JP2006221484A - Temperature anomaly monitoring method for computer - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To properly detect temperature anomaly in a CPU without any increase in a load of the CPU. <P>SOLUTION: A temperature sensor is stuck to a CPU package for the computer, and an abnormal temperature of the CPU is monitored based on a temperature detection signal from the temperature sensor. In a normal time when a temperature sensor value V is not above a stable operation upper limit value M of the CPU, monitoring is carried out at long period T1 intervals (S2-S5). When the temperature sensor value V exceeds the stable operation upper limit value M, a switch to monitoring at a short period T2 intervals is carried out (S8). When the short period monitoring condition continues beyond a set frequency (S7), a RAS function in the CPU starts notification of a warning to a user or shut down of the computer (S9). <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a monitoring method for monitoring a temperature abnormality of a computer, and more particularly to a method for determining the occurrence of a CPU temperature abnormality based on a temperature detection signal of a temperature sensor attached to a CPU package.

  Computers installed in the vicinity of various facilities and equipment, such as plant monitoring and control systems, that monitor and control them are poor in the surrounding environment (temperature, humidity, dust, etc.) in which they are installed. It is predicted that abnormalities (malfunctions or loss of function) will occur due to the above, and that it will develop into a fatal failure called system down. For example, there is a risk of CPU runaway (thermal runaway) due to a rise in computer temperature.

  Therefore, a protection method has been proposed in which the occurrence of a computer abnormality is detected in advance and appropriate protection measures (abnormality alarm, data backup, etc.) are taken (see, for example, Patent Document 1).

In the above-mentioned patent documents, the internal temperature detection of the computer by the temperature sensor, the fan operation state monitoring by the fan sensor, the power supply monitoring, etc. are performed. When an abnormality occurs, the RAS function (Reliability, Availability ( (Availability) and Serviceability (maintenance) are used to notify the user of an abnormality.
JP 2000-75952 A

  Among computer abnormality occurrence factors, CPU temperature abnormality is fatal, and it is important to detect the temperature abnormality at an early stage.

  In general, in order to detect an abnormal temperature of a CPU, a temperature sensor is attached to the CPU, and an abnormality is determined when a detection signal of the temperature sensor exceeds a predetermined threshold. Also, in the abnormality monitoring in a computer, a sensor detection signal such as a temperature sensor is acquired by a CPU at a constant cycle, the detection signal is compared with a threshold value, and if an abnormality occurs, a RAS function or the like is used. This is a method for informing the user.

  As described above, in the method of detecting anomalies by comparing the detection value of the temperature sensor with a threshold value at a fixed period, if the threshold setting and monitoring period are inappropriate, the system can be stopped unnecessarily or protective measures can be taken. There was a problem that was delayed.

FIG. 3 shows an aspect of CPU temperature monitoring, in which a temperature upper limit value (threshold value) at which the CPU can operate stably is set, and whether or not the temperature detection value by the temperature sensor exceeds the threshold value at a constant cycle (monitoring cycle). An abnormality is detected. Figure 3 In the case of (a), set low threshold, shows the case of setting the monitoring period relatively long and the detection value of the temperature sensor at time t ₁ and t ₄ is greater than the temporary upper limit, Although the thermal runaway does not occur and the CPU is preferably continued to operate, the abnormality is detected and the protection operation is started.

  In FIG. 3B, the threshold value is the same value, but the monitoring cycle is shortened and abnormality determination is performed using the average value of a plurality of detected values. In this case, unnecessary abnormality determination can be eliminated, but the CPU load increases because the monitoring cycle is shortened.

  FIG. 3C shows a case where the CPU load is reduced by extending the monitoring cycle in the abnormality detection using the average value as in FIG. In this case, although a temperature abnormality that causes the CPU to run out of heat during a long monitoring period occurs, the temperature cannot be detected and the time until the CPU is shut down cannot be secured.

  An object of the present invention is to provide a computer temperature abnormality monitoring system that can accurately detect a CPU temperature abnormality without increasing the CPU load.

  In order to solve the above problems, the present invention normally monitors the CPU temperature in a long cycle, and switches to short cycle monitoring when the CPU temperature exceeds the stable operation upper limit value in this monitoring. It is determined that a temperature abnormality has occurred in the CPU for the first time when the state continuously exceeds the set number of times, so that the user is notified and the computer is shut down.

(1) A monitoring method in which a temperature sensor is attached to a CPU package of a computer, and a temperature abnormality of the CPU is monitored based on a temperature detection signal of the temperature sensor,
Means for monitoring the temperature of the CPU in a long period under normal conditions;
Means for switching to short cycle monitoring when the temperature of the CPU exceeds the stable operation upper limit in the long cycle monitoring;
Means for determining that a temperature abnormality has occurred in the CPU when the short-cycle monitoring state continuously exceeds a set number of times;
And a means for informing the outside when the CPU temperature abnormality is determined and for starting the shutdown of the computer.

  As described above, according to the present invention, the CPU temperature is normally monitored in a long cycle, and when the CPU temperature exceeds the stable operation upper limit value in this monitoring, switching to the short cycle monitoring is performed. Since it is determined that a temperature abnormality has occurred in the CPU for the first time when the set number of times is exceeded, the user is notified and the computer is shut down, so unnecessary CPU load in the normal state can be avoided.

  In addition, it is possible to detect the temperature abnormality more reliably by setting the stable operation upper limit value of the CPU to be low, and the computer is not stopped unnecessarily.

  FIG. 1 is a processing procedure diagram of a temperature abnormality monitoring system showing an embodiment of the present invention. This process is realized by a software configuration of the computer. In addition, the temperature sensor required to implement this embodiment is affixed to the CPU package, and the detection signal of this temperature sensor is periodically converted to digital data by an A / D converter and stored in the memory of the computer. Keep it. Various set values are previously written in a refresh memory or the like.

  In FIG. 1, first, various setting values are read when the computer is started (S1). These set values are the monitoring period of the long period T1 and the short period T2, the stable operation upper limit value M of the CPU, and the temperature abnormality occurrence upper limit value N. After that, initialization of temperature abnormality occurrence number n = 0 and timer time limit T = T1 is performed (S2).

  After the initialization, the timer function waits for the elapse of T time (S3), and when this time comes, the CPU is interrupted and the CPU acquires the temperature sensor value V (S4). The CPU determines whether or not the temperature sensor value V exceeds the stable operation upper limit value M (S5). If not, the process returns to the process S2, starts the timer function, and the CPU executes other tasks. Return. Therefore, if the temperature sensor value V is equal to or less than the stable operation upper limit value M, the temperature abnormality monitoring process by the CPU is executed in the long cycle T1.

  When the temperature sensor value V exceeds the stable operation upper limit M, the CPU performs the temperature abnormality occurrence number n = n + 1 (S6), and determines whether the temperature abnormality occurrence number n is equal to or higher than the temperature abnormality occurrence number upper limit N. (S7). If the temperature abnormality occurrence count n does not exceed the temperature abnormality occurrence count upper limit N, the timer time limit T is switched to T2 (S8), the process returns to step S3, and the timer function time measurement is started with the short cycle T2, and the CPU Return to task execution. In the temperature monitoring process with the short cycle T2, the CPU repeats while incrementing the number n if the temperature sensor value V is equal to or greater than the stable operation upper limit value M, and again when the V ≦ M during the repetition, the long cycle T1 again. Return to temperature monitoring. Therefore, when the temperature sensor value V remains above the stable operation upper limit value M, the CPU repeats the abnormality monitoring with the short cycle T2.

  When n ≧ N is continuously obtained by repeating the abnormality monitoring, the CPU notifies the user of the temperature abnormality by executing the RAS function and starts shutting down the computer (S9).

  A mode of temperature abnormality monitoring by the above processing procedure is shown in FIG. As shown in FIG. 5A, monitoring is repeated in the long cycle T1 until the temperature sensor value V exceeds the upper limit value M, and monitoring is performed in the short cycle T2 from the time when the temperature sensor value V exceeds the upper limit value M. repeat. When the temperature sensor value V falls below the upper limit value M in the monitoring with the short cycle T2, the monitoring returns to the monitoring with the long cycle T1 again.

  Further, as shown in FIG. 2B, when the temperature sensor value V exceeds the upper limit value M in the monitoring with the short cycle T2, the temperature abnormality of the CPU surely occurs. It is determined that it has been performed, and notification to the user and shutdown are started, and data necessary to reproduce the current operating status is saved in a flash memory or the like.

  Therefore, according to the present embodiment, the processing is normally performed with the long cycle T1, and the load on the CPU can be reduced. On the other hand, when a temperature abnormality is detected, the process is switched to the processing with the short cycle T2, and the temperature abnormality is reliably detected, and the system down of the computer can be avoided reliably.

  Note that the periods T1 and T2, the stable operation upper limit M, and the like in the embodiment are appropriately changed according to the scale and performance of the computer and its processing target equipment / equipment. Further, when the determination of the temperature abnormality is obtained, the case where the user is notified by the RAS function or the shutdown is started is shown, but these processes can be appropriately changed.

The processing procedure figure of the temperature abnormality monitoring system which shows embodiment of this invention. The aspect figure of the temperature abnormality monitoring in embodiment. FIG. 6 is a diagram showing a conventional temperature abnormality monitoring.

Claims (1)

A temperature sensor is attached to a CPU package of a computer, and a monitoring method for monitoring a CPU temperature abnormality based on a temperature detection signal of the temperature sensor,
Means for monitoring the temperature of the CPU in a long period under normal conditions;
Means for switching to short cycle monitoring when the temperature of the CPU exceeds the stable operation upper limit in the long cycle monitoring;
Means for determining that a temperature abnormality has occurred in the CPU when the short-cycle monitoring state continuously exceeds a set number of times;
A computer temperature abnormality monitoring system, comprising: an external notification when a CPU temperature abnormality is determined, and means for starting a computer shutdown.

Images