JPH0449435A - Method for correcting time at the time of recording trace information - Google Patents

Abstract

PURPOSE:To prevent the generation of a trouble in the merging of trace information by reading out reference time and the time of a reference clock and correcting the time of a self-clock when the reference time stored in an I/O device is effective, and then starting the recording of trace information. CONSTITUTION:The I/O device 3 is provided with a storage means 5 for storing reference time applied from plural master devices 1, 2 and the validity of the reference time. If the reference time stored in the storage means 5 of the I/O device 3 is effective when either one of the master devices 1, 2 records trace information, the reference time stored in the means 5 and the time of a reference clock 4 are read out, the time of the self-device, i.e. the master device 1 or 2, is corrected and then the recording of the trace information is started. Since the clock deviation of plural computers can be corrected at the time of succeeding merging, highly reliable trace information can be obtained.

Description

[Detailed Description of the Invention] [Summary] When a plurality of higher-level devices trace the issuance history of accesses to input/output devices, the time at which trace information is recorded allows each higher-level device to trace based on the same time. Regarding the correction method, the purpose is to ensure that there is no problem in merging trace information, and it is shared by multiple higher-level devices that each record trace information that traces the issuance history of accesses to input/output devices. In a system equipped with an input/output device, the input/output device M is provided with a reference clock, a storage means for storing the reference time given from the host device and the validity of the reference time, and one When a host device records the trace information, if the reference time stored in the input/output device is invalid, it causes the input/output device to store a reference time based on its own clock, and After storing the validity and resetting the reference clock, recording of trace information is started, and when the two higher-level devices record the trace information, the reference time stored in the input/output device is valid. In this case, the device starts recording trace information after reading the reference time stored in the device and the time of the reference clock and correcting the time of its own clock.

[Industrial Field of Application] The present invention provides a method for recording trace information that enables each host device to trace the access history to an input/output device based on the same time as a reference. This invention relates to a time correction method.

In recent years, the history of accesses issued to input/output devices has been recorded and used for performance analysis and fault investigation of the input/output devices, but the most common method is to record the history of accesses issued to input/output devices. , access issuance history is traced and recorded on a magnetic tape device, etc. In this case, even if the input/output device is shared by multiple host computers, the records of accesses made by the host computer are not recorded. It is only done.

However, input/output devices are often shared by multiple host computers, and there may be conflicting accesses.In this case, there is a post computer that has to wait under exclusive control, which often causes performance problems for the input/output devices. .

Therefore, it is necessary to perform performance analysis by merging trace information of access issuance histories recorded by a plurality of postcomputers. For this purpose, the trace information collected by each host computer must be traced based on the same time, otherwise the information at the time of merging will not be reliable. However, the accuracy of the clocks of each host computer is too high for practical use, and it is necessary to perform time correction.

[Conventional technology]

For example, when a single magnetic disk device is shared by a plurality of host computers, conventionally, when one host computer records trace information, it only records its own access issue history. The recorded contents include the access issuing time, the access end time, the identifier of the accessed magnetic disk device, and the record position.

[Problem to be solved by the invention]

When performing performance analysis or fault investigation of magnetic disk devices,
If it is only accessed by one host computer, it is enough to analyze the trace information collected by the host computer, but if it is accessed by multiple host computers, do not merge the trace information collected separately by multiple host computers. I am unable to make a comprehensive judgment.

When merging trace information collected individually by multiple host computers, the authenticity of the trace information will be lost unless the access issue time, end time, etc. of each host computer are corrected using one reference time. Post computers only have separate clocks, and merging them will inevitably result in considerable errors.

In other words, the accuracy of each host computer's clock has a maximum error of several seconds, and considering that the access time to the magnetic disk device is several tens of milliseconds, there is a problem that the accuracy is impractical. be.

For this reason, it may be possible to collect trace information on the magnetic disk control device side, but in this case, the problem of clock accuracy error between host computers does not occur because the time can be recorded using the same clock. However, since information on the accessing side cannot be recorded, and the storage capacity of the magnetic disk control device itself is not sufficient for this purpose, there is a problem that it is not suitable for long-term recording.

In view of these problems, the present invention aims to make it possible to acquire time correction information on the host combination router side, so that there is no problem in merging trace information. There is.

[Failure to solve the problem]

FIG. 1 is a block diagram illustrating the invention in detail.

The input/output device 3 is provided with a base clock 4 and a storage means 5 for storing the reference time given from the plurality of host devices 1 or 2 and the validity of this reference time.

The plurality of higher-level devices 1 and 2 record trace information obtained by tracing the history of accesses to the shared input/output device 3 in the external storage devices 6 and 7, respectively. At this time, the operations are as follows.

When one higher-level device, that is, the lower device 1 or 2, records trace information, if the reference time stored in the storage means 5 of the input/output device 3 is invalid, the input/output device 3 records its own special note, That is, the reference time based on the clock of the host device 1 or 2 is sent out and stored in the storage means 5, and the validity of this reference time is stored.
After performing step 1, recording of trace information is started.

Also, there are two higher-level devices, that is, higher-level device 1 is 2 and 1 is 1.
When recording the L.-space information, if the reference time stored in the storage means 5 of the input/output device 3 is invalid, the base 1'11 time stored in the storage means 5 of the input/output device 3, After reading the time on the reference clock 4 and correcting the time on the clock of the host device W1 or 2, recording of trace information is started.

[Effect]

By configuring as described above, when the host device 1 first records trace information in the external storage device 6, the storage means 5 of the input/output device 3 stores the stored base (1t time), so that the The device l stores the time indicated by its own clock in the storage means 5 of the input/output device 3 as (i#-time), and also stores the validity of this reference time. Thus, the access issuance history can be traced and sequentially recorded in the external storage device 6.

The host device 2 that starts recording trace information later indicates that the %r% time stored in the storage means 5 of the input/output device 3 is valid, so it compares it with the reference time stored in the storage means 5. , the host device 1 reads the time counted since the clock 4 was turned on, and sets its own clock based on the time obtained by adding the time of the reference clock to this reference time. Therefore, the access issuance history can be traced and sequentially recorded in the external storage device (2) in accordance with the corrected time.

Therefore, even if the trace information recorded by the external storage devices 6 and 7 is merged, there is no time difference, so the high level of Buddhist axis 8
You can obtain information on rivers 2 and 2.

〔Example〕

FIG. 2 is a block diagram of a circuit showing one embodiment of the present invention.

In accordance with instructions from the computer controller 1, the host I to the computer 8 send input/output commands to the channel interface circuit 14 of the disk controller W10 to request access to the disk device 11. At this time, the host computer 8 records trace information obtained by tracing this access issuance history in the magnetic tape device 12 according to instructions from the trace program. and the flag 23 recorded in the memory 211 is set to J, a ; 414 time 01) (g 22
) Check if it is valid.

If the flag 23 indicates invalidity, the host-lumpiator 8 reads the time from its own clock and sets the time of l.
' Write the mail as OD value 22, and at the same time set flag 23 to 6.1 o'clock 1:12.
Lyse 0, [sal! :Ru. Therefore, the time 1it20 starts measuring time from the time sent by the postcomputer 8.

Based on the input/output command sent by the host computer 8, the processor 19 sends the data received via the channel interface circuit 14 to the disk device 11 via the data transfer circuit 17 and the device interface circuit 18 in the case of data writing. If data is to be written or read from a specified record position, the disk device 11
The data is sequentially read from the specified record and transferred to the host computer 8.

Along with the above operation, the magnetic tape device 12 has the following information:
The access issue time, the access end time, the identifier of the accessed disk device 11, the record position, etc. are recorded, and the access issue time and the access end time are recorded based on the time sent by the clock of the host computer 8.

Further, the host computer 9 sends an input/output command to the channel interface circuit 15 of the disk control device 10 to request access to the disk device 11 according to instructions from the application program. At this time, the host computer 9 causes the magnetic tape device 13 to record this access information according to instructions from the trace program. First, in order to obtain a reference time, the host computer 9 instructs the processor 19 via the channel interface circuit 15 to record the access information in the memory 21. Check whether the flag 23 indicating that the reference time TOD value 22 is valid is checked.

At this time, the TOD value 2 has already been set by the host computer 8.
2 is written in the memory 21 and the flag 23 indicates valid, the host computer 9 reads the TOD value 22 stored in the memory 21 and also reads the time on the clock 20.

Then, the time obtained by adding the read TOD value 22 and the time of the clock 20 is set as a reference time, and the time of the own clock is corrected. Then, using this corrected time, the access issue time and access end time are recorded in the magnetic tape device 13.

When the host computer 8 finishes recording the trace information, it instructs the processor 19 to invalidate the flag 23 in the memory 2I.

The same applies when a host computer is connected to the channel interface circuit 16.

FIG. 3 is a flowchart illustrating the operation of FIG. 2.

When the trace program of the host computer 8 or 9 starts tracing processing, it reads correction information from the disk controller 10 in step (3). That is, the flag 23 in the memory 21 is read. Then, in step (2), it is checked whether the flag 23 is valid.

If invalid, the disk controller 10 is instructed to store the TOD value 22 and reset the clock 20 in step (3). That is, it stores the time on its own clock.

Then, in step ■, set the TOD value offset to O,
End trace start processing. That is, it starts tracing using the time sent by its own clock.

If the flag 23 indicates validity in step (2), the TOD value correction offset is obtained in step (2), and the trace start process is ended. That is, the TOD value 22 stored in the memory 21 and the time of the clock 20 are read, the difference between the time of the own clock is determined, the time of the own clock is corrected using this difference, and tracing is started.

When acquiring time correction information using the procedure described above, there are problems with the execution time and accuracy of the instructions for setting and reading the correction information, but the instruction execution time is usually several hundred microseconds, and the correction information is It must be taken into consideration that this includes an error of several hundred microseconds. However, considering that this correction information is used to ensure consistency of trace information between host computers, it is sufficient that it is sufficiently shorter than the access execution time for the input/output device.

The access execution time for an input/output device is usually several tens of milliseconds, even for a magnetic disk device, which is the fastest of the input/output devices, and is at least a few milliseconds, so this error will be merged later. It is small enough.

Further, the clock 20 measures time in units of at least 100 microseconds, and must be able to start measuring time from 0 by a reset signal.

[Effects of the Invention] As explained above, the present invention allows a plurality of host computers to record trace information of access issuance histories to input/output devices, and when merging the information later, corrects the difference in the clocks of the host computers. Therefore, trace information with high reliability can be obtained.

4 is a reference clock, 5 is a storage means, 11 is a disk device, 12.13 is a magnetic tape device,
14 to 16 are channel interface circuits, 17 is a data transfer circuit, 1g is a device interface circuit, 19 is a processor, 2o is a clock, 21 is a memory, 22 is a TOD value, and 23 is a flag.

[Brief explanation of drawings]

FIG. 1 is a block diagram explaining the present invention in detail, FIG. 2 is a block diagram of a circuit showing an embodiment of the present invention, and FIG. 3 is a flow chart explaining the operation of FIG. In the figure,

Claims (1)

[Claims] A plurality of host devices (1
) (2) and an input/output device (3) shared by the plurality of higher-level devices (1) and (2), the input/output device (3) includes a reference clock (4) and , a storage means (5) for storing a reference time given from the higher-level device and the validity of the reference time, and when one higher-level device records the trace information, the input/output device (3) If the reference time to be stored is invalid, the input/output device (3) stores the reference time based on its own clock, and also stores the validity of the reference time and resets the reference clock (4). After that, recording of trace information is started, and when one higher-level device records the trace information, if the reference time stored in the input/output device (3) is valid, the input/output device (3) 3) When recording trace information, the recording of the trace information is started after reading the reference time stored in the reference clock and the time of the reference clock (4) and correcting the time of the own clock. Time correction method.