JPS61117633A - Capture ratio measuring system - Google Patents

Abstract

PURPOSE:To measure an occupation time ratio of each job from the total CPU time, which the operating system obtains on a basis of identifiers stored in a trace buffer, by providing a hardware tracer in the CPU. CONSTITUTION:A hardware tracer 13 is provided in a CPU 1, and identifiers (address space discriminating information) 33 stored in a main storage device MSU 3 are sampled each time when an operating system OS 31 switches a job JOB 32. That is, the tracer 13 performs sampling at a specific period independently of the software operation and stores identifiers in a trace buffer 34 which the OS 31 prepares on the MSU 3. The OS 31 reads identifiers 33 stored in the buffer 34 to measure the total CPU time. Thus, since identifiers of jobs 32 which are operated in the interrupt inhibiting state can be sampled, the occupation time ratio of each job 32 is measured with a high precision.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to the ratio of the time each job occupies the central processing unit (CPU) of a data processing device to the total CPU time;
That is, it relates to a method for accurately measuring cabture ratio.

Generally, in an online data processing device, each job is executed under the management of an operating system (OS).

In this case, for example, the ratio of the time occupied by the central processing unit (hereinafter referred to as CPU) to the total CPU time for execution of a user job [this is referred to as user CPU occupancy rate (capture ratio)] By looking at
It is possible to know the operating status of the data processing apparatus, in other words, the amount of overhead time required for each user job.

For example, if the cabture ratio is large, it means that the overhead of the job is small.

Therefore, the accuracy of measuring the CPU usage rate of the user must be as high as possible.

[Conventional technology]

FIG. 3 is a diagram schematically showing a mechanism for calculating user CPU occupancy using a conventional method, in which l is a central processing unit (CPU), 2 is a memory access control unit (M
MC) is in charge of memory access control when CPU l accesses MSII 2, and 3 is the main memory (hereinafter referred to as
? It is called lSυ).

In the present invention, at least the CPU l is provided with a hardware timer 11 and an interrupt timer 12.
ζ, MSU 3 has an operating system (O
S) 31 and each job (JOB) are stored.

In a data processing device with such a hardware configuration, when determining the user CPU occupancy rate (cab chair ratio), the CPU occupancy time (!
The operating system (OS) running on CPt11 (referred to as CPU time) is the CPU time for each job (JOB
) The total CPU time can be measured accurately by reading the hardware timer 11'c at the beginning and end of the execution of 32. At this point, the address space identifier # information (!ll identifier) for each user stored in the fixed area 33 of MSu 3 is sampled, and it is calculated by multiplying the number of sampled identifiers and the sampling period. Ta.

[Problem that the invention seeks to solve]

Therefore, in the above conventional method, each job (JOB)
If interrupts are disabled during the execution of 32 (often implemented during paging processing, etc.), the timer interrupts for sampling the address space identification information (J alias) for each user will be disabled. Therefore, the exact total C
There is a problem in that the Pυ time cannot be measured, and the measurement accuracy of the rate inside the user CPU is reduced.

In view of the above conventional drawbacks, the present invention provides the above address space identification information (!a
The purpose of this invention is to provide a method that can sample ti identifiers).

[Means for solving problems]

The purpose of this is to install a hardware tracer in the central processing unit (CPU) and use the hardware to identify the job-corresponding address space identification information (identifier) that the laser stores in a specific area (trace buffer) on the main memory. is achieved by the :1-yapture ratio negotiation method of the present invention, which measures the cab chair ratio for each job from the total CPυ time collected by the operating system (O5).

[Effect]

That is, according to the present invention, the operating system (O5) of the data processing device provides address space identification information (identifier) in the CPU to be stored in a fixed area on the MSII every time the job (JOB) is switched. The hardware tracer samples the operating system (O
Since the operating system (OS) measures the total CPU time by reading what is stored in the trace buffer prepared by S) in the trace buffer prepared on the MSII, jobs running with interrupts disabled Since the above-mentioned identifier of (JOB) can be sampled, there is an effect that the measurement accuracy of cab chair radio can be improved.

〔Example〕

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

FIG. 1 is a diagram schematically showing an embodiment of the present invention, and FIG. 2 is a flowchart showing the operation when calculating the total CPU time by implementing the present invention.

In FIG. 1, the same symbols as in FIG. 3 indicate the same objects, 13.34 are necessary for carrying out the invention, and 1
3 is a hardware tracer, and 34 is a trace buffer.

The trace buffer 34 is prepared on the MSU Z by the cab chair regio measurement software described later, and the address space identification 18 report corresponding to the job (JOB) is set in the fixed area 33 on the MSU a. It is set by the operating system (O5) when switching the CPU control right.

When the measurement of the total cpυ time is started, the measurement software prepares the trace buffer 34 on the ISU 3, and stores its address, size, and trace period, as well as the address space identification information field address on the hardware. After notifying the hardware tracer 13, it requests the hardware tracer 13 to start the tracer.

The measurement software uses the hardware to collect the address space identification 1 information stored in the trace buffer at a cycle when the trace buffer 34 becomes full, and restart the trace from the beginning of the trace buffer 34. Ask Tracer 13.

The hardware tracer 13 is set in the fixed area 33 at specified time intervals using an internal timer.
The address space identification information ζ is sequentially traced to the trace buffer 34.

Next, the details of the operation of the measurement software will be explained with reference to the flowchart (a) in FIG. 2.

Step 40: Prepare the trace buffer 34 on H5 (13).

Step 41 Instructs the hardware tracer 13 to start measuring the total CPU time. At this time, the trace period, address space identification information field address, address and size of the trace buffer 34 are notified.

Step 42: Wait for a certain period of time until the tracer buffer 34 becomes full.

Step 43 Instructs the hardware tracer 13 to finish the above measurement.

Step 44; Collect and analyze the address space identification information stored in the trace buffer 34, and
Find the u time.

Step 45: Look at the instructions from the operator or the control statement of the measurement software to see if it is OK to end the measurement. If the instruction is to end the measurement, step 46
However, if it is necessary to continue the measurement, the process returns to step 41.

Step 46: Return (cancel) the trace buffer 34.

The above is the details of the operation of the measurement software.

Next - C++ by operating system (O5)
The switching operation of the U control right will be explained using (b).

Step 50; When each user type (JOB) is switched by the operating system (OS), the job (JOB) is stored in the fixed area 33 on the MSυ3.
Set address space identification information corresponding to .

By sampling this address space identification information at regular intervals, the CPU occupation time of the job can be determined.

Next, regarding the operation of the hardware tracer 13, (c
).

Step 60: When the hardware tracer 13 is instructed to start tracing in step 41, it waits for the designated time using an internal hardware timer.

Step 61: Copy the address space identification information i stored in the fixed area 33 on MSυ3 to the trace buffer 34 at each specified time.

Step 62; Step 43 in the measurement software
It is checked whether there is an instruction to end the measurement, and if the instruction is an instruction to end the measurement, the tracing operation is ended, but if it is not an instruction to end, the process returns to step 60.

In this way, c
The total CPU for each user type is
You can ask for time. That is, the total C for each user type
P[1 hour - measurement elapsed time x (number of traces for each user type/total number of traces)].

Note that the trace period specified in step 41 is a factor that determines measurement accuracy.

Further, as a method for the hardware tracer 13 to access the Isu 3, there are a known method using memory stealing, a method at a predetermined timing, etc., but considering the gist of the present invention, it is not possible to use a specific method. Needless to say, it is not limited to.

〔Effect of the invention〕

As described above in detail, the capture ratio measurement method of the present invention allows the operating system (O5) of the data processing device to store data in a fixed area on the MSIJ for each user each time the operating system (O5) switches jobs (JOB). A hardware tracer installed in the CPU samples address space identification information (identifier) at a specific cycle, independent of software operations, and stores it in a trace buffer prepared by the operating system (OS) on the MStl. the operating system (OS)
Since it is designed to measure the total CPυ time by continuing, the above identifier of a job (JOB) running with interrupts disabled can be sampled.
This has the effect of improving the measurement accuracy of cab chair radio.

[Brief explanation of the drawing]

FIG. 1 is a diagram schematically showing an embodiment of the present invention. FIG. 2 is a flowchart showing the operation when implementing the present invention and measuring CPU time. FIG. 3 is a diagram schematically showing a measurement system for measuring total CPU time using a conventional method. In the drawing, l is a central processing unit (CP[I), and 11 is a hardware timer. 12 is an interrupt timer. 13 is the hardware tray. 2 is a memory access control unit (MMC). 3 is the main memory (MSIJ). 31 is an operating system (OS). 32 is a job (JOB), 33 is a fixed area. Reference numeral 34 indicates a trace buffer 1; 40 to 46, and 50 to 62 indicate each step 1 of the cab chair regio measurement operation. 1st cake

Claims (1)

[Claims] To measure the time ratio (capture ratio) occupied by each job on the central processing unit (CPU) of a data processing device,
S) is address space identification information (
Total CP calculated by counting the number of samples (identifier)
This method is calculated from the U time, and the central processing unit (C
A hardware tracer is provided in the PU), and the operating system (OS) collects the address space identification information (identifier) stored in a specific area (trace buffer) on the main storage by the hardware tracer. A capture ratio measurement method characterized by measuring the capture ratio of each job from the determined total CPU time.