JPH05265985A - Data managing system - Google Patents

Abstract

PURPOSE:To prevent a deadlock state or the breakdown of data by enabling a central processing unit to know the fault of a distributed processor. CONSTITUTION:The fault of the distributed processor 100 in the course of the update processing of the data is detected by a fault detecting means 13, and in the case that the fault occurs, the lock of the data which the distributed processor 100 is being processing, is canceled by a fault processing means 14, and an update invalidating means invalidates update information from the distributed processor 100 in which the fault has occurred.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data management system in a distributed processing system.

[0002]

2. Description of the Related Art As a conventional data management system, for example, there is one disclosed in Japanese Unexamined Patent Publication No. 3-252756. This will be described with reference to FIGS. 3 and 4. FIG. 3 is a block diagram showing a distributed processing system in a conventional data management system, and FIG. 4 is an explanatory diagram showing an example of each management information.

In FIG. 3, 1 is a central processing unit, 2 is a main storage unit, 3 is block management information, 4 is an update data management file, 5 is a communication line, 100 and 100a are distributed processing units, and 101 and 101a are updates. This is the file to be processed.

Next, the operation will be described with reference to FIGS. 3 and 4. First, the distributed processing device 100 to the central processing device 1
It is assumed that a reference / update request targeting the records from record number 301 to record number 500 is notified. The central processing unit 1 searches for a block including the corresponding record in the block management information 3 shown in FIG. 4, and sets the processing flags of the corresponding blocks 34 and 35 to ON.

Therefore, a block in which this processing flag is set to ON indicates that the records in the range of the block are subject to reference / update by any of the distributed processing devices. When there is a reference / update processing request from another distributed processing device X for a block whose processing flag is set to ON, the central processing device 1 notifies the distributed processing device X that the request cannot be accepted. ..

Next, the central processing unit 1 is the distributed processing unit 100.
The blocks 34 and 35 that have been requested by the system notify that the reference / update processing is possible.
00 can freely access the data in the range from the record number 301 to the record number 500 of the file 101.

At this time, in the update processing, the contents of the file 101 are directly updated, and at the same time, the updated data is transferred to the central processing unit 1. The transferred update data is stored in the update data management file 4 of the central processing unit 1. However,
When stored, a pair of the update data and the recognition state flag set to the initial state is stored as the update record 41 as shown in the update data file 4 of FIG.

When this recognition state flag is set to ON, it means that the update data of the update data management record is not reflected in the file of the corresponding distributed processing device. Therefore, in the initial state of the recognition state flag, only the part corresponding to the distributed processing device that has actually performed the update processing is off, and the other parts corresponding to the distributed processing device are set to on.

When the series of reference / update processing in the distributed processing apparatus 100 is completed, the distributed processing apparatus 100 notifies the central processing unit 1 of the end. The central processing unit 1 that has received the end notification sets the processing flags of the blocks 34 and 35, which were the target of the reference / update processing by the distributed processing apparatus 100, to OFF and cancels the blocks.

When a reference / update processing request is issued from the distributed processing apparatus 100a after that, the central processing unit 1 checks the block management information to see if the block to be processed is available. Records 41, 42, and 4 in which the part corresponding to the distributed processing apparatus 100a (corresponding to the terminal A in FIG. 4) of the recognition state flag in the update data management file 4 is set to ON together with the notification that the processing is permitted.
The update data of No. 5 is transferred to the distributed processing device 100a. The distributed processing device 100a receiving the update data reflects the updated data in the file 100a, and then accepts the permitted reference / update processing.

When the central processing unit 1 transfers the update data to the distributed processing unit 100a, the part corresponding to the distributed processing unit 100a of the recognition state flag is set to OFF in the corresponding update data management records 41, 42 and 45. To do. Further, by turning off the part corresponding to the distributed processing device 100a, all the recognition state flags are turned off.
Is released from.

[0012]

Since the conventional data management system operates as described above, when a distributed processing device which is permitted to process a certain data goes down due to some failure in the middle of the process, its lock is locked. There was a problem that it could not be canceled and no one could access the data.

There is also a problem that when a machine that has been down due to some failure is restarted, the data destroyed by the failure is sent as update data.

The present invention has been made to solve the above problems, and it is possible to release the deadlock caused by a failure during the data reference / update operation and prevent the update data from being destroyed. The purpose is to achieve it.

[0015]

A data management system according to the present invention comprises a time counter for measuring a fixed time, a failure detecting means for detecting whether or not a failure has occurred, and a processing for processing the failure. There is provided a failure processing means for performing the above, and an update invalidating means for invalidating the update information from the failed machine.

[0016]

According to the present invention, by counting the time from the update work and automatically checking whether or not a failure has occurred, the failure of the distributed processing device is detected, the lock is released, and the update is invalidated. It is possible to do.

[0017]

FIG. 1 is an explanatory diagram showing a data management system according to an embodiment of the present invention. In FIG. 1, 1 is a central processing unit, 2 is a main memory, 3 is block management information, 4 is an update data management file, 5 is a communication line, 10 is a time counter, 11 is a block exclusive control means, and 12 is update management. Means, 13 is failure detection means, 14 is failure processing means, 15
Is update invalidation means, 100 and 100a are distributed processing devices, 1
01 and 101a are files to be updated.

Next, the operation will be described in detail with reference to the flowchart of FIG. First, a file update request is received from the distributed processing device, for example, the distributed processing device 100 (step 201). When receiving the data reference / update processing request, the block exclusive control means 11 checks the block management information 3 to see if the block is locked (step 202). If it is locked, the process proceeds to step 214.

If the data is not locked, the block exclusive control means turns on the processing flag of the block management information 3 (step 203). The update management means 12 checks the update data management file 4 and transfers the update information regarding the file to the distributed processing device 100 (step 204).

The fault detecting means 13 sets a time counter for the block (step 205). It is checked whether update data has been transferred to the block (step 206). If not transferred, the process proceeds to step 208. When the update data is transferred, the update management means 12 stores the information in the update data management file 4 (step 20).
7).

It is checked whether or not the end notice of the data reference / update processing has been transferred (step 208). If the end notification has been transferred, the process moves to step 211.
If the end notification has not been transferred, the process starts and it is checked whether a fixed time has passed (step 2).
09). If the fixed time has not passed, step 20
Return to 6.

If the fixed time has elapsed, the failure detection processing means 13 checks whether or not the distributed processing apparatus 100 is down due to a failure (step 210). If it is normal, the process returns to step 205 and the time counter is set again. If the distributed processing device 100 is down, the failure processing means 14 turns off the processing flag of the block management information 3 (step 211) and deletes the update information sent up to that point.

After that, it is checked whether or not the update data is sent from the distributed processing device 100 (step 21).
2). Step 21 if the update data is not sent
Go to 4. When the update data is sent, the update is invalidated, and the update invalidation unit 15 notifies the distributed processing device 100 of the invalidation of the update (step 213).

If the update data is not sent, the update management means 12 checks whether all the recognition status flags of the update data management file 4 are off (step 214).
If not all off, return to step 201 and wait for the request. If all are off, the update management means 12 erases the update information from the update data management file 4 (step 215).

The request from the distributed processing device 100 is a file update request, but if the block is locked, the block exclusive control means 11 issues a request invalidation notification (step 216).

[0026]

As described above, according to the present invention, even if the distributed processing device is down or destroyed due to some trouble during the data reference / update process, the other distributed processing device will not Since the deadlock state that the reference and update processing of can not be performed is automatically canceled and the data is prevented from being destroyed, the reliability and the real-time property of the system can be improved.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a data management system according to an embodiment of the present invention.

FIG. 2 is a flowchart showing the operation of the embodiment of the present invention.

FIG. 3 is a configuration diagram showing a distributed processing system in a conventional data management system.

4 is an explanatory diagram showing an example of each management information in the system of FIG.

[Explanation of symbols]

 1 Central Processing Unit 2 Main Storage Device 3 Block Management Information 4 Update Data Management File 5, 5a Communication Line 10 Time Counter 11 Block Exclusive Control Means 12 Update Management Means 13 Failure Detection Means 14 Failure Handling Means 15 Update Invalidation Means 100, 100a Distributed Processing device 101, 101a file

Claims (1)

[Claims] 1. A distributed processing system in which a central processing unit and a plurality of distributed processing units are connected by a communication line, and files having the same data are distributedly arranged in each distributed processing unit. Update management that guarantees the consistency of the data of each file by centrally managing the stored update data management file and the update data management file by the central processing unit and referring to the recognition status flag of the update data management file. Means, block management information that stores the current reference / update processing status of each distributed processing device, and block management information that centrally manages the block management information by the central processing device and guarantees the exclusivity of the process for the data of each file. A fixed time is counted from the time when the control means and the distributed processing device start reference / update processing of data of each file. And a failure detection means for confirming whether or not the distributed processing device performing the reference / update processing of the data is operating properly when the fixed time has been notified from the time counter, As a result of the failure detection means, when a failure occurrence is detected in the distributed processing device, the lock of the data is released, the reference / update process of the other distributed processing device is permitted, and the distributed processing device operates normally. If it is, the value of the time counter is reset and the failure processing means for restarting the counting of the fixed time and the update information transferred from the distributed processing device in which the failure occurred during the data reference / update processing are invalidated, A data management system characterized by comprising an update invalidating means for notifying the invalidation of the update.