JPH06214883A - Data saving system - Google Patents

Abstract

PURPOSE:To shorten time till the restart of processing after the generation of a fault by activating a save judging means in a fixed cycle and executing a saving means corresponding to the lapse of time or the number of times of update after the last save. CONSTITUTION:An access monitoring means 1 performs exclusive control between the saving operation of data on a non-volatile storage device 9 and the data operation and when there are updated data, a save judging means 2 decides whether the save is really performed or not corresponding to the lapse of time or the number of times of save after the last save. A time monitoring means 3 activates this save judging means 2 in the fixed cycle. A saving means 4 really copies the data on the non-volatile storage device 9 to a non-volatile storage device 12, and an update counting means 5 counts the number of times of update after the save. An invalid area judging means 6 judges valid and invalid areas in the data saved areas on the non-volatile storage device 9 at the time of generating the fault of this non-volatile storage device 9 and suppresses the amount of data to be recovered at a minimum.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a non-volatile memory device for storing data in a volatile memory device when data in the non-volatile memory device for low-speed access is used in a volatile memory device capable of high-speed access in a computer system. The method of saving data to the.

[0002]

2. Description of the Related Art In recent years, it has become possible to manufacture large-capacity storage devices made of semiconductors, and a part of data stored in a low-speed access storage device such as a magnetic disk is loaded onto the semiconductor high-speed access storage device to store data. A method for realizing a substantially high-speed reference and update has become popular. In this method, in a storage device applying magnetism represented by a magnetic disk, the stored contents are not lost due to a failure other than the storage medium itself such as the stop of the power supply. The memory content will be lost because the continuity of is not guaranteed. As an example of such a case, an operating system (OS)
In a storage system that uses a semiconductor extended storage device as a high-speed buffer of a magnetic disk storage device, if the OS stops and the continuity of the power supply of the extended storage device seen from the OS is lost, a part or the whole of the extended storage device is lost. There is a case where a failure occurs, or a power supply is stopped in an electronic disk storage system in which a storage element is composed of a semiconductor.

When such a failure occurs, since the semiconductor memory device itself has a large capacity, the number of related files or data sets is large, and it takes a lot of time to restore the data, which has a great influence on the business. Therefore, as a software countermeasure against this influence, conventionally, a method has been adopted in which the OS or the business program accumulates the update history of data in the nonvolatile storage device, and when the data is lost, the update history is used to reproduce the data. ing. On the other hand, as measures against hardware, for example, Japanese Patent Laid-Open Publication No.
No. 25456), there is a method of notifying the software that the saved data has become invalid and warning the software of re-saving as well as providing the saving means to the software. Further, as proposed in Japanese Patent Laid-Open Publication No. Sho 64-3758, there is also known a system in which a standby power supply (backup power supply device) is installed in a semiconductor memory device as a countermeasure against a power failure.

[0004]

However, in the method using the update history, since the process up to the occurrence of a failure is re-executed for the initial value of the data and the latest data is reproduced, from the occurrence of the failure to the restoration of the data. There was a problem that it took a long time. Further, the hardware function that gives the software a warning of saving does not guarantee the data on the semiconductor. Further, the method using a backup power source has a problem that the cost of the device is increased, the installation floor is expanded, and the failure including the backup power source cannot be dealt with. The present invention has been made in order to solve such a problem, does not require data reproduction processing at the time of occurrence of a failure, can shorten the time until the processing is restarted after a failure, and has a facility such as a standby power supply. The purpose is to provide a data saving method that does not require.

[0005]

According to the data saving method of the present invention, access monitoring means for judging whether or not access to the data is possible for each usage unit of the area of the volatile storage device, and the area of the volatile storage device. Evacuation determination means for deciding whether or not to save the data to the corresponding non-volatile storage device for each usage unit, time monitoring means for activating the evacuation determination means at fixed time intervals, and storage area of the volatile storage device. Save means for saving the data to the corresponding area of the non-volatile storage device for each usage unit; update counting means for storing the number of updates of the data for each usage unit of the volatile storage device since the last save; And an invalid area determining means for determining a valid area and an invalid area in the saved area of the data in the volatile storage device when a failure occurs in the local storage device.

[0006]

In the present invention, the access monitoring means performs exclusive control between the data saving operation on the volatile storage device and the data operation, and the saving determining means when updated data exists,
Whether or not the evacuation is actually performed is determined based on the elapsed time after the previous evacuation or the number of updates, and the time monitoring means activates the evacuation determination means in a constant cycle. Further, the saving means actually copies the data in the volatile storage device to the nonvolatile storage device, and the update counting means counts the number of updates after the saving. Here, the update counting means counts the number of data updates since the previous save operation, and the save determination means detects that the data on the volatile storage device is updated and the need to save again occurs with this count value. By monitoring the number of times the data has been updated, the number of save operations with respect to the number of updates is maintained constant. In addition, the invalid area determination means determines a valid area and an invalid area of the saved area of the data on the volatile storage device when a failure occurs in the volatile storage apparatus, and the invalid part of the saved data is thereby determined. It becomes possible to limit the amount of data to be recovered.

[0007]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the data saving method of the present invention. In FIG. 1, reference numeral 1 denotes an access monitoring unit for determining whether or not the data can be accessed for each usage unit of the area of the volatile storage device, and 2 corresponds to the data for each usage unit of the area of the volatile storage device. The evacuation determination means 3 for deciding whether or not to evacuate to the non-volatile storage device is a time monitoring means for activating the evacuation determination means at fixed time intervals. Further, 4 is a saving means for saving the data to the corresponding area of the non-volatile storage device for each usage unit of the area of the volatile storage device, and 5 is the last saving of the data for each usage unit of the volatile storage device. Update counting means for storing the number of times of subsequent updates and 6 are invalid area determining means for determining the valid area and the invalid area in the saved area of the data in the volatile storage device when the failure occurs.

2 and 3 are flow charts showing the processing of the access monitoring means 1. As shown in the figure, the access monitoring means 1 includes an access information input unit 11, an access management table entry number calculation unit 12, a save state determination unit 13, an access type determination unit 14, an access management entry lock unit 15,
Evacuation completion confirmation unit 16, access information input unit 17, access management table entry number calculation unit 18, update state determination unit 1
9, update count means entry number calculation unit 1A, update count value count unit 1B, access management entry unlock unit 1
It is composed of C and.

FIG. 4 is a flow chart showing the processing of the time monitoring means 3. As shown in the figure, the time monitoring means 3 includes a current time input section 31, a save determination means activation section 32, and a restart time calculation section 3
3 and a restart time registration unit 34. Further, FIG. 5 is a flow chart showing the processing of the evacuation judging means 2, and comprises an update counting means searching section 21, an all entry completion judging section 22, an update count judging section 23, and a saving means starting section 24.

FIG. 6 is a flow chart showing the processing of the saving means 4.
The saving unit 4 includes a save area information input unit 41, an access management table entry number calculation unit 42, an update state determination unit 43, an access management entry table lock unit 44, a cache memory save unit 45, an access management table entry unlock unit 46, and an update. The counting unit entry number calculating unit 47 and the update counting initializing unit 48 are included.

FIG. 7 is a diagram showing the structure of the update counting means 5. The update counting means 5 is composed of a table for storing the number of updates after the last save is performed for each storage area of the cache memory. The update counting means 5 is updated by the access monitoring means 1 and the saving means 4, and the saving determination means 2
Referenced by.

FIG. 8 is a diagram showing the configuration of the access management table 1a shown in FIG. The access management table 1a is referenced and updated by the access monitoring means 1 and the saving means 4. Further, the access management table 1a stores, for each storage area of the cache memory, an exclusive flag indicating that the data in the area is being updated or saved. FIG. 9 is a flow chart showing the processing of the invalid area judging means 6, which is the update counting means searching section 61 and the all entry completion judging section 6.
2, invalid entry determination unit 63, invalid address calculation unit 6
4, an invalid address output unit 65.

Next, the operation of the above embodiment will be described by taking the case where the cache memory 901 on the semiconductor memory device 9 of FIG. 1 is saved in the magnetic disk device 12 as an example. In this embodiment, the cache memory 901 is 1K (K = 1024).
It shall be divided into fixed length areas of bytes,
The area starting from the byte shall be saved.

First, the operating system 7 activates the time monitoring means 3 once, and by this activation, the time monitoring means 3 inputs the current time from the operating system 7 by the current time input section 31, and the evacuation determination means activation section 3
2, the evacuation determination means 2 is activated. Next, the time monitoring unit 3 calculates the time when the next restart is to be performed by the restart time calculation unit 33. In the present embodiment, assuming that the activation interval is the first known time T1, the next activation time is a value obtained by adding the activation interval T1 to the current time input by the current time input unit 31, This is registered in the operating system 7 by the restart time registration unit 34,
Wait for the next startup.

When the evacuation judging means 2 is activated, the update counting means searching section 21 fetches the entries of the update counting means 5 in order from 0. When all the entries have been searched, the all-entry completion determination unit 22 ends the process. If there is an unprocessed entry, control is passed from the all entry completion determination unit 22 to the update count determination unit 23. The update number determination unit 23 refers to the update number counted by the entry of the update counting means 5 which is being examined, and the update number after the data of the cache memory area corresponding to this entry is last saved is externally determined. It is determined whether or not the given threshold is exceeded.

If the number of updates does not exceed the threshold, it is determined that there is no need to save this time. If the number of updates exceeds the threshold, it is determined that this cache memory area is saved this time. When the evacuation determination unit 2 determines that evacuation is necessary, the evacuation unit activation unit 24 activates the evacuation unit 4. Next, the evacuation determination unit 2 executes the update count determination unit 23 or the evacuation unit activation unit 24, and then executes the update counting unit search unit 21.
Return to the next entry processing.

When the evacuation means 4 is activated, the evacuation area information input unit 41 inputs information on the position of the cache memory area to be evacuation target from the evacuation determination means 2. In the present embodiment, since the cache memory is divided into fixed lengths of 1 Kbyte, if the entry number of the corresponding update counting means 5 is input, the address of the corresponding cache memory can be found from equation (1). If the entry number of the update counting means 5 is n and the address of the cache memory is a, then a = nK (1)

Next, the saving means 4 causes the access management table entry number calculation unit 42 to calculate the entry number of the access management table 1a corresponding to the cache memory area input by the save area information input unit 42. In the present embodiment, the cache memory is divided into fixed lengths of 1 Kbyte, and the correspondence with the cache memory area of the access management table 1a has the same structure as the update counting means 5, so the update counting means 5 The entry number n becomes the access management table entry number. Next, the evacuation means 4 uses the update state determination unit 43.
By referring to the exclusion flag Fn of the access management table entry n, if the exclusion flag Fn is "updating",
The process ends. If the exclusion flag Fn is not "updating", the access management table entry lock unit 44 sets the exclusion flag Fn of the access management table entry n of the cache memory area to be saved to "saving", and the cache memory The save unit 46 requests the operating system 7 to save the 1 Kbyte area from nK to nK-1 to be saved. Further, after the evacuation is completed, the access management table entry unlocking unit 46 cancels the display of the "saving" of the exclusive flag Fn of the access management table entry n of the cache memory area to be saved, and the update counting unit entry number calculating unit 47. Then, the entry number of the update counting means 5 corresponding to the cache memory area to be saved is calculated. In this embodiment, the update counting means entry number n is obtained as an input from the evacuation judging means 2. afterwards,
The update count initialization unit 48 sets the count value of the entry n of the update counting means 5 to zero.

Next, the exclusive control of the access to the cache memory 901 for the save processing by the save means 4 and the access to the cache memory 901 by the application program 8 will be described. When the cache memory area to be saved is updated by the application program 8 during the save processing by the save means 4, the saved contents include both updated and unupdated parts by the application program, and the data is disturbed. The means for preventing this phenomenon will be described below.

When the application program 8 requests the operating system 7 to access the cache memory area, the operating system 7 starts the access monitoring means 1 from the starting point 1. When the access monitoring unit 1 is started from the starting point 1, the access information input unit 11 inputs the address of the cache memory requested for access and the access type of reference or update from the operating system 7, and the access management table entry number. The calculator 12 calculates the entry number of the access management table 1a corresponding to the cache memory area requested to be accessed. This calculation is already calculated in the description of the operation of the access management table entry number calculation unit 43 of the saving means 4, and the address nK of the cache memory corresponds to the access management table entry number n by the equation (1).

Next, the access monitoring means 1 refers to the access management table entry n by the evacuation state judging section 13,
It is determined whether the evacuation process is in progress. If the area of the address nK of the cache memory is being saved, the access management table entry locking unit 44 of the saving means 4 sets the exclusive flag Fn of the access management table entry n to "saving", so the saving is completed. The confirmation unit 16 suspends the processing for the second known constant time T2. When the saving process of the area of the cache memory at the address nK is completed, the saving means 4
The access management table entry unlocking unit 47 of the access management table entry n of the exclusion flag Fn of the access management table entry n is “saving”.
Since the display is canceled, the evacuation status determination unit 13 and the evacuation completion confirmation unit 16 can confirm the evacuation process completion by monitoring the evacuation flag Fn of the access management table entry n. The access monitoring unit 1 is the evacuation state determination unit 1
If it is confirmed that the evacuation process is not being performed by 3, the access type determination unit 14 determines the access type input by the access information input unit 11. If the access is a reference, the control is returned to the operating system 7, and the application program 8 Access is possible. If the access is an update, the access management table entry lock unit 15 sets the exclusive flag Fn of the access management table entry n to “updating”, returns control to the operating system 7, and enables access by the application program 8. To do.

Next, when the application program 8 calls the operating system 7 to return the cache memory area to the operating system 7, the operating system 7 starts the access monitoring means 1 from the starting point 2. When the access monitoring unit 1 is activated from the starting point 2, the address of the cache memory requested to be accessed by the access information input unit 17 is input from the operating system 7, and the access management table entry number calculation unit 18 requests the access. The entry number of the access management table 1a corresponding to the cache memory area is calculated. This calculation is already calculated in the description of the operation of the access management table entry number calculation unit 43 of the saving means 4, and the address nK of the cache memory corresponds to the access management table entry number n by the equation (1).

The access monitoring means 1 has an update state judging section 19
Refers to the access management table entry n, and if the exclusion flag is other than "updating", the process is terminated and control is returned to the operating system 7, and the exclusion flag is "updating".
If so, the update counting means entry number calculation unit 1A
The update counting unit entry number corresponding to the cache memory area input by the access information input unit 17 is calculated. In this calculation, if the cache memory address a is known, the update count entry number n can be calculated by the formula (2) which is a modification of the formula (1).

N = a / K (2) Further, the monitoring means 1 adds 1 to the content of the update counting means 5 by the update counting section 1B. Further, the monitoring means 1 cancels the display of "update in progress" of the access management table entry n by the access management table entry unlock unit 1C, terminates the processing, and returns the control to the operating system 7. As described above, the exclusive flag of the access management table entry enables exclusive control between the update of the corresponding cache memory area and the processing of the escape unit.

Next, the identification of the valid part and the invalid part of the data in the semiconductor memory device 9 when a failure occurs in the semiconductor memory device 9 used as a cache memory will be described. The effective portion means the semiconductor memory device 9
The above is an area that has not been updated since the last save was performed, and the invalid portion is an area that has not been saved since the last update was performed. When a failure occurs in the semiconductor memory device 9, the operating system 7 activates the invalid area determination means 6. When the invalid area determination means 6 is activated, the update counting means 5 is searched by the update counting means searching section 61.
Are sequentially taken out from 0. When all the entries have been searched, the all entry completion determination unit 62 ends the processing. If there is an unprocessed entry, control is passed from the all-entry completion determination unit 62 to the invalid entry determination unit 63. Whether the saved data of the data in the area of the cache memory corresponding to the entry is the latest is known from the update count value of the entry.

If the update count value is set to zero, the cache memory area corresponding to that entry should have not been updated since the last save, so the saved data is valid, and at this time the output is Absent. On the other hand, if the update count value is not zero, the cache memory area corresponding to that entry has been updated since the last save, so the saved data is invalid because it does not match the latest cache memory contents. The address calculation unit 64 calculates the corresponding cache memory address by the equation (1), and the invalid address output unit 65 notifies the operating system 7 of the address of the cache memory corresponding to the invalidated saved data.

[0027]

As described above, according to the data saving method of the present invention, the data in the volatile storage device is stored in the nonvolatile storage device within a certain time after the update or when the data is updated a certain number of times. Therefore, even if a failure such as a power failure occurs that the data on the volatile storage device is lost, it is highly likely that the data will be saved on the non-volatile storage device, and therefore data recovery processing by update history is unnecessary. Or the amount of data that requires data recovery processing is limited and lost, that is, the invalid part of the saved data is known and the amount of data to be restored can be limited, so The advantages are that the time can be shortened, the operating rate of the computer system can be improved, and expensive equipment such as an auxiliary power source is unnecessary.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a data saving method of the present invention.

FIG. 2 is a flowchart showing the processing of the access monitoring means of FIG.

FIG. 3 is a flowchart showing the processing of the access monitoring means of FIG.

FIG. 4 is a flowchart showing the processing of the time monitoring means of FIG.

5 is a flowchart showing the processing of the evacuation determination means in FIG.

FIG. 6 is a flowchart showing the processing of the saving means in FIG.

FIG. 7 is a diagram showing a configuration of update counting means in FIG.

FIG. 8 is a diagram showing a configuration of an access management table of FIG.

FIG. 9 is a flowchart showing the processing of the invalid area determination means in FIG.

[Explanation of symbols]

 1 access monitoring means 1a access management table 2 evacuation determination means 3 time monitoring means 4 evacuation means 5 update counting means 6 invalid area determination means 7 operating system 8 application program 9 semiconductor memory device 901 cache memory 10 main memory 11 CPU 12 magnetic disk device

Claims (1)

[Claims] 1. In a hierarchical storage management system that uses a high-speed volatile storage device as a cache memory of a low-speed nonvolatile storage device, whether or not to access the data is determined for each usage unit of the area of the volatile storage device. The access monitoring means for judging, the evacuation judging means for judging whether or not to save the data to the corresponding nonvolatile memory device for each usage unit of the area of the volatile memory device, and the evacuation judging means at fixed time intervals. A time monitoring means to be activated, a saving means for saving the data to a corresponding area of the non-volatile storage device for each usage unit of the area of the volatile storage device, and a last part of the data for each usage unit of the volatile storage device. An update counting unit that stores the number of updates after the saving, and a valid area and an invalid area of the saved area of the data in the volatile storage device when a failure occurs in the volatile storage device And an invalid area determining means for performing the data saving method.