresponding logical block is stored in the memory. If it does not exist, the above address A partial area of the address translation map on the memory is registered in the address map difference table by the up-paging mechanism, and after writing to the secondary storage device after the next checkpoint, a logic corresponding to this area is written. The entry of the address translation map corresponding to the block is read from the secondary storage device, and data is read from the physical block having the latest logical block image for the logical block. And the entries registered in the address map difference table are deleted, and at the time of recovery, each entry registered in the address map difference table is written back to the address map difference table. Input and output control method in the machine system. 9. A memory area having a checkpoint / rollback mechanism and not returning to the state at the latest checkpoint by recovery for mapping a logical block and a physical block of a secondary storage device, and the secondary storage device In a computer system having an address conversion map above and an input / output module for writing and reading data to and from the secondary storage device, a logical block for which writing processing has been performed since the time of the latest checkpoint is referred to as a logical block. A block, a difference recording mechanism for recording a set of physical blocks allocated to the logical block before performing the write processing, and an address conversion map for paging the address conversion map between the secondary storage device and the memory On the paging mechanism and on the secondary storage device An invalid block management module that manages a physical block having no valid data, and when the input / output module receives a write request to the secondary storage device, the address conversion map corresponding to a corresponding logical block. Is not present in the memory, the address translation map paging mechanism writes the partial area of the address translation map in the memory to the secondary storage device after the next checkpoint, and then corresponds to this area. The entry of the address translation map corresponding to the logical block to be read is read from the secondary storage device, and the mapping of the logical block and the physical block on the address translation map at this time is recorded in the difference recording mechanism, and the invalid block is recorded. Physical block newly secured from the management module Is registered in the address translation map as a physical block for this logical block, a write process is performed on this physical block, and when the input / output module receives a read request from the secondary storage device, If the address translation map corresponding to the address does not exist in the memory, while registering a part of the address translation map on the memory in the address map difference table by the address map paging mechanism,
After writing to the secondary storage device after the next checkpoint, an entry of the address translation map corresponding to the logical block corresponding to this area is read from the secondary storage device, and the latest logical block image for the logical block is read. In the checkpoint process, all physical blocks recorded in the difference recording mechanism are released to the invalid block management module, and all logical blocks recorded in the difference recording mechanism are At the time of recovery, the physical block mapped to the logical block in the address conversion map is passed to the invalid block management module for all logical blocks recorded in the differential recording mechanism. And release the above address translation In the computer system, the address conversion map at the latest checkpoint is restored by writing back a set of logical blocks and physical blocks registered in the difference recording mechanism. I / O control method. 10. A memory area which has a checkpoint / rollback mechanism, does not return to the state at the latest checkpoint by recovery for mapping logical blocks and physical blocks of the secondary storage device, and A computer system having an input / output module for writing and reading data to and from the secondary storage device having an address conversion map arranged in the secondary storage device; The logical translation block having the address translation map paging mechanism for performing paging of a map, and for which a logical block has been written after the latest checkpoint, is assigned to this logical block and a physical block assigned to this logical block before the write processing is performed. Recording mechanism that records a set of An invalid block management module that manages a physical block having no valid data on the secondary storage device. When the input / output module receives a write request to the secondary storage device, If the address translation map corresponding to the block does not exist in the memory, the address translation map paging mechanism causes a partial area of the address translation map on the memory to be moved to the secondary storage device after the next checkpoint. After writing the data in the address conversion map corresponding to the logical block corresponding to this area, the entry is read from the secondary storage device. If the specified logical block is not registered in the difference recording mechanism, Mapping of blocks and physical blocks on the above address translation map After recording in the difference recording mechanism and registering the newly secured physical block from the invalid block management module as a physical block for this logical block in the address translation map, a write process is performed on this physical block, and If the logical block is registered in the difference recording mechanism, data is written to the physical block mapped to the logical block in the address conversion map, and the input / output module receives a read request from the secondary storage device. When the address translation map corresponding to the relevant logical block does not exist in the memory, the address map paging mechanism registers a partial area of the address translation map in the memory in the address map difference table. With the next check After writing to the secondary storage device after the time point, the entry of the address translation map corresponding to the logical block corresponding to this area is read from the secondary storage device, and the latest logical block image for the logical block is read. In the checkpoint process, all physical blocks recorded in the difference recording mechanism are released to the invalid block management module, and all logical blocks and physical blocks recorded in the difference recording mechanism are read out. At the time of recovery, for all the logical blocks recorded in the difference recording mechanism, the physical blocks mapped to the logical blocks in the address conversion map are passed to the invalid block management module. Release and use the above address translation map On the other hand, the address conversion map at the time of the latest checkpoint is restored by writing back a set of a logical block and a physical block registered in the difference recording mechanism. Method. 11. A checkpoint / role for performing a process while periodically collecting checkpoints and, when a failure occurs in the system, restoring the state of the system to the immediately preceding checkpoint and re-executing the process. A computer system having a back mechanism and an input / output module for writing and reading data to and from a secondary storage device is allocated and stored in the storage device. Address translation map for allocating at least one physical block to each logical block and assigning a pointer to a physical block having a logical block image at the latest checkpoint and a pointer to a physical block having the latest logical block image to each logical block Is allocated and retained Storage means, invalid block management means for managing a physical block having no valid data on the secondary storage device, and a logical block when the input / output module receives a write request to the secondary storage device. Writes data to a physical block different from the latest checkpoint image, registers the physical block as the latest logical block image in the address conversion map, and requests the input / output module to read from the secondary storage device. Is received, the data is read from the physical block having the latest logical block image for the logical block. In the checkpoint processing, the physical block having the latest logical block image is read for each logical block on the address modification map. Block this logical block At the time of recovery, the physical block having the logical block image at the time of the checkpoint is used as the checkpoint image for the logical block. An input / output control device in a computer system, comprising: an input / output control unit configured as a physical block having the latest image. 12. A checkpoint / role for performing processing while periodically collecting checkpoints, and when a failure occurs in the system, returning the state of the system to the point of the checkpoint collected immediately before and re-executing the processing. In a computer system having a back mechanism and an input / output module for writing and reading data to and from a secondary storage device, one or more physical blocks are assigned to each logical block of the secondary storage device. Allocate, for each logical block, assign a pointer to the physical block having the logical block image at the latest checkpoint, assign a pointer to the physical block having the latest logical block image,
In addition, an address conversion map that records the time of the last write processing for each physical block is allocated and retained, and an invalid block that manages physical blocks that do not have valid data on the secondary storage device When the management unit and the input / output module receive a write request to the secondary storage device, data is written to the logical block in a physical block different from the latest checkpoint image, and the physical block is updated to the latest. And the time at which the write processing was performed on this physical block is registered in the address translation map. When the input / output module receives a read request from the secondary storage device, the latest Physical block with a logical block image of In the checkpoint process, the current time is stored as a checkpoint timestamp, and at the time of recovery, each logical block on the address translation map has a newer timestamp than the checkpoint timestamp. When the pointer to the physical block is registered, the latest physical block is passed to the invalid block management module to be released, and the pointer to the physical block at the time of the check point is set as the pointer to the latest physical block. And an input / output control means for registering the input / output in the computer system. 13. A checkpoint / roll which periodically proceeds with processing while collecting checkpoints, and when a failure occurs in the system, returns the state of the system to the point of the checkpoint collected immediately before and executes the processing again. A computer system having a back mechanism and an input / output module for writing and reading data to and from the secondary storage device, and an address translation map for mapping logical blocks and physical blocks of the secondary storage device, A difference recording mechanism for recording a set of a logical block that has been subjected to a write process after the latest checkpoint, and a set of physical blocks allocated to the logical block before the write process; Manage physical blocks that do not have valid data on the device When the input / output module receives a write request to the secondary storage device, the mapping of the logical block and the physical block on the address translation map at this time is stored in the difference recording mechanism. The physical block newly recorded from the invalid block management module is recorded in the address conversion map as a physical block for the logical block, and a write process is performed on the physical block. All physical blocks recorded in the mechanism are released to the invalid block management module, all the logical block and physical block pairs recorded in the difference recording mechanism are deleted, and at the time of recovery, the difference recording mechanism is deleted. For all recorded logical blocks, the above A physical block mapped to the logical block in the address translation map is passed to the invalid block management module to be released, and a set of a logical block and a physical block registered in the difference recording mechanism is registered in the address translation map. Input / output control means for restoring the address conversion map at the time of the latest checkpoint by writing back the data. 14. A checkpoint / roll function for performing a process while periodically collecting checkpoints and, when a failure occurs in the system, returning the state of the system to the immediately preceding checkpoint and re-executing the process. A computer system having a back mechanism and an input / output module for writing and reading data to and from the secondary storage device, and an address translation map for mapping logical blocks and physical blocks of the secondary storage device, Difference recording means for recording a set of logical blocks for which logical processing has been performed since the latest checkpoint and physical blocks allocated to the logical blocks before performing the write processing; and a secondary storage device. Manages physical blocks that have no valid data on them An effective block management mechanism, and when the specified logical block is not registered in the difference recording mechanism when the input / output module receives a write request to the secondary storage device, The mapping of the block on the address translation map is recorded in the difference recording mechanism, and the newly secured physical block from the invalid block management module is registered in the address translation map as a physical block for this logical block. If the specified logical block is registered in the difference recording mechanism, data is written to the physical block mapped to this logical block in the address conversion map. All physical blocks recorded in the difference recording mechanism Block is released to the invalid block management module, and all the pairs of logical blocks and physical blocks recorded in the difference recording mechanism are deleted. For the logical block, the physical block mapped to the logical block in the address translation map is passed to the invalid block management mechanism to be released, and the logical block registered in the difference recording mechanism for the address translation map. And an input / output control means for restoring the contents of the address conversion map at the latest checkpoint by rewriting a set of physical blocks. 15. A checkpoint / roll function for performing a process while periodically collecting checkpoints, and when a failure occurs in the system, returning the state of the system to the immediately preceding checkpoint and re-executing the process. At least one physical block is allocated to each logical block constituting the secondary storage device, and a pointer to a physical block having a logical block image at the latest checkpoint and the latest logical block image are stored. Used in a computer system having an address conversion map for allocating a pointer to a physical block having the I / O module for writing and reading data to and from the secondary storage device, the I / O module receiving a write request When the logical address When the latest logical block image is registered for the specified logical block by referring to the address conversion map and the function for checking whether or not the latest logical block image exists, A function to write to the physical block registered as the latest logical block image, and if the latest logical block image is not assigned to the specified logical block, a new A function of securing a physical block, registering this physical block as an effective physical block having the latest image of the specified logical block in the address translation map, and writing the physical block is programmed and stored. Computer readable storage medium. 16. In the checkpoint process, physical address entries that are logical block images at the time of the latest checkpoint registered in the address conversion map are released by passing them to the invalid block management module, and all the latest checks are performed. A function to set a physical address entry that is a logical block image at the point in time to a specific value indicating that a physical block has not been allocated, and in the recovery processing, refer to the address conversion map to obtain the latest checkpoint time. For each logical block whose physical address entry that is a logical block image is not the specific value, the corresponding valid physical block is released by passing it to the invalid block management module, and the logical block image at the latest checkpoint is released. A function of moving the value of the physical block address entry that becomes the latest logical block image to the physical block address entry that becomes the latest logical block image, and setting the specific value to the physical block address entry that becomes the logical block image at the time of the latest checkpoint. A computer-readable program which is programmed and recorded with a function of storing the contents of the address conversion map in a nonvolatile storage area at the time of system shutdown, and restoring the contents of the address conversion map to the state at the time of shutdown completion at the next boot. 16. The storage medium according to 15. 17. A checkpoint / roll function for performing a process while periodically collecting checkpoints and, when a failure occurs in the system, returning the state of the system to the immediately preceding checkpoint and re-executing the process. At least one physical block is allocated to each logical block constituting the secondary storage device, and a pointer to a physical block having a logical block image at the latest checkpoint and the latest logical block image are stored. Used in a computer system having an address conversion map for allocating a pointer to a physical block and a time stamp, and an input / output module for writing and reading data to and from the secondary storage device. When you receive it, A function to check whether the time stamp of a block serving as a logical block image is older than the latest check point time stamp, and by referring to the contents of the address translation map, If the timestamp of the physical block is not older than the checkpoint timestamp, the function to write to the physical block registered as the currently valid physical block and the latest logical block image for the specified logical block If the timestamp of the physical block is older than the checkpoint timestamp, a function to further check whether an address which is a logical block image at the latest checkpoint exists, and the latest checkpoint of the specified logical block If the physical block address that is the logical block image at the time is not the specific value, the physical block is released by passing the physical block to the invalid block management module, and for the specified logical block,
A function for registering the address of the currently valid physical block, which is the latest logical block image, as the physical block address of the logical block image at the time of the latest checkpoint, and securing a new physical block from the invalid block management module, A computer-readable storage medium in which a block is registered as an effective physical block of a designated logical block in the address conversion map, and a function of writing the physical block is programmed and stored. 18. A function for registering the current time as a checkpoint time stamp in the checkpoint processing, and for each logical block on the address translation map, a valid physical block corresponding to the logical block in the recovery processing. If the timestamp is newer than the checkpoint timestamp, the valid physical block is released by passing it to the invalid block management module, and the value of the physical block address entry which is the logical block image at the latest checkpoint is updated to the latest logical block. A function to move to a physical block address which is a block image and set the value of a physical block address entry which is a logical block image at the time of the latest checkpoint to the above-mentioned specific value; 16. The computer-readable storage medium according to claim 15, wherein a function of restoring the address conversion map to a state at the time of shutdown completion at the next boot is programmed and recorded by storing the conversion map in a nonvolatile storage. 19. A checkpoint / roller for performing a process while periodically collecting checkpoints and, when a failure occurs in the system, returning the state of the system to the point of the checkpoint just collected and re-executing the process. Used in a computer system having a back mechanism, an address translation map for mapping logical blocks and physical blocks of a secondary storage device, and an input / output module for writing and reading data to and from the secondary storage device. When the input / output module receives the write request, a function for registering the current logical block and physical block pair in the differential registration mechanism with respect to the specified logical block, and a new physical block from the invalid block management module. Allocate this physical block to the specified logical block. A function of registering in the address translation map as a mapping for the click, a computer-readable storage medium which functions and to perform a write to this physical block after the completion of registration is programmed and stored. 20. In the checkpoint process, all physical blocks registered in the difference recording mechanism are released by being passed to the invalid block management module, and all logical block addresses registered in the difference registration mechanism are released. And a function of deleting the registration of the set of the physical block address and the physical block address, and in the recovery process, for all the logical blocks registered in the difference recording mechanism, Release by passing to the invalid block management module, and overwrite the address conversion map with the mapping of the logical block and the physical block registered in the difference recording mechanism, thereby obtaining the address conversion map at the latest checkpoint. Function to restore the contents of 20. The computer-readable storage medium according to claim 19, wherein are stored and programmed. 21. A checkpoint / role for performing a process while periodically collecting checkpoints, and when a failure occurs in the system, returning the state of the system to the immediately preceding checkpoint and re-executing the process. Used in a computer system having a back mechanism, an address translation map for mapping logical blocks and physical blocks of a secondary storage device, and an input / output module for writing and reading data to and from the secondary storage device. When the input / output module receives the write request, if the specified logical block is not registered in the differential recording mechanism, the current logical block and physical block pair of the specified logical block is stored in the differential recording mechanism. New function from the invalid block management module And a function of registering the physical block in the address translation map as a mapping for the specified logical block, and performing a write operation on the physical block mapped to the logical block after completing the registration. A computer-readable storage medium in which a function to be programmed is programmed and stored. 22. In the checkpoint processing, all physical blocks registered in the difference recording mechanism are released by being passed to the invalid block management module, and all logical block addresses registered in the difference registration mechanism are released. And a function of deleting the registration of the set of the physical block address and the physical block address, and in the recovery processing, for all the logical blocks registered in the difference recording mechanism, the physical block mapped to the logical block on the address conversion map is deleted. By releasing by passing to the invalid block management module, by overwriting the mapping of the logical block and the physical block registered in the difference recording mechanism with respect to the address translation map,
22. The computer-readable storage medium according to claim 21, wherein a function of restoring the contents of the address conversion map at the time of the latest checkpoint is programmed and stored.