JPS60189041A - File management system - Google Patents

Abstract

PURPOSE:To secure the continuity of data on a file by writing file update data on a buffer device, which a broken-down CPU uses, onto a file which is used in accordance with this buffer device. CONSTITUTION:If a trouble occurs in a logical CPU#1 and it breaks down, a CPU#3 in the stand-by system detects it and loads a backup program to its own main storage device 2C and executes this program. A pointer for a buffer 5A and a pointer for a file directory 17A which correspond to the faulty CPU#1 are searched from a correspondence relation management table 12. At this time, the number of the CPU#3 is changed to the number of the faulty CPU#1, and data on the buffer 5A indicated by the pointer is written on a file 4A corresponding to a file directory 16A. Thus, contents of the file 4A are equal to those of the buffer block of the buffer 5A to obtain updated data, and the continuity of data in the file 4A is secured.

Description

Detailed Description of the Invention (Field of Application of the Invention) The present invention relates to a file management method used to efficiently process data in a computer system, and particularly relates to a file management method that uses a buffer device for processing. .

More specifically, the present invention provides a central processing unit lflJ
Ichinoji. Data discontinuity that occurs at 2 stores 11} La 1
(Regarding a suitable file management method for erasing data.)

(Prior Art) File access processing in a conventional PC system uses a buffer device 16- as shown in FIG.
・Transfer of data between file 4 on auxiliary storage 119: 3 and area 6 on main storage 2, which is managed by user logsim. The second son was supposed to do this.

That is, each time a file access instruction in a user program is executed, information is transferred between file 4 and area 6 managed by the user program. .

Using this method, each time a file access command is executed, a transfer of information (i.e., file ■/0) occurs between the auxiliary storage device 1' value and the main storage device.
Also, save time on file I/O, and save time on storage! i!
Since the data transfer time is much longer than the data transfer time within 2, there is a problem in that the overhead of file access processing becomes large.

For this reason, especially in a total W machine system where the processing speed can be improved easily, a total of 1749: sufficient (high-speed sensitivity) performance as a system can be obtained, and one color can be processed in a time of 11 f,g. was occurring.

For this reason, as shown in FIG. 2, a buffer device 5 is provided on the main storage device 2, and file access processing is performed using the buffer device 11. This reduces the overhead of file access processing. A small computer system has been realized.

FIG. 3 shows an example of -f14 configuration of the t7 machine/stem using the above-mentioned buffer device. In the same figure,
The same reference numerals as in FIGS. 1 and 2 represent the same or equivalent 71b. Hereinafter, the file and buffer device will be briefly explained using the third factor.

The file 4 on the auxiliary storage device 3 consists of a plurality of blocks 71 to 7n each containing three pieces of data. Each block is a physical 115th location between the main storage device 2 and the file 4 that is accessed via the file controller 11 device 11.

The buffer device 5 transfers the data in the blocks transferred from the file 4 to the user program's management interface! buffer blocks 91 to 9m as slow working areas for transfer to the area 8 for transfer, and the buffer blocks 9, to 9
It consists of a buffer management table 10 for managing blocks stored in m.

Buffer blocks 91-9m have the same length as blocks 71-7n, which are physical access units of file 4, and store blocks read from file 4 and blocks about to be written into the file.

That is, the buffer device 5 is an auxiliary storage device +1 that stores data that cannot be stored in the main storage device 2.
Store a copy of some blocks on the file 3,
This device enables file access to blocks on the buffer device 5 without generating the file I10.

Hereinafter, file access processing on the computer system shown in FIG. 3 will be briefly explained using a specific example shown in FIG. 4.

In FIG. 4, 7 is a block included in file 4 of auxiliary storage device i'ffi 3 (in this example, 1 of A-J).
0 block), and 9 is the buffer device i 5
(3 blocks in this example) included in the buffer block.

For the file data reading macro in the user program, (1) Buffer device i! ? If the data to be accessed exists in the buffer device 5, directly from the buffer device 5 without using the file I10. (2) If the data does not exist in the buffer device 5, first , from the file 4 of the auxiliary storage device 3, copy the block 7 containing the data to /(sofa block 9
After transferring the data to the soil, transfer the data to the specified area 8 (Figure 3) managed by each Yuri 5 program (Figure 41).
Zo I (see al).

For example, when executing the following macros (1) to (4), tl) READ A ...... ...-...
・ (1)+2) READ B ・・・・・・・・・
・・・・・・ (2)+3) READ E ・・・・
・・・・・・ ・・・・・・ (3)+AI 11. E
Go to I”l R... File I10γj and the above (
For the macros 1) to (3), as shown in FIG. , 5 to block 9.

However, when the macro (4) is executed, the necessary data B is already stored in the buffer block, so the file I10 is not generated.

File/data update during programming (ampdate)
For macros, only the block on buffer block 9 is updated with the data updated by the user. At the time when the amplifier date macro ends, the block 7 in the file 4 corresponding to the updated data is not updated (see +rα in FIG. 4(b)).

That is, from the state shown in Figure 4 (11), (51UPI
)ATE A→A' (61UPDATE B-B' (71READ A' (8) UPDATFJ A'→A"), block 7 in file 4 is executed sequentially.
remains unupdated, and only block 9 of buffer device 5 is updated.

The updated buffer block 9-hi block is actually reflected on file 4 when a read macro is executed for a block without Vc on buffer block 9, or when file processing Exit macro (close/
Only when a macro) etc. is executed.

That is, from the state of Figure 4 (b), t9) REA
L) C・-[41 tin+ l circle EAD O・ f5), (6)(1
υ CLOS E ・If the macros in (7) are executed sequentially, only then will the fourth
File I10 shown in (4) to (7) in Figure (C) is generated, and block 7 of file 4 is updated.

As is clear from the above description, when the data existing on the buffer device 5 is accessed, the file I10 is not generated, so that the overhead of file access processing can be reduced.

However, as explained above, the series of processing for the buffer device 4 etc. during macro execution is performed by the operating system or file management system, and the buffer device and the processing are isolated from the user program. Therefore, problems as described below (C) occur.

That is, when the update macro ends, the user program assumes that the data in the existing IC file 4A has not been updated, even though it is in i3jkL.
Actually, at the festival, the blocks in the buffer loading device 5 have been updated, and the data in the file 4 has been updated.
The state of f, f (for example, the state of Fig. 4 (1))
occurs.

Therefore, if a failure occurs in the central processing unit 1 in this state, there will be a difference between the state of the file that the user program is writing at that time and the actual state of the file.

If such a situation occurs, it becomes possible to re-listen to the user program's processing, and when processing is restarted, the data that is recognized to have been updated by the previous file access processing is actually has not been updated on the file - that is, data continuity in the file will be lost.

Eventually, the user program will no longer be able to recognize how much of the previously executed file access processing was valid. FIG. 5 is a diagram showing an example of the difference between the file status recognized by the user program and the actual file status.

Now, after the user program executes the following macros (1) to (6), +11 READ A +2t READ I:1 [3) READ c +41 u P DATE A-A' (51U p
D A T E B -B'+61 U P D A
'r E A -A'' (7) Central processing unit failure M occurrence As shown in (7), it is assumed that a failure occurs in the central processing unit [L].

At this time, as is clear from the above explanation, in the user program, the contents of block 7 of file 4 are recognized as shown in FIG.
As shown in Figure + b), block 7 of actual file 4 is not updated at all and remains 1c, and buffer block 9
is merely updated as shown in FIG. 2(b).

By the way, the main storage device in which the buffer device 5 is located,
2. In recent years, the mainstream has been shifting from non-volatile core memory to volatile IC memory due to demands for low-cost device implementation and faster access speeds.

For this reason, the main storage device 2 using an IC memory has a problem in that all data on the buffer device 5 is lost when the main storage device is powered down.

Furthermore, in order to deal with the problems of file access processing performance and loss of data continuity on files, a buffer device is used to improve IJ-do(REAL)).
・Update (UPDATE) is one way to reduce file I10 when executing a macro. ・When executing a macro, update data is also reflected directly in file 4 at the same time - so-called ``write through j'' File access processing is also performed using this method.

However, when this write-through force method is used, there is a drawback that the effect of reducing the overhead of file access processing due to the use of a buffer device is significantly reduced.

Yet another technique is to use buffer devices and file logs (
In other words, information indicating what processing was performed on which block) is acquired, and after the central processing unit is recovered, the log information is used to return files and buffer devices to the state before the failure of the central processing unit. Method 1: A method using so-called recovery processing is also used.

However, even with this recovery method, there is an increase in processing time to obtain logs, an increase in processing time to return files and buffer devices to their original state after recovery of the central processing 8+ device, and an increase in processing time to obtain logs. Problems have arisen, such as an increase in the storage capacity of auxiliary storage devices for storage.

(Object of the invention) The present invention eliminates the drawbacks of the above conventional method and
The purpose of this invention is to provide a file management method that can guarantee the continuity of data on files when performing recovery processing from a failure in the central processing unit in a computer system that performs access processing using a buffer device. .

(Summary of the Invention) In order to achieve the above object, the present invention is arranged between the auxiliary storage device 6 and the user program 9 management area in the storage device, and is provided with the Correspondence between the buffer device functioning as a buffer area, the buffer device being used by the user program, and the file being used correspondingly (YJ and using the buffer device and file) When the central processing unit is brought up after the central processing unit goes down, (or when the central processing unit goes down) 7) Based on the contents of the correspondence management table, the central processing unit on standby (when the central processing unit backs up the By writing the above file update data onto the corresponding file that was being used, the data in the file is updated to 1? The feature is that it is configured to do so.

(Embodiment of the Invention) The present invention will be described below as a total system having a single central processing unit.
The present invention will be explained in detail using examples in which the present invention is applied to a machine system and a machine system having a plurality of central processing irons, respectively.

FIG. 7 shows an embodiment in which the present invention is applied to a computer system having a mantissa central processing unit. In this figure, the same reference numerals as in FIG. 3 represent the same or equivalent parts.

In this computer system, the buffer devices #5A, 5B, 5C on the shared main storage device 13 and the shared auxiliary storage device 16 are
Files 4A and 4B on 4 and central processing unit ii ((
As information about the correspondence relationship f+'i with (not shown), (1) the file contents of files 4A and 4B used by user programs 15A, 15B, and 35C, Pointers and (2) respective corresponding buffer devices it 5 A,
5 C and (3) the logical number #1# of the corresponding central processing unit R are retained in the correspondence management table 12 in a certain predetermined area on the shared main storage device 13. are doing.

Further, buffer units R5A, 5B, 5C and file directories 17A, 17B are held on the shared main storage device 13.

With such a configuration, as will be described later, even if a failure occurs in any central processing unit, processing for the buffer devices 58 to 5C and files 4A to 40 can be performed by any other Chuo University 3!
]! It is also 5T be from device t.

Furthermore, since the correspondence management table 12 is held at a specific address, it is possible to access the table 12 by giving that address in advance to a program that needs to access it (C).

In addition, the buffer management tables in the buffer devices 5A to 5C also have a certain structure, and all central processing bags fi
It must be accessible from t.

An example of a buffer management table suitable for this purpose is shown in FIG.

FIG. 9 shows the logic number # in the embodiment shown in FIG.
This is a flowchart showing a processing method for ensuring data continuity among four files by the standby central processing unit with logical number #3 when a failure occurs in the central processing unit of No. 1. It is 1.

In addition, in the 11th [η, when executing the above processing procedure (・
cp fire station fi'i, I A (logic #1), IC (
Logic #3), buffer soldier li'' 4.5 A J two-piece sofa block 9A and file 4A are block transitions shown in four tables.

Below is the 7th r4. Referring to FIGS. 9 and 10, this implementation 1.・Explain how to process data by 11, 4 cows, 1 mackerel, 81 tons (11iii).

Central processing using n'&& thread (for Banokag) 1
i'f: IC is another central processing unit] A, I B
The operating state of ``2'' is congratulated on hunting (step S1 in Fig. 9, A), Fushihiro in Fig. 10, 1). State 2 in Fig. 10.
Now, let's talk about the central issue of CPU #lij! I! Sozui I
A is sequentially processing (1) Rii; 'AD A t2i READ C +31 READ D +41 U l) ]) AT E D-+D'.

As a result, as is clear from the above description of FIG. 4, etc., 1c, the shared auxiliary storage device? Block 4 of t, 14
Data A, C, and D in A are written into the buffer unit (block) 9 hVCll of the shared main memory 13, and data D is replaced with D'.

Suppose that in this state, the central processing unit IA develops a failure H and goes down.

The central processing unit IC, which is a standby system,
When detecting the failure of the bank amplifier (step S2 in FIG. 9), the bank amplifier program is loaded into its own main storage device 2C and this step 1 is executed.

In this bank amplifier program, first, from the correspondence management table 12, the pointer to the buffer device 5A and the file bref corresponding to the central processing unit IAO logic jI41 number #1 that caused the damage. )'J1
Discover the pointer to 7A (9th chart step S4)
.

At this time, the central processing unit IC transfers its own logical number #3 to the central processing unit I in which the failure has occurred.
i #1 (step 83 in FIG. 9).

Next, following the pointer (address) pointed in the previous step S4, the data (buffer lock 9A) of the buffer device 5AJ pointed by the pointer is transferred to the file directory 16A. File 4
It is set aside on A (step S5 in FIG. 9).

As a result, as shown in state 3 of FIG. 10, the contents of file 4A become the same as buffer block 9A of buffer device 5A in state 2, and updated data is obtained.

The above procedure ensures the continuity of data in the file 4A.

Furthermore, the central processing unit IC, the central processing unit, r
! +! The data obtained during 'tt31 viewing to device I and the backup of the program 15A that was being executed by central processing unit IA are stored in its own main storage device i? , 2 C upper l? - Read (step S6 in FIG. 9).

In this way, the central processing unit IC, which has been in a standby state, continues the processing of the central processing unit i1A as an online system and performs online processing.

On the other hand, after the downed central processing unit IA recovers from the failure,
Work as a standby system.

As is clear from the above explanation, according to this embodiment, there is no need to perform processing such as log acquisition during normal processing, so the overhead of file access processing increases and the memory capacity increases. It is possible to guarantee data continuity in a file without having to do so.

Furthermore, backup processing by a standby central processing unit for a downed central processing unit can be achieved relatively easily compared to conventional recovery processing, thereby minimizing the downtime of the computer system.

Furthermore, since high-speed backup by the standby central processing unit of a down central processing unit is possible, it is possible to minimize losses due to file restoration processing caused by data discontinuity.

FIG. 6 is a functional block diagram showing another embodiment in which the present invention is applied to a computer system having an early central processing unit. In this figure, the same reference numerals as in FIG. 7 represent the same or equivalent parts.

In this maintenance system, the buffer devices R5A and 5B on the storage device 2 and the files 4A and 4 on the auxiliary storage device 3 are used.
As correspondence relationship information with B and 4C, the files 4A and 4C used in the user program 15 are
Directory 17A.

The correspondence management table 12 holds a pointer to 17B and pointers to the corresponding buffer devices i'l 5 A and 5 B, respectively.

If a failure occurs in the central processing unit, it becomes impossible to access the file control blocks 16A,', 16B-\, and so on.

The data in 5B-hi is transferred to files 48-4C.
It becomes impossible to write upwards.

As a countermeasure for this, in this embodiment, a program that executes the processing described below is added in advance to the program that is executed when the central processing unit starts up the ancient texts.

That is, when retrieving old texts from the central processing unit, the correspondence management table 12 is referenced to obtain the pointers of the buffer device 5A and file directory 17A, and the data on the buffer block in the buffer device tft5A indicated by the pointer is transferred to a file.・On file 4A corresponding to directory 17 A+? : Become disgusted.

Furthermore, the correspondence management table 12 is sequentially checked and the same processing as above is performed.

In addition, the processing procedure in this case is shown in Fig. 7 and Fig. 91-
This is the same as the explanation given in Figure 5, Diagnosis 4110, and can be easily inferred from this, so a detailed explanation will be omitted.

File access processing is performed by the above processing.]
Data continuity can be guaranteed for all files that have been previously stored.

As is clear from the above explanation, this implementation iFI Ic
This is because there is no need to process 1 blow and 9 times when processing a message, so the overhead of file access processing increases and the storage capacity increases 4y,゛,if,i. 5N 'i of the data in the file
It is possible to guarantee 'l'.

In addition, the processing when the central processing unit is raised can be realized relatively easily compared to conventional recovery processing, so that downtime of the computer system can be minimized.

In addition, as an application of each of the above-mentioned embodiments, C1 Correspondence in the above two embodiments ・i7 [Table 12 and buffer device 5 people, 5B, 5C are stored in the main storage device 2 or shared main storage device It13 If we decide to realize this area with non-volatile memory elements such as core memory,
Even when the power of the (shared) main storage device is turned off, the continuity of data in files can be guaranteed using the same method as shown in the embodiment.

In the above, the present invention has been explained in detail using a specific embodiment. Even for systems that provide a buffer area on a storage device that is faster than the system and perform processing using it, it is necessary to guarantee continuity of output data when restarting after a processing device failure occurs. It can be easily inferred that the present invention can be applied to ICs in such cases.

(Effects of the Invention) As explained above, according to the present invention, by using a simple device and method, it is possible to guarantee the continuity of data in a file even when the central processing unit goes down. , it is possible to improve the operating rate and reliability of a computer system that performs file access processing.

[Brief explanation of the drawing]

Figure 11 is a schematic block diagram of a computer system that performs file access processing without using a buffer, Figure 2 is a schematic block diagram of a computer system that performs file access processing using a buffer, and Figure 3 is a schematic block diagram of a computer system that performs file access processing without using a buffer. 4 is a diagram for explaining the file access example in FIG. 3, and FIG. 5 is a diagram for explaining the file access example in FIG. 3. Figure 6 is a diagram for explaining the difference between the file status and the actual block contents. The figure shows multiple central processing units [
FIG. 8 is a functional block diagram of another embodiment of the present invention necessary for a computer system with 44 buffer management tables.
9 shows the processing procedure for guaranteeing data continuity in FIG. 7; It is a diagram showing the state transition of the 'ηiso system. 1. Central processing unit (including IA, etc., hereinafter referred to as the same), 2.
・・Main memory・ 3・Auxiliary storage (hk, 4・・
・File・ 5°'/9 tsphag value, 6-・・User program′#physical area・ 7・・Block, 8
...A certain area in the user log 7 area, 9
・Buffer Bron'/, 10 ・Buffer management table, 1] ・File lul + control device, 12 ・・
Correspondence management table, 13... shared main storage device,
14... Shared main storage device location, 15... User program, 16... File control block, 17...
File/Directory Ward 6 ← 2/6 Figure 12 Shi・1 House 7 Figure 4 8 Figure 0 (

Claims (10)

[Claims] (1) A main storage device that is directly accessed by the central processing unit, an auxiliary storage device that includes a plurality of files and has a slower access speed than the main storage device, and a user storage device in the auxiliary storage device and the main storage device. Correspondence between a buffer device placed between the GLODARAM management areas and functioning as a buffer area during data transfer between the two, the buffer device in use by the user program, and the corresponding file in use; and a correspondence management table for storing relationships, and when the central processing unit is restarted after going down, file update data on the currently used buffer device is updated based on the contents of the correspondence management table. , update the data in the file by writing to the file that was being used correspondingly (a file management method with characteristic l-9). (2) The file management method described in the patent, characterized in that the correspondence management table is located on the main storage device. (3) The file management system according to claim 1 or 2, wherein the correspondence management table is a nonvolatile memory. (4) The file management system according to any one of claims 1 to 3, wherein the buffer device is arranged 11 on the main storage device. (5) The file management system according to any one of claims 1 to 4, wherein the buffer device is a nonvolatile memory. (6) A mantissa central processing unit including a standby thread, a main memory provided corresponding to each of the central processing units and directly accessed by the central processing unit, and a shared main storage unit accessed by each of the central processing units. between a storage device, a shared auxiliary storage device that includes a plurality of files and has a slow access speed compared to the main storage device, and a user program management area in the device that stores the shared auxiliary storage device and the 1λ. a buffer device that is arranged and functions as a buffer area during data transfer between the two, a buffer device that is being used by the 21-2 program and the files that are being used correspondingly; It is equipped with a correspondence management table that stores the correspondence between the buffer device and the central processing units using the file, and detects that one of the central processing units is down, and When backing up this, based on the contents of the correspondence management table, the file update data on the buffer device that was being used by the downed central processing unit is transferred to the file that was being used correspondingly. File 9 characterized in that the data of the file is updated by writing it onto the file.
Question method. (7) The file management system according to claim 6, wherein the correspondence management table is arranged on a shared main storage device. (8) The file management method described in the patent or item 7, characterized in that the correspondence management table is a non-volatile memory. (9) The file management system 6' according to any one of claims 6 to 8, characterized in that the buffer device is arranged above the shared main memory. (10) The file management system according to any one of claims 6 to 9, wherein the buffer device is a nonvolatile memory.