JPS583047A - Load factor data displaying method of data processing system - Google Patents

Abstract

PURPOSE:To easily obtain a load factor data of each processor, by constituting a system so that an operator or a program gives an instruction suitably to each processor. CONSTITUTION:A program which runs in case when no processing to be executed by each processor 1, 2 and 3 exists, and sets a load factor of a corresponding processor to the maximum value when it is executed independently, and a counter are made resident in main memories 4, 5 and 6. The program is started when an operator gives an instruction from a display and operation panel 7. When the program has been started, it repeats its execution unless other processing exists, the counter is driven at its every constant repetition, and whenever its driving is executed, contents of the counter are displayed on the display and operation panel 7.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for measuring and displaying a processor load factor in a data processing system. In particular, in a data processing system consisting of a single or multiple processors, a main memory, peripheral devices, and a display/operation panel, the present invention is capable of reducing the amount of time that a processor spends working on a task in response to an external processing request. This invention relates to a method of displaying data obtained by measuring during non-operating time.

Conventionally, the following three methods have been used to determine the processor load factor.
Nineteen things are used.

(1) When the processor runs out of work to process, the clock device reads the time, and when the next task to process occurs, the processor reads the time again from the clock device, and calculates the difference between this and the previous time. Display in addition to. Counter display f after a certain period of time: A method of reading and calculating the load factor.

(2) Interrupt the processor at appropriate time intervals,
It is determined whether the processor was processing something immediately before the interrupt was issued, and if the processor was processing something, the first
The counter is incremented by +1, but if no processing is being performed, it is left as is. On the other hand, at the same time, the second counter O is
, +1 count, 0I10* at a certain time
How to calculate the load factor.

(3) Regarding the processing time when a program that uses only the processor is given to the system and is processed independently within the system, and the processing time of the above program when it is processed in parallel with other tasks. However, method (1) requires that the time display of the clock device is sufficiently accurate down to the minute time.
It cannot be applied to all processing devices; the method of (2) requires processing to generate one interrupt, and the overhead time for this is a hindrance. And (3) has the disadvantage that one task must be input separately and cannot be measured under the normal usage environment of the system.

Furthermore, in a system such as an asymmetric multiprocessor where each processor has a different function, measuring the load factor of each processor requires a signal line indicating whether the processor is in operation or not. It is necessary to take it out and connect it to an external measuring device, and the preparation work for that is troublesome. In cases where the processing time for a job is short and the overhead for job replacement is large, the utilization rate of each processor is not necessarily a guideline for determining the system's margin. It is our goal to provide a method for measuring and displaying processor load factor data that does not require the overhead of interrupts, does not require a precise clock device, does not change the system usage environment, and also provides a good indication of the processor's actual operating margin. be.

A further object of the present invention is that the work input into the system is sequentially assigned to the same processor, the processor continuously executes the work according to instructions stored in the main memory, and at a certain time the current work is The intermediate state of the first job is written to the main memory, and the intermediate state of the first job is read and executed. Particularly, in a system configured with a plurality of processors, it is possible to easily obtain load factor data for each processor by giving appropriate instructions to each processor by an operator or a program.

The present invention has the following configurations: a single or multiple processors, a main memory that is subordinate to each processor and stores instructions to be executed by each processor, and a system worker who monitors the system status.
? - In a data processing system that has a display and operation panel for issuing commands to the system, each processor runs when there is no processing to be performed, and calculates the load factor of the corresponding processor when it is executed independently. A program and a counter to maximize the maximum are resident in the main memory, and the program is started when an operator gives an instruction from the display/operation panel, and when started, the program is executed unless there is other processing. is repeated, and the counter is incremented at each constant repetition.

There is provided a load factor data display method characterized in that the contents of the counter are displayed on the display/operation panel each time the step is performed.

FIG. 1 shows a schematic configuration of a multiprocessor type data processing system in which the method of the present invention can be implemented. , P, and 4.5, 6 are the main memories MM, , M4 ., subordinate to the processors P1 s P*bP, respectively. The MM is a display/operation panel for monitoring the system and issuing necessary operational input instructions. Processor P is in charge of arithmetic processing, and processor PIs Pa is in charge of input/output processing that occurs during the execution of processing by processor P, and is activated by p when necessary.

Each processor PIs P** Pl has a one-processing schedule control function.

Figure 2 (The tolerance indicates the contents of the processing unit management table provided for each processor used for this purpose.
The example of the processor P1 is shown in FIG.

Currently there are 4 1 locations 3M from 1 processing 10' to 1 processing 13'
is given as a task for the processor. From 1st process 11” to 1st process 13# of 2nd wA(a),
This is a task imposed on the system from the outside. @Processing 10'' has execution control of the load factor measurement program prepared in the processor in the present invention.Each processor updates its respective management table when one process being executed is interrupted. It is controlled to search, execute the next 1 process, and execute ``process 10'' when there is no 1 process to be done.
given priority. ``Processing 10'' is interrupted if there is an interrupt to the processor and has to yield to another process 1. In order to make it possible to recover from the interrupted state at that time, as shown in FIG. 2(b), Instruction address at the time of suspension in the management table.

The values of registers within the processor are stored.

1 process 10'' starts execution when another process 10'' is in a suspended state or has finished. Therefore, one process 101 is
If the processor is not added to the processor's processing target according to an operator's command, the processor occupies the period in which it was originally inactive, which is K.

FIG. 3 shows an outline of the load factor measurement program of "Processing 10". When the program starts running, it first clears the counter. Certain operations are then performed. This calculation may be of an appropriate length that can be executed by only the processor (in this case, Ps), and the calculation processing time does not change depending on the number of repetitions. This operation is
This is repeated until a certain number N is reached. When N is reached,
The counter is incremented by +1 and its contents are displayed on the display section of the display/operation panel 7.The display disappears after a certain period of time, but other methods may be used to notify the change in the counter contents, and then the contents are displayed again. The operation is repeated as before. 1 process 10# cannot execute the other 1 processes 11' to 1 process 13#, but when the execution is completed, that is, the execution control is taken over for all the times when the processor P1 is normally idle. ,in the end,
The counter display will show a value proportional to the inactive time of processor P1.

FIG. 4 shows a ninth time chart illustrating the present invention in detail for the case where the processor is performing the compiling process. FIG. FIG. 4(b) shows the time when the load factor measurement program (@processing 10') is started and runs when the processor becomes idle due to preparation, interruption, and termination of the compilation process. This shows a situation where the counter is incremented from q to q.

FIG. 3(C) will be described later.

FIGS. 5(b) and 5(bl) are flowcharts showing the measurement work performed by an operator to obtain load factor data according to the method of the present invention.

Figure 5 (hot) shows the measurement work when the processor PR to be measured is processing a certain program given from the outside. to start the charge rate measurement program. Pursessa Karasu completes the measurement environment conditions,
That is, upon completion of inputting the source program, the compiling process is stopped by a command from the compiler. As a result, the processor P executes the load factor measurement program and displays the value q of kakunta. At that time, the operator starts the stopwatch and records the counter value as the initial value. Thereafter, the operator cancels the stopped state of the compiler, processing continues, and when the calculations in the load factor measurement program are repeated and reach a certain number of times, a counter is displayed. After completing the compiling process, the operator needs to prepare a comparison standard for the measurement results shown in Figure 5 (-).
The measurement work shown in is performed. In the case of FIG. 5(b), the measurement is performed in a state where the processor Pl does not have any program for one process from the outside.
This is exactly the same as in Figure (1); first, a counter display appears, and at 9, the stop count is started and all records are recorded to that value. Next, when the counter is displayed, the stopwatch is stopped and the counter display is recorded together with the value f.

by the way! One period T is as shown in FIG. 4 (0).

Since it is the sum with Durham's execution time, the load factor ρP1 is expressed as follows.

T THere, since the load factor measurement program execution time per one count of the load factor data display caranta is expressed as T'10;-dt, the load factor measurement program time is T'(Ol-Ol
) / (hehe), and by substituting this into the above equation for ρν, we can obtain the following equation ρpIT(T - T/"a double) = 1 On-0/. Based on the previous measurement results, By calculating the above formula, the load factor can be easily determined.

Part 6 shows a schematic configuration of another multiprocessor type data processing system in which the method of the present invention can be implemented. In Figure 2, processor P receives work ('
Processor Pls PI
is invoked by PK based on needs that arise during the execution of P. Processor P.

does not have the function to manage and control its own tasks. When Pl finishes one task, the task control function 't-4
P, and receives instructions for the next task.

The processor P also has a display function. P
, performs seven other I10 processes, and processor P
ls PIs Pl is configured to allow parallel processing.

The main memory M contains a program PG as shown in FIG.
・1, PO2, PO2 are stored. These programs PGI, PG2, and PO2 correspond to the Marugame cart measurement program described in the explanation of the embodiment shown in FIG. However, program PGI is capable of saturating Karasu itself when executed by processor P, and program PG2, when executed by processor P1, starts processor P and saturates Karasu itself. can be saturated. Similarly, program PG
3 is capable of activating processor P and saturating P when executed by processor P.

Each of the programs PGI to PG3 has a corresponding counter, and each time a certain process is repeated a certain number of times, the counter is incremented by +1, and the contents of the counter are displayed on a display/operation panel (not shown) of the system. The operator selects a program PGI to PG3 corresponding to the processor whose load factor is to be measured, gives a command from the display/operation panel, and executes the program. The subsequent steps are as follows:
It is the same as shown in the figure.

Although the invention has been described in terms of two preferred embodiments, it can be implemented in many other forms as well! Ruchipuwosetsa・
It is obvious that the present invention can be easily applied to a single processor system and is also effective in a single processor system.

[Brief explanation of drawings]

FIG. 1 shows a schematic configuration of a multiprocessor type data processing system in which the present invention can be implemented. FIGS. 2(a) and 2(b) show processing unit management tables. FIG. 3 is a schematic flow diagram of the load/rate measurement program. FIG. 4 is a timing diagram showing an example of execution of the load factor adjustment program. The 5th WA is at 70% for load factor measurement work.
-Illustration. FIG. 6 shows a schematic configuration of another i-multiprocessor type data processing system in which the present invention can be implemented. FIG. 7 shows the arrangement of a load factor measurement program used in the embodiment shown in FIG. show. In the figure, 1 to 3 are processors, 4 to 6 are main memories, and 7 is a display/operation panel t-. Patent Applicant: Fujitsu Ltd. Agent/Patent Attorney Mori 1) Genzai 2kake Sai°7 Saijizu (4) (b)

Claims (2)

[Claims] (1) A single unit has a plurality of processors, a main memory subordinate to each processor for storing instructions to be executed by each processor, and a main memory that is used by a system operator to monitor the system state and to operate the system. In a data processing system having a display/operation panel for issuing commands, a program that runs when each processor has no processing to perform and maximizes the load factor of the corresponding processor when executed independently. and a counter resident in the main memory, the program is started by an operator's instruction from the display/operation panel, and when the program is started, it repeats execution unless there is other processing; A method for displaying load factor data, characterized in that the counter is incremented at each constant repetition, and the contents of the counter are pressed to be displayed on the display/operation panel each time the increment is performed. (2) The load factor data display method according to claim (1), wherein the value of the counter can be read from another program that is being executed simultaneously.