JPH11345137A - Check point roll-back type computer system, memory management control method for the same system and storage medium stored with the same method therein - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain improvement in performance by avoiding unwanted backup by adding a function to the memory management of an OS software, completely eliminating the need for the backup of write from an I/O to a memory and decreasing the number of bus transactions. SOLUTION: This system is provided with CPU 1a,..., 1n, backup device 2, I/O device 3, memory 4, system bus 5, check point modules 51 and memory managing module 52 and for the memory which is free at a recent check point time, backup is dispensed with from the time point of memory acquiring request to the next check point. By setting such a memory as a non-backup object, performance can be improved while decreasing the bus transactions.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer high-reliability technology, that is, a computer of a memory checkpoint / rollback system, a memory management control method in a checkpoint rollback computer, and program information of the method are stored. Computer-readable storage medium.

[0002]

2. Description of the Related Art Conventionally, in a computer system employing a memory checkpoint / rollback method, a consistent state of the system is periodically stored in a memory. When a failure / failure has occurred, the failure / failure recovery is realized by returning to the state of the latest checkpoint stored in the memory.

For example, in a typical implementation, a checkpoint device is placed on a system bus, the device captures a write transaction from a CPU (central processing unit), and stores data before rewriting in a buffer. The OS (operating system) software periodically flushes the CPU registers and dirty entries on the cache, and then clears the buffer of the device.

[0004]

However, in the above-described technique, it is necessary to capture a write transaction and store data before rewriting in a buffer.
At this point, performance degradation occurs due to an increase in bus transactions. Thus, the smaller the part to be backed up, the better the performance.

Accordingly, the present invention has been made in view of the above circumstances, and solves the above-mentioned disadvantages, adds a function to memory management of OS software, avoids unnecessary backup, and improves I / O. A checkpoint rollback computer system which realizes that backup of writing from O to memory (read from I / O) is completely unnecessary, thereby reducing bus transactions and improving performance. An object of the present invention is to provide a memory management control method in a checkpoint rollback computer and a computer-readable storage medium storing program information of the memory management control method in a checkpoint rollback computer system.

[0006]

In order to achieve the above object, the present invention employs a recovery method based on a memory checkpoint / rollback. A backup device is provided with means for setting a non-backup target area by software. In the provided system, a counter that records a checkpoint generation added at the time of a checkpoint, and a checkpoint generation at that time are recorded in a software management structure of a minimum unit of memory acquisition / release at the time of a memory release request. A checkpoint generation storage area, a value stored in the checkpoint generation storage area of the memory management structure acquired at the time of a memory acquisition request,
Compare the checkpoint generation at that time, and if the value stored in the checkpoint generation storage area is smaller than the current checkpoint generation area, set the page as a non-backup target area and then request the The memory is returned, and at the time of the next checkpoint, means for resetting the backup non-target area set by the above means in the previous checkpoint section is provided.

[0007] According to such a configuration, for a memory that was free at the time of the most recent checkpoint, backup is not required from the time of a memory acquisition request to the next checkpoint. By setting such a memory as a non-backup target, it is possible to reduce bus transactions and improve performance.

[0008] In order to achieve the above object, the present invention provides:
A recovery method based on memory checkpoint / rollback is adopted, and the backup device is provided with a means that can set by software whether or not to back up data written from the I / O to the memory. A counter for recording a checkpoint generation to be added at the time of a checkpoint, a checkpoint generation storage area for recording a checkpoint generation at that time in a software management structure of a minimum unit of memory acquisition / release at the time of a memory release request, At the time of a memory release request, the checkpoint generation at that time is stored in the checkpoint generation storage area of the memory management structure to be released, and the checkpoint generation storage area is stored for the memory allocation for the write request from the I / O to the memory. Value stored in Characterized in that a means for returning the smaller memory than the checkpoint generation at that time.

According to such a configuration, a memory that has been free at the time of the most recent checkpoint is allocated to a write request from the I / O to the memory, so that backup of the memory is not required. Become. As a result, bus transactions can be reduced, leading to an improvement in performance.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows a computer to which the present embodiment is applied, and includes a CPU 1a,..., 1n, a backup device 2, an I / O device 3, a memory 4, and a system bus 5 for connecting the above elements.

FIG. 2 conceptually shows information provided in the backup device 2 on which the present invention is based. In FIG. 2, one “back-up non-target area” is expressed by one line. “Valid / Invalid” stores whether the information of the bank is valid or invalid. The “address information” stores the start address of the non-backup target area, and the “size” stores the size of the non-backup target area. Now, assuming that the start address is α and the size is A, (α, α +
The data in the address area in the range A) is not backed up.

FIG. 3 is a flowchart showing the operation of the backup apparatus 2 shown in FIG.
The operation will be described with reference to FIG. 4 which is a flowchart showing the operation of the software using the functions of FIG.

First, it is determined whether the address is registered as a non-backup target area (step S31 in FIG. 3). If the area is registered as a non-backup target area (Yes in step S31), the validity of the information is subsequently determined (step S32).
If it is determined that the information is valid (Yes in step S32), a backup operation is performed (step S33).

On the other hand, if the area is not registered as a backup non-target area in step S31 (No in step S31), a backup operation is performed (step S33).

Next, the flow of the operation of software for setting information in the non-backup target area will be described. When the process of taking the exclusive lock is performed (step S4 in FIG. 4)
1), a process of searching for an invalid entry is performed (step S4)
2). Here, if there is an invalid entry (step S
43 (Yes), the designated address and the size are set, and a valid entry is made (step S44). Then, the exclusive lock is released (step S45).

On the other hand, when it is determined in step S43 that there is no invalid entry (No in step S43), the exclusive lock is released (step S45).
Next, in the system of the present embodiment, that is, in the system in which the recovery method based on the memory checkpoint / rollback is adopted, and the backup device 2 is provided with means for setting the non-backup target area by software, A counter for recording a checkpoint generation to be added at the time of a checkpoint, a checkpoint generation storage area for recording a checkpoint generation at that time in a software management structure of a minimum unit of memory acquisition / release at the time of a memory release request, and a memory acquisition The value stored in the checkpoint generation storage area of the memory management structure obtained at the time of the request is compared with the checkpoint generation at that time, and the value stored in the checkpoint generation storage area is the checkpoint at that time. If it is smaller than the generation, After the page is set as a non-backup area, the memory is returned to the requester, and at the next checkpoint, there is provided a means for resetting the non-backup area set by the above means in the previous checkpoint section. A description will be given with reference to FIG. 5 showing an example of a data structure in a system characterized by the following.

The processing program according to the present embodiment exists in software of an operating system (OS). In the checkpoint module 51, a checkpoint generation 511 managed by the module 51 is provided. The checkpoint module 51 adds “1” to the checkpoint generation 511 every time the checkpoint generation 511 is checked.

The memory management module 52 has a page management structure 521, which is a logical structure of software for managing physical pages. This page management structure 52
No. 1 always exists in an operating system that implements ordinary paged virtual storage. In the page management structure 521, a checkpoint generation storage area (CPA) 522 is provided.

In the memory management module 52, a registration list 523 which stores pages set as “backup non-target areas” in the backup device 2 is provided.

The above checkpoint module 51
The memory management module 52 exists inside the OS stored in the memory 4 shown in FIG. less than,
The processing when a memory release occurs will be described with reference to FIG. 6, the processing when a memory acquisition has occurred with reference to FIG. 7, and the processing when the checkpoint is completed will be described with reference to FIG.

When a memory release request is issued (step S61 in FIG. 6), the page management structure and the point storage generation are checked, and the point generation is stored (step S62).

On the other hand, when a memory acquisition request occurs (FIG. 7)
(Step S71), the checkpoint (α) generation stored in the management structure of the memory to be allocated is compared with the current checkpoint (β) generation (Step S7).
2).

If the current checkpoint (β) generation is larger than the checkpoint (α) generation stored in the memory management structure to be allocated (Yes in step S73)
s), the process of setting information in the backup non-target area shown in the flowchart of FIG. 4 is performed (step S7).
4), information is set in the backup non-target area setting information (step S75).

If it is determined in step S73 that the current checkpoint (β) generation is smaller than the checkpoint (α) generation stored in the memory management structure to be allocated (No in step S73), the process proceeds. It ends.

When the checkpoint is completed, all the valid information in FIG. 2 of the address stored in the registration list 523 is invalidated. Subsequently, when the registration list 523 is scanned (step S81 in FIG. 8), a comparison process between the registered address and the address in FIG. 2 is performed (step S8).
2).

Here, if they match (step S83)
Yes), the valid / invalid of the entry is set to invalid (step S84), and the corresponding entry is deleted from the registration list (step S85). If they do not match (No in step S83), the process ends.

According to the above embodiment, for the memory that was free at the time of the most recent checkpoint, no backup is required from the memory acquisition request time to the next checkpoint. By setting such a memory as a non-backup target, it is possible to reduce bus transactions and improve performance.

In the following, in response to a request to secure a memory for writing from I / O to memory (generally, reading from I / O), a checkpoint generation that releases a page is always performed.
The logic of memory management that guarantees <(current checkpoint generation) will be described with reference to FIG.

That is, a recovery method based on a memory checkpoint / rollback is adopted, and the backup device is provided with a means for setting by software whether or not to back up data written from the I / O to the memory. And a checkpoint generation storage that records the checkpoint generation at the time in the software management structure of the minimum unit of memory acquisition / release at the time of a memory release request. The checkpoint generation at that time is stored in an area and a checkpoint generation storage area of a memory management structure to be released when a memory release request is issued. Point generation storage area Stored value for the system characterized by comprising means for returning a smaller memory than the checkpoint generation at that time, the logic will be described with reference to FIG. 9.

When the I / O memory acquisition (input is size) process is performed (step S91 in FIG. 9), it is checked whether a page satisfying α <β for the designated size can be secured (S step). 92).

If the specified size cannot be secured (No in S93), it is determined whether a sufficient time has elapsed since the previous checkpoint processing (S94). If the time has not passed sufficiently (No in step S94), an event waiting list process is performed (step S94).
Step 95), and return to the process of step S92. On the other hand, if a sufficient time has elapsed (Yes in step S94), checkpoint collection is performed (step S9).
6), the process returns to the step S92.

If it is determined in step S93 that the designated size has been secured (Yes in step S93).
s), the process ends. As described above, according to the above-described embodiment, for a write request from an I / O to a memory, a memory that is free at the time of the latest checkpoint is allocated, so that backup of the memory becomes unnecessary. . As a result, bus transactions can be reduced and performance can be improved.

[0033]

As described above, according to the present invention, O
By adding a function to the memory management of the S software, it is possible to avoid unnecessary backup and completely eliminate the need for backup of writing from the I / O to the memory, thereby reducing bus transactions and improving performance. Can be improved. That is, the number of backup operations for a memory space that does not need to be backed up is reduced, thereby reducing bus transactions and improving the performance.

[Brief description of the drawings]

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

FIG. 2 is an exemplary view conceptually showing information included in a backup device according to the embodiment;

FIG. 3 is a flowchart showing an operation of the backup device 2 according to FIG. 2 according to the embodiment;

FIG. 4 is an exemplary flowchart showing an operation flow of software for setting information in a non-backup target area according to the embodiment;

FIG. 5 is a diagram showing an example of a data structure according to the embodiment.

FIG. 6 is a flowchart showing a flow of processing when memory release occurs according to the embodiment;

FIG. 7 is a flowchart showing the flow of processing when memory acquisition has occurred according to the embodiment;

FIG. 8 is a flowchart showing a flow of processing at the time of completion of a checkpoint according to the embodiment;

FIG. 9 shows memory management according to the embodiment, which always guarantees (checkpoint generation that released pages) <(current checkpoint generation) in response to a request for securing memory for writing from I / O to memory. 4 is a flowchart showing the flow of the processing.

[Explanation of symbols]

 1a, 1n CPU 2 backup device 3 I / O device 4 memory 5 system bus 51 checkpoint module 52 memory management module

Claims (6)

[Claims] 1. A computer system which employs a recovery method based on a memory checkpoint / rollback and provides a backup device with means for setting a non-backup target area by software. Counter means for recording a point generation; checkpoint generation storage area means for recording a checkpoint generation at that time in a software management structure of a minimum unit of memory acquisition / release at the time of memory release request; A value stored in the checkpoint generation storage area means of the obtained memory management structure;
A comparing means for comparing the checkpoint generation at that time with the checkpoint generation; and if the value stored in the checkpoint generation storage area is smaller than the checkpoint generation at that time by the comparing means, the page is backed up. Means for returning the memory to the requester after setting it in the non-target area, and resetting the backup non-target area set by the above means in the previous checkpoint section at the time of the next checkpoint. Checkpoint rollback computer system. 2. A recovery method based on memory checkpoint / rollback is adopted, and a backup device
What is claimed is: 1. A computer system, comprising: means for setting, by software, whether or not to back up data written from / O to a memory; counter means for recording a checkpoint generation added at the time of a checkpoint; Checkpoint generation storage area means for recording the current checkpoint generation in the software management structure of the minimum unit of memory acquisition / release at the time of request, and checkpoint generation storage area of the memory management structure released at the time of the memory release request Means for storing the checkpoint generation at that time, and for allocating memory for the write request from the I / O to the memory, the value stored in the checkpoint generation storage area means Smaller than checkpoint generation at A checkpoint rollback computer system, comprising: means for returning a memory. 3. A computer system that employs a recovery method based on a memory checkpoint / rollback and provides a backup device with means for setting a non-backup target area by software. The point generation is counted and recorded. When a memory release request is made, the checkpoint generation at that time is recorded in the software management structure of the minimum unit of memory acquisition / release, and the checkpoint of the memory management structure acquired at the time of the memory acquisition request is recorded. The value stored in the generation storage area is compared with the checkpoint generation at that time, and if the value stored in the checkpoint generation storage area is smaller than the checkpoint generation at that time, Will not back up that page After that, the memory is returned to the requester, and at the time of the next checkpoint, the backup non-target area set above is reset in the previous checkpoint section. Memory management control method in 4. A recovery method based on memory checkpoint / rollback is adopted, and a backup
In a computer system that provides means for setting whether or not to back up data written from / O to a memory by software, a checkpoint generation to be added at the time of a checkpoint is counted and recorded at the time of a memory release request. Then, the current checkpoint generation is recorded in the software management structure of the minimum unit of memory acquisition / release, and at the time of the memory release request, the current checkpoint generation is stored in the checkpoint generation storage area of the released memory management structure. When the memory is allocated to the write request from the I / O to the memory, a memory whose value stored in the checkpoint generation storage area is smaller than the checkpoint generation at that time is returned. Checkpoint row characterized by Memory management control method in the LUVAC computer system. 5. A computer system which employs a recovery method based on memory checkpoint / rollback and provides a backup device with means for setting a non-backup target area by software. A function of counting and recording the point generation; a function of recording the checkpoint generation at that time in a software management structure of a minimum unit of memory acquisition / release at the time of a memory release request; and a memory management acquired at the time of the memory acquisition request. A function that compares the value stored in the checkpoint generation storage area of the structure with the checkpoint generation at that point in time. This comparison function allows the value stored in the checkpoint generation storage area to be checked at that time. If it is smaller than the point generation, After setting a page as a backup non-target area, the memory is returned to the requester, and at the next checkpoint, a function that resets the set backup non-target area in the previous checkpoint section is executed. A computer-readable storage medium storing a memory management control program in a checkpoint rollback computer system. 6. A recovery method based on memory checkpoint / rollback is adopted, and a backup device is
In a computer system which provides means for setting whether or not to back up data written from / O to a memory by software, a function of counting and recording a checkpoint generation added at a checkpoint, and releasing memory. A function to record the checkpoint generation at that time in the software management structure of the minimum unit of memory acquisition / release at the time of request; and a check at that time in the checkpoint generation storage area of the memory management structure released at the time of the memory release request. For the function of storing the point generation and the memory allocation for the write request from the I / O to the memory, the value stored in the checkpoint generation storage area is smaller than the value of the checkpoint generation at that time. Check to execute the function to return the small memory A computer-readable storage medium storing a memory management control program in a lockpoint rollback computer system.