JPH04248631A - Data processor - Google Patents

Abstract

PURPOSE:To perform majority operation of high reliability even in the condition that replacement, repair, or the like of arithmetic units is limited. CONSTITUTION:A comparing device 2 classifies operation results to groups having the same values. A weighting means 3 totalizes weighted values indicating reliabilities of respective arithmetic units in each group. A majority arithmetic unit 4 judges the operation result of the group having the largest weighted value to be accurate and outputs it. Weighted values of respective arithmetic units are properly updated based on past actual results or the like of arithmetic units.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data processing device, and more particularly to a data processing device that performs majority logic operations used in FTC (fault tolerant computers, hereinafter referred to as FTC). In recent years, demands for reliability in computer processing have become extremely large. Especially in areas where the FTC is required,
Reliability under continuous time operation is strictly required. Normally, in banking operations where FTC is used, the entire system continues to operate even if an error occurs, and the parts where errors occur can be replaced or repaired relatively quickly and completely. It will be done.

[0002] However, the fields in which FTC is used are not limited to these, and can be applied to, for example, planetary exploration machines, deep-sea working machines, and other situations where manual replacement or repair is almost impossible. The present invention provides a data processing device that performs more reliable processing even under such circumstances where device replacement, repair, etc. are restricted.

0003

2. Description of the Related Art In conventional data processing by the FTC, a processing method based on majority logic has often been used. This was a method of determining the truth or falsity of data provided by a plurality of arithmetic units by majority vote. For example, the calculation results of three calculation devices for one problem are compared, and if all three calculation results match, the result is determined to be correct. Also, if the two operation results match, the remaining 1
If they are different, the matching calculation result is determined to be the correct answer, and the other calculation results are determined to be errors.

As a result, the reliability of the system has been ensured by disconnecting the arithmetic unit that has caused an error and replacing or repairing it while the system as a whole continues to process.

[0005]

[Problems to be Solved by the Invention] In the above-mentioned processing method based on majority logic, an arithmetic unit that has caused an error is replaced or repaired by a human being with a spare normal arithmetic unit. It was assumed that the two had approximately the same level of reliability, and that the number of arithmetic units that would cause errors would be extremely small. However, under special circumstances, such as when FTC is used in planetary exploration equipment,
It is nearly impossible to replace or repair a normal device, and the number of spare computing devices is also limited.

[0006] Therefore, if an arithmetic unit that has caused an error is no longer used, the life of the system is greatly limited by the number of spare arithmetic units. On the other hand, if the arithmetic device that caused the error is used in the conventional manner,
The reliability of the system will be extremely reduced. Furthermore, because the conventional method simply takes a majority vote on the calculation results of each calculation device, many unreliable calculation devices are used, such as when a calculation device that has caused an error is used as is. In such a case, there is a possibility that the erroneous calculation results coincide with each other and the correct calculation result is determined to be an error, which poses a major problem beyond reliability.

An object of the present invention is to provide a data processing device that performs highly reliable majority logic operations while effectively using a limited number of arithmetic units even under the above-mentioned limited circumstances. It is in.

[0008]

[Means for Solving the Problems] In order to solve the above problems, the invention according to claim 1, as shown in FIG. a comparing device 2 that compares the calculation results of the calculating device and divides them into groups; and a weighting device 3 that calculates a weighting value for each group divided by the comparing device. ,
It is configured to have.

Further, the invention according to claims 2, 3, 4, and 5 provides a data processing system having a plurality of arithmetic units 11 to 15 and a majority arithmetic unit 4 that takes a majority vote of each arithmetic result, as shown in FIG. In the processing device, a comparison device 2 that compares the calculation results of the calculation devices and divides them into groups, a weighting means 3 that calculates a weighting value for each group divided by the comparison device, and weighting of each calculation device. The weighting value updating means 5 is configured to update the weighting value as appropriate.

0010

According to the first aspect of the invention, the comparison device 2 groups the calculation results output from each calculation device according to their values, and outputs the calculation results belonging to each group. The weighting means 3 calculates a weighting value of the calculation device corresponding to the calculation result of each group for each group, and outputs a total weighting value of the entire group. The majority calculation device 4 selects the group with the largest total weighting value for each group, determines the calculation result belonging to that group as correct, and outputs it.

Further, according to the invention as set forth in claim 2, 3, 4, or 5, the weighting value of each arithmetic unit calculated by the weighting means 3 is appropriately updated by the weighting value updating means 5 based on past performance. Updated.

0012

Embodiments Next, embodiments of the present invention will be described based on the drawings. [i] First Embodiment FIG. 1 shows a first embodiment of the present invention. First, the configuration of this embodiment will be explained.

In FIG. 1, five arithmetic units 11 to 1
5 is used. Each arithmetic device has the same structure, performs arithmetic processing on the same given task, and outputs each arithmetic result to the next-stage comparison device 2. here,
The calculation results of each calculation device are expressed as Result 1 to Result 5, respectively. Comparison device 2 compares input results 1 to 5,
Classify into groups of items with the same result value. In other words, the same operation results belong to the same group, and the number of groups is the number of types of operation results. The comparison device 2 outputs the calculation result value of each group and information on the calculation device that outputs the result value to the weighting means 3.

The weighting means 3 has a weighting value storage device for storing weighting values corresponding to each arithmetic unit, and calculates the sum of the weighting values for each group based on the information of the arithmetic units given by the comparator 2. Calculate. The weighting means 3 outputs the calculation result of each group and the total weighting value of the group to the majority calculation device 4 at the next stage.

The majority calculation device 4 compares the total weighting values of each group, selects the group with the largest weighting value, and outputs the calculation result of that group as the correct value. Next, the operation of this embodiment will be specifically explained based on FIG. 3. Here, each arithmetic device 11
Let us consider the case where the calculation results shown in FIG. The comparison device 2 compares and classifies the calculation results 1 to 5 provided by the calculation devices 11 to 15. In other words,
Arithmetic devices 1, 2, and 4 that output arithmetic result A belong to group 1, arithmetic device 3 that outputs arithmetic result B belong to group 2, and arithmetic device 5 that outputs arithmetic result C belong to group 3. It turns out. Therefore, comparison device 2
is the calculation result A and the calculation devices 1, 2, belonging to group 1,
4 as the information of group 1, the next stage weighting means 3
Output to. Similarly, the calculation result B and the calculation device 3 are output to the weighting means 3 as information of group 2, and the calculation result C and calculation device 5 are output to the weighting means 3 as information of group 3.

The weighting means 3 at the next stage calculates a weighting value for each group based on the information for each group outputted by the comparison device 2. That is, for group 1, based on the given information of arithmetic units 1, 2, and 4, the corresponding weighting values are read from the weighting value storage device, and the total value is calculated, and this value is set as the total weighting value of group 1. Similarly, the total weighting value of groups 2, 3 is calculated based on the information of groups 2, 3 provided by comparison device 2. In this example, the total weighting value of each group is the value shown in FIG. Since the weighting values stored in the weighting value storage device indicate the inherent reliability of each calculation device, the total weighting value of each group indicates the reliability of the calculation results of that group. The total weighting value of each group calculated in this way is sent to the next stage majority calculation device 4 together with the calculation results of that group.

The majority calculation device 4 outputs the calculation result value of the group with the largest total weighting value as the correct answer, based on the fact that the weighting value of each group indicates the reliability of the calculation result value. In this example, as shown in FIG. 3, since the total weighting value of group 1 is the largest, calculation result A, which is the calculation result of group 1, is determined to be the correct answer. On the other hand, calculation results B and C for groups 2 and 3 are both determined to be errors.

Next, the operation of this embodiment in another case will be explained based on FIG. 4. In this case as well, calculation processing is performed in the same way as in the case of FIG. 3, and calculation result A determined to be correct is output. However, in the case shown in FIG. 4, the results obtained by the method according to the first embodiment of the present invention are different from the results obtained by the conventional method using simple majority logic. In other words, in the conventional method using simple majority logic, the calculation result B
is determined to be correct. In this way, when the number of arithmetic devices that output error calculation results is greater, there is a risk that the error calculation results may actually be determined to be correct. In contrast, in the method according to the first embodiment of the present invention, the correct calculation result A is determined to be the correct answer. Therefore,
According to the present invention, even if the reliability of each arithmetic device is low and there are more arithmetic devices that output error arithmetic results as in the case of FIG. 4, the individual reliability of each arithmetic device is taken into consideration. , the correct calculation result can be selected.

Although the present embodiment has been described with reference to five arithmetic units, the number of arithmetic units is not limited to this number. Further, although the weighting value storage device for storing the weighting value of each arithmetic device is configured to be included inside the weighting means 3, it may be provided separately from the weighting means 3, or it may be provided individually in each arithmetic device. It's okay. Therefore, these are
Any of these can be determined depending on the system to which the present invention is applied.

In addition, in the weighting means 3 of this embodiment, the weighting value of each group is calculated by simple addition of the weighting values of each arithmetic unit, but depending on the characteristics of the system, for example, the weighting value may be It may be configured to be calculated using other methods such as using a coefficient and multiplying the calculation result. [ii] Second Embodiment FIG. 2 shows a second embodiment of the present invention. In this example, the first
In addition to the configuration of the embodiment, an example will be disclosed in which a weighting value updating means 5 for updating the weighting value of each arithmetic unit is provided.

Next, the operation of this embodiment will be explained. In the parts of this embodiment that are configured in the same way as the first embodiment, majority calculation is performed in consideration of the weighting values as in the first embodiment. Here, in this embodiment, a weighting value updating means 5 is provided outside the weighting means 3 for updating the weighting value of each arithmetic unit based on past performance. Therefore, the weighting value of each arithmetic unit is updated as appropriate based on the information output by the weighting value updating means 5, so that majority arithmetic processing is performed that is more suitable for the current operation of the system.

Next, as a method for updating weighting values, a method using a so-called bathtub curve will be explained. Generally, the reliability of a system is expressed by the failure rate shown by a bathtub curve as shown in FIG. That is, system failures often occur in the early stages of system operation. Thereafter, as the operation of the system approaches a stable state to some extent, the failure rate gradually decreases, and a stable operation period with a low failure rate continues for a certain period of time. After stable operation for a certain period of time, the system reaches the end of its service life, and the failure rate begins to increase again. Therefore, in response to such temporal changes in the failure rate of the system, the weighting value updating means 5 is configured to reduce the weighting value at the initial stage of system operation and during the life period, and increase the weighting value during the stable period. By controlling it, it becomes possible to perform majority calculations that suit the operating conditions of the system, improving the accuracy and reliability of the calculations.

Next, as another method of updating the weighting values,
A method based on the actual occurrence frequency of errors, as set forth in claims 4 and 5, will be explained. According to the majority arithmetic processing of the present invention, only one arithmetic result out of a plurality of arithmetic results output from each arithmetic device is determined to be correct, and all the remaining arithmetic results are determined to be errors. Therefore, since there is a high possibility that the arithmetic device that outputs the arithmetic result determined to be an error contains some kind of intermittent failure or the like, the weighting value is decreased. On the other hand, since a calculation device that has not outputted an error calculation result for a certain period of time is less likely to have an internal failure or the like and is considered to have high reliability, the weighting value is increased. In this way, by appropriately increasing or decreasing the weighting value based on the determination result of the majority arithmetic device 4, the arithmetic results of a highly reliable arithmetic device are given a large proportion, and the arithmetic results of a less reliable arithmetic device are given a small percentage. Since this method is used for majority calculation, it is possible to improve the accuracy and reliability of the calculation.

In this embodiment, the weighting value updating means 5
Although it is provided separately from the weighting means 3, other configurations may be used, such as providing it inside the weighting means 3, for example.

[0025]

Effects of the Invention According to the present invention, in a data processing device based on majority logic, the majority decision operation is performed in consideration of the weighted value indicating the reliability of each arithmetic unit, so that the reliability of each arithmetic unit is constant. However, even in a situation where the number of arithmetic units is limited, it is possible to provide a data processing device that performs highly reliable majority logic operations.

[0026] Furthermore, in the present invention, the weighting values of each arithmetic unit are updated as appropriate based on past results, etc., so that the precision of the arithmetic operation can be improved by performing majority arithmetic that is suitable for the actual operating situation of the system. can be further improved.

[Brief explanation of the drawing]

FIG. 1 is a diagram illustrating the principle of the invention according to claim 1.

FIG. 2 is a diagram illustrating the principle of the invention according to claims 2, 3, 4, and 5.

FIG. 3 is an explanatory diagram of a specific operation of the invention according to claim 1.

FIG. 4 is an explanatory diagram of another specific operation of the invention according to claim 1.

FIG. 5 is a diagram showing a bathtub curve according to the invention according to claim 3.

[Explanation of symbols]

11-15... Arithmetic device 2... Comparison device 3...Weighting means 4...Majority calculation device 5...Weighting value updating means

Claims (5)

[Claims] 1. A data processing device comprising a plurality of arithmetic units and a majority arithmetic unit that compares the arithmetic results of each arithmetic unit and calculates majority logic, wherein the majority arithmetic unit is configured to perform weighting that is assigned to each arithmetic unit. A data processing device characterized in that it is configured to perform majority calculation according to a value. 2. The data processing device according to claim 1, wherein the weighting value assigned to each arithmetic device is updated as appropriate based on the past performance of each arithmetic device. 3. The data processing device according to claim 1, wherein the weighting value assigned to each arithmetic device is assigned based on a failure rate value of a bathtub curve indicating reliability of each arithmetic device. . 4. The data processing apparatus according to claim 1, wherein among the arithmetic units, the weighting value assigned to the arithmetic unit that causes an error in the arithmetic operation is reduced as appropriate. 5. Among the arithmetic units, if the normal operating time exceeds a certain period of time, the assigned weighting value is increased as appropriate. data processing equipment.