JP2000322334A - Automatic monitoring system for input and output performance - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an automatic monitoring system for input and output performance surely and early monitoring the deterioration of the input and output performance of an input and output device. SOLUTION: In this automatic monitoring system for input and output performance, the estimated maximum number of input and output competition, maximum length of transmission data, and performance monitor interval value are stored in a main storage device 20 by an inputting means 11 of information specific to a system, and input and output execution with the maximum input and output data length to a real device and estimated input and output performance upper limit calculation is operated by an input and output time sampling means 12, and the load of each input and output path is averaged by an input and output load distribution controlling means 13, and the accumulating totals of the necessary input and output time of each input and output path and the number of times of executed input and output are stored into the device 20 by an input and output statistical information obtaining means 14, and an abnormal part is specified from an input and output performance upper limit value by an input and output performance monitoring means 15, and the path of the abnormal part is closed at the time of the invalidity of use by a device re-constituting means 16, and the fault contents and fault parts are communicated to a fault phenomenon communicating means 17.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for performing load distribution control of a plurality of input / output paths for peripheral devices of a computer system, and constantly monitors the time required for each input / output for each access path. In addition to detecting the degradation in performance and localizing the effect on system operation in conjunction with the device configuration control function, the maintenance and operation department is notified of the opportunity to perform preventive maintenance by acquiring the factor information. The present invention relates to an automatic input / output performance monitoring system designed to improve the availability of stable system operation.

[0002]

2. Description of the Related Art Conventionally, a decrease in input / output performance with respect to a peripheral control device is caused by the presence or absence of a cache or RAID (Redundant
Since it is greatly affected by device-specific functions (static information) such as a configuration of rray of inexpensive discs, it has been difficult to dynamically judge that the performance of each input / output is degraded. In addition, the performance degradation often depends on the delay of the higher-level application and the timeout of the higher-level application.
I had to judge.

[0003]

These problems are as follows. First, the cause of performance degradation is frequent thrashing and many factors such as a job CPU loop can be considered. Analyze the trace data to determine that a failure has occurred,
There is a problem that it is necessary to investigate the cause, and there is a possibility that even if the performance degradation is actually caused by a failure, it may not be noticed for a long time. As a result, there is a problem that the opportunity for preventive maintenance is lost, and at the same time, there is a concern that the subsequent device failure may affect the system. Second, when the performance is degraded, there is a problem that the work of investigating the cause, identifying the content of the fault, and executing the workaround must be performed manually.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problems, and it is possible to more reliably, early, and easily monitor a decrease in input / output performance with respect to an input / output device. It is an object of the present invention to provide an input / output automatic performance monitoring system that can automatically avoid performance degradation due to an input / output path failure by disabling a failed path.

[0005]

In order to achieve the above object, an automatic input / output performance monitoring system according to the present invention obtains a maximum number of input / output conflicts assumed in advance as a performance influence factor, and obtains a logical unit. An input means for system-specific information for acquiring a maximum transfer data length that can be subjected to one input / output and an appropriate performance monitoring interval value according to each system operation, and storing the acquired information in a main storage device; Performs input / output with the maximum input / output data length to / from the actual device with reference to the system-specific information stored in the main storage device, and executes the performance result, the maximum number of input / output conflicts, and the hardware stored in the main storage device. An input / output time sampling unit for calculating an upper limit value of expected input / output performance based on the hardware configuration information and storing the calculated upper limit value in the main storage device, and usable from the main storage device I / O load distribution control means that obtains appropriate route information, determines a use route for each I / O request by a round robin method, and averages the load of each I / O route regardless of the host connection state and the I / O type. A total of input / output time required for each input / output path, and input / output statistical information acquisition means for storing the number of executed input / output in the main storage device, and periodically stored in the main storage device. Input / output performance monitoring means for specifying an abnormal point from the input / output performance upper limit value stored in the main storage device with reference to the input / output statistical information, and the input / output path by the input / output performance monitoring means. Device reconfiguration control that, when it is determined that the abnormal path is unusable and the path is abnormal, the path of the abnormal point is closed, and the hardware configuration information is changed to make the target path an unusable path for input / output. means , Characterized in that it comprises a remote maintenance terminal, and a fault event notification means for notifying the fault content and fault location predicting through remote monitoring terminal to operations department.

In order to realize the above, the maximum number of input / output conflicts assumed in advance by the input means of the system-specific information,
Obtain the maximum transfer data length that can be the target of a logical single input / output, an appropriate performance monitoring interval value according to each system operation, store it in the main memory, and use the input / output time sampling means to Perform I / O with the maximum I / O data length, and calculate the expected upper limit of I / O performance based on the performance results, the maximum number of I / O conflicts, and the hardware configuration information stored in the main storage device. The available path information is obtained from the main storage apparatus by the input / output load distribution control means, the used path is determined by a round robin method for each input / output request, the host connection state, the input / output type Regardless of this, the load of each input / output path is averaged, and the total input / output time required for each input / output path by the input / output statistical information acquisition means and the number of executed input / output times are stored in main memory. The I / O performance monitoring means periodically specifies the abnormal part from the I / O performance upper limit value stored in the main storage device with reference to the I / O statistical information stored in the main storage device, In addition to closing the abnormal path determined to be unusable by the device reconfiguration control means, changing the hardware configuration information to make the target path an unusable path for input / output, the remote maintenance terminal, operation department Since the failure content and the location of the failure to be predicted are notified through the remote monitoring terminal using the failure event notification means, it is possible to more reliably, early, and easily monitor the input / output performance degradation of the input / output device, and automatically By setting the faulty path to the unusable state, it is possible to automatically avoid performance degradation due to an input / output path fault without depending on the device type.

[0007]

Next, an embodiment of an automatic input / output performance monitoring system according to the present invention will be described with reference to the drawings. In the first embodiment according to the present invention, as shown in FIG. 1, input means 11 for system-specific information, sampling means 12 for input / output time, input / output load distribution control means 13 for input / output requests, The system includes an input / output statistical information acquisition unit 14, an input / output performance monitoring unit 15, a device reconfiguration control unit 16, an external failure event notification unit 17, and a main storage device 20. These details will be described later with reference to FIG.

In general, input / output performance is caused by abnormal events such as processing performance of input / output devices, input / output load, and faults.
Therefore, in the first embodiment, the performance of an input / output device that can simultaneously process input / output requests from a plurality of nodes is monitored for each input / output path, and performance degradation due to an abnormal event and abnormal It is intended to dynamically detect a performance drop caused by an increase in the load required for input / output, and also to perform a reconfiguration of a device, a preventive maintenance operation, and a warning notification for an operation function.

Next, an example of an operating environment configuration to which the first embodiment is applied will be described. FIG. 2 is a block diagram showing a configuration example of this operation environment. 2, the input / output subsystem 5 has a plurality of input / output devices 6, and is capable of simultaneously processing input / output requests from a plurality of connected hosts # 1 to # 3. Output system. Further, a plurality of input / output paths 4 are provided between the host # 1 to the host # 3 and the input / output subsystem 5. The host # 1 to the host # 3 have a monitoring terminal 7 for monitoring operation from a remote place.

Next, the configuration example of the first embodiment according to the present invention shown in FIG. 1 will be described in detail. FIG. 1 is a block diagram. In FIG. 1, the input unit 11 for system-specific information is a unit for externally inputting input / output performance factor information that does not depend on HW (hardware) characteristics. In an environment in which the system of the present invention operates, FIG.
As described above, since a conflict may occur due to input / output requests from a plurality of hosts # 1 to # 3, it is necessary to consider input / output queuing in the input / output subsystem 5 in advance as a factor of input / output performance degradation. is there.

[0011] Therefore, input means 1 for system-specific information
In step 1, the maximum number of input / output conflicts assumed in advance is acquired as a performance affecting factor. In addition, a maximum transfer data length that can be a target of logical single input / output and an appropriate performance monitoring interval value according to each system operation are acquired. These are separate from the HW performance factors of the I / O subsystem,
This is a performance factor due to the system operation method. The information acquired by the system unique information input unit 11 is stored in the system unique information area 21 on the main storage device 20 and the main storage device 20.
It is stored in the performance monitoring interval value area 22 above.

In the input / output time sampling means 12,
With reference to the system-specific information stored in the system-specific information area 21, input / output with the maximum input / output data length for the real device is performed. Based on the HW configuration information stored in the HW configuration information area 23, the upper limit of the input / output performance expected in the system is calculated and stored in the input / output performance upper limit area 24 in the main storage device 20. Main storage device 20
The HW configuration information stored in the HW configuration information area 23 is fixed information (input / output paths to the subsystem 5, the number of input / output devices, etc.) due to the HW configuration of this system, and is stored as static information in the main storage device. 20 is stored in advance.

The input / output load distribution control means 13 obtains usable path information from the HW configuration information stored in the HW configuration information area 23, and determines a use path for each input / output request by a round robin method. The load of each input / output path is averaged by the input / output load distribution control means 13 irrespective of the host connection state and the input / output type.

The input / output statistical information acquiring means 14 stores the total input / output time required for each input / output path and the number of executed input / output operations in the input / output statistical information area 25 in the main storage device 20. . The input / output performance monitoring means 15 periodically refers to the input / output statistical information stored in the input / output statistical information area 25 on the main storage device 20 and calculates the average of the time required for a single input / output. The calculation is performed for each route, and the abnormal point is specified based on the calculation result and the input / output performance upper limit value stored in the input / output performance upper limit area 24. Here, when an abnormal part is specified, the abnormal path is closed through the device reconfiguration control means 16 and the failure event notification means 17 notifies the details of the failure.

When it is determined that the route is unusable due to a failure or the like, the apparatus reconfiguration control means 16 changes the HW configuration information in the HW configuration information area 23 to make the target route an unusable route for input / output. . The fault event notifying means 17 notifies the maintenance terminal and the operation department in the remote place of the fault content and the fault location to be predicted through the remote monitoring terminal 7 in FIG. This enables quick maintenance work.

In the first embodiment, the input / output load distribution control means 13 eliminates load imbalance depending on the type of input / output and dynamically monitors the average required time for a single input / output for each input / output path. In general, it is easy to judge a decrease in input / output performance of a peripheral control device which is difficult to be affected by device-specific functions such as a cache mechanism and a RAID configuration, and a performance degradation event independent of a device type is monitored. It is possible to do that.

Next, the operation of the first embodiment configured as described above will be described with reference to the flowchart of FIG. In the input of the system-specific information by the system-specific information input unit 11 (step 31), as the predictable performance influence factor information due to the system operation method,
The number of input / output conflicts, the maximum input / output transfer data length, and the performance monitoring interval value are acquired and stored in the system specific information area 41 and the performance monitoring interval value area 42 in the main storage device 20.

In the sampling of the input / output time by the input / output time sampling means 12 (step 32), an actually measured value of the input / output performance at the maximum input / output data length is measured based on this information. Further, the number of input / output paths included in the HW configuration information 43 in the HW configuration information area 23 of the main storage device 20,
Considering the number of input / output devices 6 in the subsystem 5,
Calculate the upper limit of expected input / output performance. The calculated upper limit of the expected performance value is stored in the input / output performance upper limit value 44 of the input / output performance upper limit area 24 on the main storage device 20 so as to be used as a performance monitoring threshold value.

The specific processing of the input / output load distribution control means 13 is based on the input / output path information in the HW configuration information 43 of the HW configuration information area 23 in the main storage device 20, and performs input / output in a round-robin manner. The input / output possible path to be used is selected (step 331), the input / output path is determined (step 332), and the load of each input / output path is distributed.

Further, the input / output statistical information obtaining means 14 obtains the time when the input / output is started, and stores the input / output start time in the main storage device 20 (step 33).
3) Execute input / output with respect to the input / output device (step 3)
34) After the actual input / output is completed, the input / output time is calculated from the difference from the time at that time, and added to the total required input / output time 45 in the main storage device 20. Is updated (step 335). Here, the total required input / output time 45 and the total number of input / output times 46 are managed separately for each input / output path.

The input / output performance monitoring processing by the input performance monitoring means 15 (step 34) waits for a time corresponding to the performance monitoring interval value 42 in the main storage device 20, and operates every time the processing is completed (step 340). . As shown in FIG. 3, first, the total input / output required time 45 for each input / output path on the main storage device 20 recorded by the input / output information acquisition processing (step 33) and the total number of input / outputs 46 are used. First, an average input / output time for a single logical input / output is calculated for each input / output path (step 34).
1).

The calculated average input / output time for each route is compared (step 342). If there is a clear difference between the input / output routes (step 348), the route with the biased input / output time is determined. Is determined to be only the device connected to the specific device in the subsystem 5 (step 343), and the failure location is specified. Thereafter, the failure route is closed by the device reconfiguration control means 16 of FIG. 1 (step 34).
4), the failure information specified by the failure event notification means 17 and the blocked path information are notified to the remote monitoring terminal 7 in FIG. 2 (step 345).

If it is determined in the comparison of the average input / output time for each route in the step 342 that no deviation to a specific route is found (step 348), the upper limit 44 of the input / output performance in the main storage device 20 Compare (Step 34)
6) As a result of the comparison, if the average input / output time of each path exceeds the input / output performance upper limit value 44, the performance of the remote monitoring terminal 7 is determined as performance degradation due to an event other than a failure of the specific path.
Is notified (step 347).

As described above, in the first embodiment, the performance of an input / output device that can simultaneously process input / output requests from a plurality of nodes is monitored for each input / output path, and performance degradation due to an abnormal event is prevented. In addition, it is possible to dynamically detect a performance decrease caused by an abnormal increase in the load on an input / output request, and also to perform a reconfiguration of a device, a preventive maintenance operation, and a warning notification to an operation function.

Next, a second embodiment of the present invention will be described. Depending on the operation of the system, even if the input / output performance decreases due to something other than a specific path failure,
It is assumed that the suspected device is quickly disconnected from the system, and another input / output device or another host can be used as a substitute for the system. Therefore, in the input of the system unique information in step 31 in FIG. 3, in addition to acquiring the unique information determined by the operation mode of the system, the device reconfiguration control means in the case where the corresponding threshold is violated is also registered as information in advance. By doing so, in the input / output performance monitoring processing in step 34, it is possible to perform device reconfiguration suitable for each system operation.

FIG. 4 is a flow chart for explaining the operation of this embodiment, that is, the operation of the second embodiment. The processing in FIG. 4 differs from the processing in FIG. 3 in the following points. In FIG. 4, the same processing steps as those in the flowchart of FIG.
The input of the system-specific information (step 51) is performed when the input / output performance exceeds the upper limit of the input / output performance, in addition to acquiring the factor information affecting the predictable performance required for performance monitoring. Get the operation for the configuration process. The device reconfiguration information when the acquired performance upper limit value is exceeded,
The device reconfiguration information 52 is stored in the device reconfiguration information area on the main storage device 20 shown in FIG.

In the input / output performance monitoring process (step 53), as in the case of FIG. 3, first, the total required input / output time and the total number of input / output times for each input / output path on the main storage device 20 are used. Calculate and compare the time required for a single input / output operation unit and specify it. As a result, if it is recognized that the performance degradation is caused by a cause other than the specific input / output path failure and the specific device failure, the device reconfiguration processing (step 5)
By referring to the device reconfiguration information 52 according to 4), the specified device reconfiguration such as the device closing process and the system forced stop process is performed. In this way, even if the cause cannot be specified and manual avoidance of performance degradation cannot be performed, by defining appropriate workarounds in advance according to system operation, automatic A means for monitoring performance degradation can be provided.

[0028]

As described above, according to the present invention, it is determined whether the input / output performance abnormality is due to a failure in a specific route or a failure in a specific device. The suspected route can be blocked immediately by the device reconfiguration control, and the effect on system operation can be localized when a performance abnormality occurs. In addition, the response time for a single I / O request is statistically compared in a state in which erroneous judgments due to bias in performance depending on the I / O type are reduced by the load distribution means for I / O requests, so that the cluster configuration etc. A performance abnormality can be detected without depending on the connection configuration of the input / output subsystem and the device type. Furthermore, no special hardware is required for determining the deterioration of the input / output performance, and the performance comparison based on the actually measured data in the actual operation system is performed. Abnormal performance can be detected without being aware of characteristics. In addition, the number of parts that depend on the normal input / output system is reduced, and the performance monitoring function can be realized as an independent function.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment of an input / output automatic performance monitoring system according to the present invention.

FIG. 2 is a block diagram showing a configuration example of an operating environment applied to the input / output automatic performance monitoring system according to the present invention.

FIG. 3 is a flowchart for explaining the operation of the first embodiment of the input / output automatic performance monitoring system according to the present invention;

FIG. 4 is a flowchart for explaining the operation of the second embodiment of the input / output automatic performance monitoring system according to the present invention;

[Explanation of symbols]

11 input means for system-specific information, 12 input / output time sampling means, 13 input / output load distribution control means, 14 input / output statistical information acquisition means, 15 input / output performance monitoring means, 16 ... Device reconfiguration control means, 17 ...
... Failure event notification means, 20... Main storage device, 21... System specific information area, 22... Performance monitoring interval area, 23... HW configuration information area, 24. Output statistics.

Claims (12)

[Claims] The present invention acquires a maximum number of input / output conflicts assumed in advance as a performance influencing factor, a maximum transfer data length that can be a target of a logical single input / output, and an appropriate performance according to each system operation. A means for inputting system-specific information for acquiring a monitoring interval value and storing the acquired information in the main storage; and a maximum input / output data length for the real device with reference to the system-specific information stored in the main storage. The I / O is performed, and the upper limit value of the expected I / O performance is calculated based on the performance result, the maximum number of I / O conflicts, and the hardware configuration information stored in the main storage device. Sampling means for input / output time to be stored, obtainable path information from the main storage device, determine a path to be used for each input / output request by a round robin method, I / O load distribution control means for averaging the load on each I / O path regardless of the connection status and I / O type, total I / O time required for each I / O path, and I / O executed I / O statistical information acquisition means for storing the number of times in the main storage device; and the I / O performance stored in the main storage device by periodically referring to the I / O statistical information stored in the main storage device An input / output performance monitoring means for specifying an abnormal point from the upper limit value, and when the input / output performance monitoring means specifies an abnormal point of the input / output path and determines that the path cannot be used, the path of the abnormal point is closed. In addition, the device configuration control means that changes the above hardware configuration information to make the target path an unusable path for input / output, the maintenance terminal at a remote location, and the failure content predicted through the remote monitoring terminal to the operation department Obstacle O automatic performance monitoring system comprising: the fault event notification means for notifying Tokoro, a. 2. The input device according to claim 1, wherein the input means for the system-specific information is a performance factor caused by the way of operating the system, separately from a hardware performance factor of the input / output subsystem. Output automatic performance monitoring system. 3. The input / output time sampling means,
2. The automatic input / output performance monitoring system according to claim 1, wherein the hardware configuration information is fixed information resulting from the hardware configuration of the system and is stored in the main storage device in advance. 4. The input / output performance monitoring means, when an abnormal point is identified, closes an abnormal path through the apparatus reconfiguration control means and notifies the failure event notification means of the content of the failure. The input / output automatic performance monitoring system according to claim 1, wherein: 5. The input / output statistical information obtaining means obtains a time at the time of starting the input / output, and calculates an input / output time from a difference from the time at the time when the actual input / output is completed. 2. The automatic input / output performance monitoring system according to claim 1, further comprising adding the total required input / output time stored in said main storage device to update the total number of input / output times stored in said main storage device. . 6. The automatic input / output performance monitoring system according to claim 5, wherein the total required input / output time is managed separately for each input / output path. 7. The automatic input / output performance monitoring system according to claim 5, wherein the total number of input / output operations is managed separately for each input / output path. 8. The input / output performance monitoring means, based on the total required input / output time and the total number of input / output times for each input / output path of the main storage device recorded by the input / output information acquisition processing, performs a logical single input / output operation. Calculate the average input / output time for each input / output path, compare the average input / output time for each path, and if there is a clear difference between the input / output paths, 2. The input / output automatic performance monitoring system according to claim 1, wherein it is determined whether or not only one of the devices is connected to the specific device and whether or not the failure is specified. 9. The input / output performance monitoring means, based on the total required input / output time and the total number of input / output times for each input / output path of the main storage device recorded by the input / output information acquisition processing, performs logical single input / output. Calculate the average input / output time for each input / output path, compare the average input / output time for each path, and compare the average input / output time for each input / output path. 2. The automatic input / output performance monitoring system according to claim 1, wherein a warning is notified to the remote monitoring terminal that the performance is degraded due to an event other than a failure. 10. The input / output performance monitoring means, when inputting the system-specific information by the system-specific information input means, conflicts with a threshold value other than acquiring specific information determined by an operation mode of the system. 2. The input / output automatic performance monitoring system according to claim 1, wherein the device reconfiguration control means in the case is registered in advance as information, thereby enabling device reconfiguration adapted to each system operation. 11. The input / output performance monitoring means performs a single input / output operation for each input / output path in the input / output performance monitoring process based on a total required input / output time and an input / output count for each input / output path of the main storage device. 2. The input / output automatic performance monitoring system according to claim 1, wherein the required time of the unit is calculated and compared to perform the specification. 12. The input / output performance monitoring means performs a single input / output operation for each input / output path in the input / output performance monitoring processing based on the total required input / output time and the number of input / output operations for each input / output path of the main storage device. If it is recognized that the performance degradation is caused by a cause other than a specific input / output path failure or a specific device failure by calculating and comparing the required time of the unit and performing the specification, the device is reconfigured by the device reconfiguration processing. 2. The input / output automatic performance monitoring system according to claim 1, wherein the specified device reconfiguration is performed with reference to the reconfiguration information.