JPH11191016A - Multiprocessor degeneration switching method - Google Patents

Abstract

(57) [Problem] In a multiprocessor system in which each processor has a cooling fan, a processor that cannot be cooled due to a fan failure operates by temporarily stopping and restarting the system after ensuring data consistency of the system. To provide a degeneration switching method for improving the operation rate. Kind Code: A1 In a multiprocessor equipped with a plurality of fans for cooling a processor, the presence or absence of a failure of each cooling fan is monitored, and a failure of any one of the cooling fans is cooled by the failed cooling fan. Switch to low power mode, disconnect, restart system and continue processing. Further, when all the cooling fans are faulty, an alarm or the like is output and the system is stopped.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multiprocessor degeneration technique having a plurality of highly integrated processors, and more particularly to a technique suitable for degeneration of a multiprocessor using a cooling system in which a fan is built in a radiation fin.

[0002]

2. Description of the Related Art As a degenerate operation technique for a fan failure of an electronic device, a degenerate operation method described in Japanese Patent Application Laid-Open No. 4-184515 is known. In this method, the number of allowable failed fans is determined in advance, counted each time a fan failure occurs, and when the number of failed fans exceeds the allowable number of failed fans, the operation is shifted to a degeneration operation. Further, a degenerate operation control device based on cooling state abnormality detection is disclosed in
No. 244535 is known. In this device, the processor that has become unable to cool is disconnected in a low-power mode, by turning off a power supply, and the like, and degraded without stopping the system by continuing to operate other parts.

[0003]

When a high-integration processor is cooled by blowing air from a fan, there is a problem that the operating rate is generally lowered because the fan failure rate is higher than the semiconductor failure rate. Further, in the technique described in Japanese Patent Application Laid-Open No. 4-184515 in which fans have redundancy, it is necessary to equip a large number of fans, resulting in an expensive system. Further, in the technique described in Japanese Patent Application Laid-Open No. Hei 7-244535, the processor has a master / slave configuration, the slave processor is stopped, and even if the data being processed is lost, the strictness of the system can be achieved. It can be applied to fields that do not require high data integrity. However, when operating a general operating system that does not have the function of dynamically changing the processor multiplicity in a symmetric multiprocessor, stopping some processors without stopping the operating system ensures data consistency. There is a problem that the operation of the operating system cannot be guaranteed thereafter.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above problem by temporarily switching the number of active processors of a multiprocessor system to a processor which cannot be cooled due to a fan failure while ensuring the data consistency of the system. The object of the present invention is to provide a multiprocessor degeneration switching method capable of improving the operating rate by continuing the operation separately by restarting.

[0005]

SUMMARY OF THE INVENTION In a multiprocessor equipped with a plurality of fans for cooling a processor, the presence / absence of a failure in each cooling fan is monitored. Switch the cooled processor to the low power mode, disconnect it, restart the system and continue processing.

Further, when all the cooling fans have failed,
Outputs an alarm etc. and stops the system.

[0007]

Embodiments of the present invention will be described in detail below with reference to the drawings.

First, an outline of the present invention will be described in the case of using a cooling method in which a fan is built in a radiation fin. FIG. 1 shows an example of a processor module used in the present invention. The processor module sends the processor chip 1 through the fins 102 by blowing air from the fan 103.
Cool 01. The processor chip 101 needs to have a low power consumption mode.

FIG. 2 is a block diagram of a multiprocessor system using the above processor module. Processor module (in FIG. 2, fans 201, 20
3, 205, 207 and processor chips 202, 20
4, 206 and 208 are shown separately. ) Is connected to the shared bus 211. The shared bus 211 has a shared memory 2 for storing a processor standby flag and the like.
09, a fan sensor interface 210 for monitoring the rotation state of the cooling fan, a firmware program memory 212 for controlling interrupt processing and the like, a reset interface 213 for performing a system reset and the like,
201 and 20 having low power consumption mode
3, 205 and 207 are connected. Processor 20
1, 203, 205, and 207 have cooling fans 202, 204, 206, and 208 built in cooling fins, respectively.

Next, the processing operation of one embodiment of the present invention will be described with reference to the flowchart of FIG. 3 and FIG.

If there is no fan failure, at the time when the operation of the firmware by reset is started, the rotation state of the fan is read by the fan sensor interface 210 (301), and it is determined whether the fan of its own processor has failed (302). ). If there is no fan failure, it is determined whether there is another fan failure (303), and if there is no other fan failure, all processors operate (304). Thereafter, it is determined whether there is a checkpoint (305), and if there is no checkpoint, an initial startup is performed (306), and thereafter, a business process is executed (307).

Processing when a fan stop interrupt occurs will be described with reference to the flowchart of FIG.

[0013] If a fan fails during business processing,
The fan sensor interface 210 detects this, and a fan stop interrupt is generated by the interrupt function. The checkpoint before the interruption is saved by the processing of the fan stop interruption (401). this is,
This is performed so that processing can be continued from the business processing before the interruption when the interruption processing is completed. After saving the checkpoint, the system is stopped (402), and the reset mechanism 213 is driven to perform a reset (403).

The reset process will be described with reference to the flowchart of FIG.

The processing operation when there is a fan failure of its own processor will be described below. When the operation of the firmware by the reset is started, the rotation state of the fan is read by the fan sensor interface 210 (301), and it is determined whether or not the fan of its own processor has failed (302). In this case, if there is a fan failure of the own processor, it is determined whether the fan is the last one (308). Here, since there are still fans that have not failed, the processor of the failed fan is stored in the shared memory 20.
A value indicating a fan failure is stored at a specific address of No. 9 and the processor is shifted to the low power consumption mode, thereby preventing heat generation of the processor (309). In this case, since the fans of the other processors have not failed, the operating system and the business application are started by the remaining processors and the business process is continuously executed as described later.

Next, there is a fan failure of the own processor,
The operation when the fan is the last one will be described. When the operation of the firmware by the reset is started, the rotation state is read by the fan sensor interface 210 (301), and it is determined whether the fan of its own processor has failed (302). Since there is a fan failure of its own processor, it is determined whether the fan is the last one (308). If the fan is the last one, it is impossible to shift to the degenerate operation. Therefore, a failure alarm or the like of the system device is displayed, the business process is stopped, and the system is stopped. After that, the mode shifts to the low power consumption mode and waits (310).

Next, the operation when the fan of the own processor is normal and the fan of another processor has a failure will be described. At the time when the operation of the firmware by the reset is started, the rotation state is read by the fan sensor interface 210 (301), it is determined whether the fan of its own processor has failed (302), and it is determined whether there is another fan failure. (303). When there is a fan failure of another processor, the processor determines whether a value indicating the fan failure has already been stored at a specific address of the shared memory 209 (311). Since the processor has already stored the value indicating the fan failure at the specific address of the shared memory 209, the operating system and the business application are started by the remaining processors (304).
Thereafter, it is determined whether there is a checkpoint (30).
5) Since there is a checkpoint, the operation is started from the checkpoint (313), and the business process is continuously executed (307).

Next, a description will be given of the operation in the case where the set time expires in which a value indicating the fan failure is stored at a specific address of the shared memory 209 of the processor due to a fan failure of another processor. When the operation of the firmware by the reset is started, the fan sensor interface 210
Then, the rotation state is read (301), it is determined whether the fan of its own processor has failed (302), and it is determined whether there is another fan failure (303). When there is a fan failure of another processor, the processor determines whether a value indicating the fan failure has already been stored at a specific address of the shared memory 209 (311). Since the processor has not yet stored the value indicating the fan failure at the specific address of the shared memory 209, the set time for storing the value indicating the fan failure at the specific address of the shared memory 209 of the processor has expired (the relevant processor has This indicates that it was not possible to shift to the low power consumption mode even though the fan failed.) (312). Since the setting time for storing the value indicating the fan failure at the specific address of the shared memory 209 of the processor has not expired, the processor again determines whether the value indicating the fan failure has already been stored at the specific address of the shared memory 209 (311). ). Since the value indicating the fan failure has not yet been stored in the specific address of the shared memory 209 and the value indicating the fan failure has been stored in the specific address of the shared memory 209 of the processor, the shared memory of the processor has expired. Since it is not possible to shift to a state in which a value indicating a fan failure is stored at a specific address 209, a failure alarm or the like of a system device is displayed to stop business processing and stop the system, thereby preventing overheating.

The processor has stored a value indicating a fan failure at a specific address of the shared memory 209 without the set time for storing a value indicating a fan failure at a specific address of the shared memory 209 of the processor. In some cases, the operating system and the business application are started on the remaining processors (304), and thereafter,
It is determined whether there is a checkpoint (305), and since there is a checkpoint, the operation is started from the checkpoint (313), and the business process is continuously executed (30).
7).

[0020]

According to the present invention, by switching the processor paired with the fan, the operation rate can be increased without stopping the system as long as the redundancy of the processor permits. Further, even when the cooling fans of all the processors are stopped, the system can be safely stopped and restarted from the checkpoint after repair, so that the safety and data consistency of the system can be ensured.

[Brief description of the drawings]

FIG. 1 is a schematic diagram of a processor module according to an embodiment of the present invention.

FIG. 2 is a block diagram of a multiprocessor system according to one embodiment of the present invention.

FIG. 3 is a flowchart at the time of starting by reset in one embodiment of the present invention.

FIG. 4 is a flowchart at the time of a fan stop interrupt in one embodiment of the present invention.

[Explanation of symbols]

101, 202, 204, 206, 208: processor chip, 102: heat radiation fin, 103, 201, 20
3, 205, 207 ... fan, 209 shared memory,
210 fan sensor interface, 211 shared bus, 212 firmware program memory, 2
13 Reset interface.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Osamu Osano 1st Horiyamashita, Hadano-shi, Kanagawa Prefecture Within Nichi Information Technology Co., Ltd. Inside

Claims (1)

[Claims] In a multiprocessor equipped with a plurality of processors and a plurality of fans for cooling the plurality of processors, the presence or absence of a failure of each of the plurality of cooling fans is monitored. And switching the processor cooled by the failed cooling fan to the low power consumption mode, disconnecting the processor, restarting the system, and continuing processing.