JPS6022783B2 - Peer duplicate data control method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an efficient control method for duplicate data in a distributed database system.

Today, with the development of information processing technology, distributed database systems are being discussed.

One of these is a control method for duplicate data. Duplicate data refers to data that is managed by physically different external parties, but whose contents are the same.

The basis of the duplicate data control method is to maintain the identity of one piece of data. The problem with maintaining identity is that
Each computer that manages duplicate data performs processing independently and in parallel, and the challenge is how to keep the content of the duplicate data managed by these computers the same.

To simplify the explanation, an example in which identity collapses will be explained using FIG. 1. In FIG. 1, there are a computer 1 and a computer 2 that have duplicate data.

Also, record R is one of the duplicate data. Now, computer 1 issues an update request that sets the contents of record R to a, and computer 2 sends an update request that sets the contents of record R to b.
Assume that a data update request occurs. Execute these update requests outside of each view. At this point, record R in computer 1 is a, record R in computer 2 is b, and each computer notifies the other computers of these update requests. Each computer receives these notifications with some time delay and updates the duplicate data. That is, the computer 1 updates the code R to b, and the computer 2 updates the code R to a. In the example of FIG. 1, even if a similar update is performed, the contents of duplicate data managed by each computer will easily collapse. This is because the update order differs depending on each computer. Conventionally, a majority voting method has been proposed as a method for maintaining identity, but it is difficult to put it into practical use because it takes too long to decide whether to update.

An object of the present invention is to provide a control method for duplicate data that can determine update permission in a short time.

First, the principle of the present invention will be explained. If the data group updated at a certain timing is the same data group on each computer, identity can be easily maintained.

The present invention introduces the concept of logical time to this timing. Logical time is a logical time that is not physically the same in each computer but is logically synchronized.
The method for advancing the logical time will be explained below. After confirming that all of the other computers' logical times are greater than or equal to the computer's logical time, a computer advances the computer's logical time by one and changes the logical time to the other computers. Give notice.

In this method, the logical time of another computer is one logical time earlier than, equal to, or one logical time later than the logical time of the computer.

The first embodiment of the present invention defines a method for advancing the logical time in each computer, and includes a function to store logical time change notifications supplied from other computers, and a function to advance the logical time of all other computers. A function to advance the logical time of this computer by one when it is confirmed that the time has exceeded the logical time of this computer, and a function to notify other computers of the logical time change notification when the logical time is changed. This method is characterized by the following.

That is, each computer advances its logical time while synchronizing with the logical time of other computers.

Next, the flow of data in the present invention will be explained using FIG. 2.

Assume that an update request occurs in a certain computer (hereinafter referred to as A) when the logical time of A is k. A notifies other computers (hereinafter referred to as B to Bn, simply written as B) that an update request has occurred. Let this event be a.
B stores the update request until B's logical time reaches k+N, and executes the update request when B's logical time reaches k+N. Let this event be b. From the first embodiment of the invention, it can be proven that event a ends before event b.

If update requests are executed according to the above procedure, each computer will execute the same update request at a certain logical time, and it is possible to make the update order the same in each computer.

In the second embodiment of the present invention, an update request generated from each computer at the logical time k of that computer is stored in the computer until the logical time k+N of the computer, and updated when the logical time reaches k+N. In addition to the first embodiment, it includes a buffer that stacks update requests from N logical times ago to the present, a function to stack update requests in the buffer, and a function to stack update requests from N logical times ago to the present, and a function to stack update requests in the buffer when changing the logical time. This method is characterized by having a function of executing only the update request issued N logical times ago among the update requests.

Note that N' is an integer greater than or equal to 2, and the smaller the number, the more efficient the objective achievement is. The third embodiment of the present invention is for preventing unnecessary network 10, and in addition to the second embodiment, it is characterized by preventing updating of the logical time when there is no update request in the buffer. This is the method.

The details of the present invention will be explained using the block diagram of FIG.

First, a first example will be described. In Figure 3,
The table management device 1 is supplied with a change notification of the logical time of another computer through the terminal 2, changes the contents of the logical time table 3, and starts the logical time management device 4. The logical time table 3 is a table that stores the logical time of each computer, and has the same number of counters 3 as the number of computers.
Consists of 1 to 3·n. The logical time management device 4 is
The contents of the logical time table are checked, and when it is confirmed that the logical times of all other computers have advanced beyond the logical time of the computer, the logical time of the computer is advanced by one. It also notifies other computers of the logical time change through the terminal 5. According to the first embodiment, the logical times of each computer can be logically synchronized without the need for a privileged computer, and furthermore, since there is no need to set up a privileged computer, a highly reliable system can be realized.

The second embodiment has the following functions in addition to the first embodiment.

The logical time management device 4 activates the data update device 6 when changing the logical time. The command management device 7 is supplied with all update requests for duplicate data through a terminal 8 and records them in a buffer 9 together with their logical time. The data update device 6 is activated by the logical time management device 4 and executes the update request issued at the logical time of the computer minus N from among the update requests stored in the buffer 9. According to the second embodiment, the data group updated at each logical time becomes the same data group in each independent computer.

The third embodiment has the following functions in addition to the second embodiment.

The logical time management device 4 uses the buffer 9 when changing the logical time.
Refer to , and stop changing the logical time when there is no update request in the buffer. The third embodiment prevents the logical time from advancing when unnecessary, and reduces the number of networks 10.

Next, details of the logical time management device 4 will be explained using FIG. 3. The logical time management device 4 includes a switch circuit 41, a comparison circuit 42, and a change determination circuit 43. The switch circuit 41 connects the counters 311 to 311 of the logical time table 3.
The values of 3·n are successively supplied to the comparator circuit 42. The comparison circuit 42 compares the value supplied from the switch circuit 4 with the logical time of this computer, and determines whether the logical times of all other computers are greater than or equal to (including the same) the logical time of this computer.
If so, the change determination circuit 43 is activated. The change determination circuit 43 checks the contents of the buffer 9 and determines if the update request is 1.
If there is, it determines the advance of the logical time and activates the data update device 6 to advance the logical time of the computer by one. It also notifies other computers of changes in logical time through terminal 5.

Next, details of the data updating device 6 will be explained using FIG. 3.

The data update device 6 includes a subtracter 61, a counter 62, a coincidence circuit 63, and a gate circuit 64.

The subtracter 42 is activated by the logical time management device 4, subtracts N from the logical time value of the computer, and sets the result in the counter 62. The matching circuit 63 checks the logical time in the buffer 9 and the value of the counter 62, and when they match, notifies the gate circuit 64 that there is a match. The gate circuit 64 notifies the computer of the contents of the update request in accordance with the notification from the coincidence circuit 63. According to the present invention, the data group updated at each logical time becomes the same data group in each out-of-conview unit that is processed independently and in parallel.

This allows consistency maintenance to be handled in the same way as standalone. In the embodiments, in order to simplify the explanation, the device is described as being independent of a computer, but the device according to the present invention can be built into a computer, or can be realized by software.
They do not define the scope of the invention.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram showing an example in which identity collapses, FIG. 2 is an explanatory diagram showing the flow of data in the present invention, and FIG. 3 is a block diagram showing an embodiment of the present invention. In the figure, 1 is a table management device, 2 is an input terminal, and 3 is a table management device.
is a logical time table, 4 is a logical time management device, 5 is an output terminal, 6 is a data update device, 7 is a command management device, 8
9 represents an input terminal, 9 represents a buffer, 41 represents a switch circuit, 42 a comparison circuit, 43 a change determination circuit, 61 a subtracter, 62 a counter, 63 a match circuit, and 64 a gate circuit. Oh figure soup 2 figure Sai 3 figure

Claims (1)

[Claims] 1. Each computer in a distributed database system having duplicate data has a logical time, and each computer receives a logical time change notification supplied from another computer and outputs a change signal. A management device, a logical time table that receives change signals from the table management device and changes the logical time stored in correspondence with each computer, and logical times of all other computers stored in the logical time table. Logical time management that advances the logical time of this computer by one when it confirms that it is greater than or equal to the logical time of this computer, and notifies other computers of the logical time change notification when the logical time is changed. A redundant data control method characterized by having a device. 2. Each computer in a distributed database system having duplicate data has a logical time, and each computer has a table management device that receives a logical time change notification supplied from other computers and outputs a change signal; A logical time table that receives a change signal from the management device and changes the logical time stored corresponding to each computer, and the logical time of all other computers stored in the logical time table is the logical time of this computer. A logical time management device that advances the logical time of the current computer by one and notifies other computers of the logical time change when the logical time is changed; The present invention is characterized by having a buffer that stacks update requests for duplicate data from to the present, and a data update device that executes only update requests issued at N logical times among the update requests in the buffer when a logical time is changed. Duplicate data data control method. 3. Each computer in a distributed database system having duplicate data has a logical time, and each computer has a table management device that receives change notifications of logical time supplied from other computers and outputs a change signal; A logical time table that changes the logical time stored corresponding to each computer upon receiving a change signal from the management device, and the logical time of all other computers stored in the logical time table is greater than or equal to the logical time of this computer. A logical time management device that advances the logical time of the current computer by one and notifies other computers of the change in logical time when the logical time is changed; a buffer that stacks update requests for duplicate data from 1 to 2; a data update device that executes only update requests issued N logical times ago among update requests in the buffer when a logical time is changed; and data in the buffer. and means for preventing updating of logical time when there is no duplicate data control method.