JPH02216577A - Fault detecting system in multi-processor system - Google Patents

Abstract

PURPOSE:To decrease an increase of a load of a management use processor and an operation monitoring interval of each processor and to enhance the reliability by forming two pieces each of each processor except the management use processor to a pair, and transmitting and receiving alternately an operation monitoring signal between two processors. CONSTITUTION:Between two processors in each pair, an operation monitoring signal S1 is transmitted and received each other. For instance, when a fault is occurred in a processor 32, the operation monitoring signal S1 transmitted and received in a pair 41 stops, it is detected by the other processor 31, and the processor 31 transmits the other processor fault informing signal to a management use processor 21. The management use processor 21 executes alternately transmission and reception of the operation monitoring signal S1 to and from the processor 31 in a period of a fault of the processor 32. In this regard, during this time, as well, other two pairs 42, 43 execute a normal operation, and transmit and receive alternately the operation monitoring signal S1 between the processors in each pair.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a failure detection method in a multiprocessor system.

(Prior Art) Conventionally, a failure detection method in this type of multiprocessor system involves sequentially transmitting operation monitoring signals to all processors from a management processor that manages the operating status of each processor in the system, and then returning the signals. The occurrence of a failure in all processors in the system is detected by monitoring incoming response signals and determining that the processor is at fault if the response signal is not received within a predetermined time.

FIG. 3 is a system block diagram showing an example of a failure detection method in a conventional multiprocessor system.

The multiprocessor system of this example includes one management processor 21 and three processors 31. The management processor 21 is composed of processors 31 . ~ 33, transmits an operation monitoring signal Sl to the processor 31, monitors that the response signal S3 is received within a predetermined time, and uses the result as the operating status of the processor 31. The processor 32. By sequentially repeating the same procedure for the processors 33, each processor 3
1. ~33 failures have been detected.

[Problem to be solved by the invention]

The fault detection method in the conventional multiprocessor system described above is a method in which the management processor sequentially sends an operation monitoring signal to all other processors and monitors the response signals, so when the number of processors in the system increases, However, the disadvantage is that the number of operation monitoring signals to be sent increases, increasing the load on the management processor. On the other hand, from the point of view of individual processors, there is a drawback that the interval at which the operation monitoring signal is received becomes long, and the time until a failure is detected becomes long.

[Means to solve the problem]

A failure detection method in a multiprocessor system according to the present invention is a multiprocessor system in which a plurality of processors and one management processor that manages the operating state of each processor are connected via a bus to perform distributed processing. transmitting/receiving means for alternately transmitting and receiving operation monitoring signals between the two processors in one bear, each of which has two processors; The one processor is characterized in that, when the operation monitoring signal cannot be received from the partner processor within a time, the one processor includes notification means for notifying the management processor that the partner processor is in trouble.

〔Example〕

Next, the present invention will be explained with reference to FIGS. 1 and 2.

FIG. 1 is a system block diagram showing an embodiment of the fault detection method in the multiprocessor system of the present invention, in which (a) shows the fault detection operation in a normal state, and (b) shows the operation when a fault occurs. (c) shows a failure detection operation performed by the management processor when the other processor is in failure. Further, FIG. 2 is a sequence diagram showing an example of a sequence of a series of failure detection operations in FIG. 1.

First, the fault detection operation of this embodiment will be explained with reference to FIG.

The multiprocessor system in this embodiment is configured such that a management processor 21 and general processors 318 to 36 are connected by a ring-shaped bus 11. Bear 41 for processor 31 and processor 32, Bear 42 for processor 33 and processor 34, Processor 35
and the processor 36 form a total of three bears, the bear 43,
As shown in FIG. 1(a), the two processors in each bear mutually transmit and receive operation monitoring signals S1. If, for example, a failure occurs in the processor 32 in this state, the operation monitoring signal S1 that has been sent and received within the bear 41 will stop, as shown in FIG.
1, and the processor 31 transmits a partner failure notification signal S2 to the management processor 21. The management processor 21 that received this partner failure notification signal S2
As shown in FIG. 3(c), during the period when the processor 32 is in failure, the processor 32 becomes a temporary bare partner of the processor 31.
During this period, the other two bears 42 and 43 are normal, so
As usual, the operation monitoring signal S1 is sent and received alternately between the processors in each bear.

Next, the failure detection operation sequence will be explained with reference to FIG.

FIG. 2 shows the failure detection operation sequence in the bare 41 formed by the processor 31 and the processor 32. In normal operation, the operation monitoring signal S1 is sent between the processor 31 and the processor 32 at the timing of S1 transmission timing out. The operation monitoring signals S1 which are sent to each other alternately and then received from the partner processor are monitored. If a failure occurs in the processor 32 in this state, the processor 32 is
1 detects a failure of the processor 32. At this time, the processor 31 sends the other party failure notification signal S2 to the management processor 21.
Send to. The management processor 21 that has received the partner failure notification signal S2 changes the operating state of the processor 32 in which the failure has occurred, and then transmits the operation monitoring signal S1 to the processor 31. The processor 31 receives this operation monitoring signal S.
1, and thereafter receives the operation monitoring signal S1 alternately with the management processor 21.

〔Effect of the invention〕

As explained above, the failure detection method in the multiprocessor system of the present invention makes each processor other than the management processor in the system bear two, and independently communicates between the two processors in each bear for each bear. It alternately sends and receives operation monitoring signals to monitor the operating state of the other processor, and if it cannot receive the operation monitoring signal from the other processor within a predetermined time, it determines that a failure has occurred in the other processor and notifies the management processor. The management processor that receives the fault then sends and receives operation monitoring signals to and from the processor that sent the fault notification, and the management processor does not intervene in fault detection under normal operating conditions of the system. This eliminates an increase in the load on the management processor and an increase in the operation monitoring interval of each processor due to an increase in the number of processors, and has the effect of realizing a highly reliable large-scale multiprocessor system.

[Brief explanation of the drawing]

FIGS. 1(a), (b), and (C) are system block diagrams showing an embodiment of a fault detection method in a multiprocessor system of the present invention, and FIG. 2 is an example of a series of fault detection operations in FIG. 1. FIG. 3 is a system block diagram showing an example of a failure detection method in a conventional multiprocessor system. 11... Bus, 21... Management processor, 31° 32.33.34.35.36... Processor, 41
．． 42.43...Bear, Sl...Operation monitoring signal, S
2: Other party failure notification signal. 1st country (α)

Claims (1)

[Claims] In a multiprocessor system that performs distributed processing by connecting a plurality of processors and one management processor that manages the operating state of each processor via a bus, two of the plurality of processors each form one pair, and this one
transmitting/receiving means for alternately transmitting and receiving an operation monitoring signal between the two processors in a pair; A failure detection method in a multiprocessor system, characterized in that the one processor includes notification means for notifying the management processor that the other processor is in failure when reception is not possible.