JPH07121488A - Periodical diagnosing system of multiprocessor system - Google Patents

Abstract

PURPOSE:To provide the periodical diagnosing system of a multiprocessor system whereby the deterioration of processing ability in a master processor is prevented in a multiprocessor system. CONSTITUTION:The master processor 101 group-divides slave processors 102,... 106 connected on a bus 107 into two units 108 and 109. The master processor 101 executes only the periodical diagnosis of a health check master 102 and the periodical diagnosis of the slave processors 103 and 1011 inside the unit 108 is executed by the slave processor 102. The master processor 101 sets information of the slave processor to be diagnosed by the slave processor 102 in a periodical diagnosis indication signal so as to report it. When the abnormality of the slave processors 103 and 104 inside the same unit 108 is detected, the slave processor 102 reports it to the master processor 101. The periodical diagnosis of the slave processors 105 and 106 inside the unit 109 is executed in the same way.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a regular diagnosis system for a multiprocessor system, and more particularly to a multiprocessor system in which a master processor is connected to a plurality of slave processors through a bus and which prevents a decrease in the processing capacity of the master processor. Regarding regular diagnosis method.

[0002]

2. Description of the Related Art In a conventional periodic diagnosis system for a multiprocessor system of this type, all slave processors under its control are regularly diagnosed based on data internally managed by a master processor.

FIG. 4 is a block diagram showing an example of a conventional periodic diagnosis system for a multiprocessor system.

Referring to FIG. 4, the multiprocessor system of this embodiment has a plurality of slave processors 402, 40 connected to a master processor 401 via a bus 407.
3, 404, 405 and 406.

The master processor 401 individually transmits a periodic diagnosis instruction signal to each slave processor 402, ..., 406 through a bus 407, and receives a periodic diagnosis response signal from each slave processor 402 ,. is doing.

[0006]

However, in the conventional periodic diagnosis method for a multiprocessor system, the master processor issues a regular diagnosis instruction to all slave processors and receives regular diagnostic responses from all slave processors. Therefore, there is a problem that the processing capability of the master processor is lowered and the handling is delayed as a multiprocessor system.

An object of the present invention is to divide a plurality of slave processors into groups in a multiprocessor system, to provide a health check master in each unit, and the master processor performs periodic diagnosis of only the health check master, thereby processing the health check master. It is an object of the present invention to provide a periodical diagnosis method for a multiprocessor system that prevents a decrease in capacity.

[0008]

According to the present invention, in a multiprocessor system composed of a master processor and a plurality of slave processors connected to the master processor via a bus, the plurality of slave processors are One of a plurality of slave processors in each unit divided into groups is selected as a health check master, and the master processor periodically diagnoses only the health check master and performs restart processing according to the diagnosis result. A periodic diagnosis method for a multiprocessor system can be obtained.

Further, each health check master periodically diagnoses the slave processor in the same unit and transmits the diagnosis result to the master processor, and each slave processor transmits the diagnosis result to the health check master in the same unit. It is possible to obtain a periodical diagnosis method for a multiprocessor system, which is characterized in that

Further, the multiprocessor system comprises the master processor and a plurality of slave processors which are connected to the master processor through the bus and which are duplicated by a home system and a mate system. A slave processor and a plurality of slave processors of the mate system are grouped into a plurality of groups, and one of the slave processors of the home system in each unit is selected as a health check master, and the master processor is the health check master. A periodic diagnosis method for a multiprocessor system is obtained, which is characterized by performing periodic diagnosis of only those and performing restart processing according to the diagnosis result. Then, each of the health check masters performs periodic diagnosis of the home type slave processor and the mate type slave processor in the same unit, and transmits the diagnosis result to the master processor,
A periodic diagnosis method for a multiprocessor system is obtained in which each slave processor transmits a diagnosis result to the health check master in the same unit.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a first embodiment of a regular diagnosis system for a multiprocessor system according to the present invention, and FIG.
FIG. 3 is a block diagram showing a second embodiment of a periodical diagnosis system for a multiprocessor system of the present invention, and FIG. 3 is FIG. 1 and FIG.
3 is a flowchart showing a flow of operations in FIG.

Referring to FIG. 1, in the multiprocessor system of the first embodiment, a multiprocessor 101 and a plurality of slave processors 102, ...
7 are connected.

The master processor divides a plurality of slave processors into several groups based on the data of the slave processors managed in the master processor itself, and one representative slave processor from each unit. To decide. The health check master (Health check Mas) is used as a representative of the slave processor when performing periodic diagnosis.
ter).

In the first embodiment, the master processor 10
Reference numeral 1 divides the slave processors 102, ..., 106 connected on the bus 107 into two groups.

That is, the slave processors 102, 103 and 104 belong to the unit 108, and the slave processor 102 with the smallest slot number in this unit 108 is selected as the health check master to perform the periodic diagnosis in the unit 108. Let

Similarly, the slave processors 105 and 106 belong to the unit 109, and this unit 1
Slave processor 1 with the smallest slot number in 09
05 is selected as the health check master and unit 109
Have a regular internal diagnosis.

Next, referring to FIG. 2, the multiprocessor system of the second embodiment includes a multiprocessor 201.
, Home system, a plurality of home type slave processors 202H, ..., 206H and a plurality of mate type slave processors 202M ,.
M is connected through the bus 207.

The master processor divides a plurality of slave processors into several groups based on the data of the slave processors managed in the master processor itself, and one of the units is a representative slave processor. Is selected as the health check master.

In the second embodiment, the master processor 20
Reference numeral 1 divides the slave processors 202H, ... 206H and 202M, ..., 206M connected on the bus 207 into two groups.

That is, the unit 208 includes home slave processors 202H, 203H, 204H and mate slave processors 202M, 203M,
204M belongs to the unit 108, and the slave processor 202H having the smallest slot number in this unit 108 is selected as a health check master to perform regular diagnosis of the ACT system.

Similarly, home slave processors 205H and 206H and mate slave processors 205M and 206M belong to the unit 209, and the slave processor 205H with the smallest slot number in this unit 209 becomes the health check master. Select and let the ACT system perform regular diagnosis.

Next, the operation of the first and second embodiments will be described with reference to FIG.

As described above, in the first embodiment, the master processor 101 selects the slave processor 102 in the unit 108 as the health check master (step S301), and the master processor 101 instructs the slave processor 102 of the health check master to do so. A periodic diagnosis instruction signal (illustrated by a solid arrow) for performing periodic diagnosis only is transmitted to the slave processor 102 (S302).

When the slave processor 102 receives the regular diagnosis instruction signal from the master processor 101, the slave processor to be regularly diagnosed in the same unit 108 in the regular diagnosis instruction signal (slave processor in service in the ACT system). Of the slave processor 103 and 104 under its control, a regular diagnosis instruction signal (illustrated by a solid arrow) is transmitted (S303) to perform regular diagnosis.

The slave processor 102 checks whether or not there is a periodic diagnostic response signal (illustrated by a broken arrow) from the slave processors 103 and 104 under the slave processor 102 (S30).
4) If all slave processors can be received (Yes in S304), the slave processors 103, 1
The state of 04 is normal (S305), the regular diagnosis is ended, and a regular diagnostic response signal (illustrated by a dashed arrow) is transmitted to the master processor 101 (S308) to notify that the state of the slave processor is normal.

Further, when the periodic diagnosis response signal from at least one of the slave processors 103 and 104 is not received during the periodic diagnosis between the slave processor 102 and the slave processors 103 and 104 (S
No in 304), the slave processor 102 repeats sending the periodic diagnosis instruction signal several times (that is, Y in S306).
If es, the process returns to S303).

When the regular diagnostic response signal cannot be received within the number of times of retransmitting the regular diagnostic instruction signal (S306).
In No), it is determined that the states of the slave processors 103 and 104 are abnormal (S307), and a regular diagnostic response signal is transmitted to the master processor 101 (S308) to notify the abnormal state of the slave processors.

When the master processor 101 receives the periodic diagnosis response signal from the slave processor 102 (S3
(Yes in 09), the states of the slave processors 103 and 104 in the unit 108 are checked (S310), and if there is an abnormal state processor (Yes in S311), a restart processing request of the corresponding slave processor is transmitted (S3).
13).

When this slave processor is in the process of restarting, the slave processor 102 does not perform periodic diagnosis on the slave processor in which the abnormality is detected.

When the result in S309 is No, the process proceeds to S312, in which the master processor 101 makes the slave processor 10
The transmission of the periodical diagnosis instruction signal is repeated several times in 2. That is, if Yes in S312, the process returns to S302.

If the periodical diagnostic response signal cannot be received from the slave processor 102 within the number of retransmissions (No in S312), the process proceeds to S313.

The above steps S301 to S313
The above processes are performed in the same procedure for the slave processor 105 of the health check master in the unit 109 and the slave processor 106 under it.

The same process as described above is performed in the periodic diagnosis between the master processor 101 and the slave processor (health check master) 102.

As described above, when the result in S309 is No, that is, when the master processor 101 cannot receive the regular diagnostic response signal from the slave processor 102, the master processor 101 repeats the retransmission of the regular diagnostic instruction signal several times (S312 → S302), but if the periodical diagnostic response signal is still not received (N in S312, N
o), health check master (slave processor 10
The restart processing request of 2) is performed (S313).

At this time, since the slave processor 102 is in the process of restarting, another slave processor 103 in the same unit 108 is set as the health check master according to the constraint condition.

When the health check masters are replaced in this way, when the restart process of the slave processor 102, which is the initial health check master, succeeds,
This slave processor 102 is again used as a health check master, and thereafter, the master processor 101 transmits a periodic diagnosis instruction signal to the slave processor 102.

With respect to the periodic diagnosis response signal of the slave processor (health check master) 105 in the unit 109, the slave processor 106 is also subjected to the same procedure as above.
To Jerez Check Master.

In this embodiment, the master processor alternately performs regular diagnosis of the health check masters (slave processors 102 and 105) in the units 108 and 109 at a cycle of 1 second.

Next, in the second embodiment, the slave processors 202H, 202M, ...
Since the home system is the ACT system and the mate system is the SBY system because of the redundant configuration, the periodic diagnosis of the SBY system (mate system) slave processors 202M, ...
System (home system) slave processors 202H, ..., 2
06H transmits a periodic diagnosis instruction signal (shown by a dashed-dotted line arrow), and the slave processors 202M, ..., 20
This is performed by receiving a regular diagnostic response signal (illustrated by a two-dot chain line arrow) from 6M. However, the periodic diagnosis between the ACT system and the SBY system is performed regardless of the period of the periodic diagnosis instruction from the master processor to the health check master.

[0041]

As described above, according to the present invention, the master processor can regularly diagnose all of the plurality of slave processors by performing the regular diagnosis of only the health check master of each unit. This has the effect of reducing the load on the processor and avoiding a drop in its processing capability.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of a regular diagnosis system for a multiprocessor system according to the present invention.

FIG. 2 is a block diagram showing a second embodiment of the regular diagnosis system for the multiprocessor system of the present invention.

FIG. 3 is a flowchart showing a flow of operations in FIGS. 1 and 2.

FIG. 4 is a block diagram showing an example of a regular diagnosis method for a conventional multiprocessor system.

[Explanation of symbols]

101, 201, 401 Master processor 102, 105, 202H, 205H Slave processor (health check master) 103, 104, 106, 203H, 203M, 204
H, 204M, 206H, 206M, 402, 403,
404, 405, 406 Slave processor 107, 207, 407 Bus 108, 109, 208, 209 units

Claims (4)

[Claims] 1. A multiprocessor system comprising a master processor and a plurality of slave processors connected to the master processor through a bus, wherein a plurality of slave processors in each unit are grouped into a plurality of groups. One of the slave processors is selected as a health check master, the master processor periodically diagnoses only the health check master, and restart processing is performed according to the diagnosis result. 2. The health check master periodically diagnoses the slave processors in the same unit, and transmits the diagnosis result to the master processor, and the slave processors each diagnose the health check master in the same unit. The method for periodic diagnosis of a multiprocessor system according to claim 1, wherein 3. The multiprocessor system comprises the master processor and a plurality of slave processors connected to the master processor through the bus and duplicated in a home system and a mate system. A slave processor and a plurality of slave processors of the mate system are grouped into a plurality of groups, and one of the slave processors of the home system in each unit is selected as a health check master, and the master processor is the health check master. 3. The regular diagnosis method for a multiprocessor system according to claim 1, wherein only the regular diagnosis is performed and the restart processing is performed according to the diagnostic result. 4. The health check master periodically diagnoses the home slave processor and the mate slave processor in the same unit, and transmits the diagnosis result to the master processor. 4. The regular diagnosis method for a multiprocessor system according to claim 3, wherein the diagnosis result is transmitted to the health check master in the same unit.