JP2618526B2 - Acquisition method of measured data - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] Regarding a method of acquiring measured data used for simulation of a disk cache device, the frequency of occurrence of data, such as when a wait time occurs when measuring the usage time of a disk device, In the case where large fluctuations are expected, the purpose is to obtain measured data accurately and flexibly, and (elapsed time / unit time) + (IO count / unit count) is constant to acquire the measured data Is configured to sample data in a period until the reference value is reached.

[Industrial applications]

The present invention relates to a method for acquiring measured data used for simulation of a disk cache device.

In recent years, the processing power of CPU has improved remarkably,
In comparison with this, it has been regarded as a problem that the processing capability of the disk device has not been improved so much.

Disk cache devices have been developed to solve this problem. Although a disk cache device can be expected to greatly reduce access time when used properly, it is expensive and therefore poses a large risk in its introduction.

Therefore, it is necessary to sufficiently study the effect of introducing the disk cache device before the disk cache device is introduced, and a simulation is performed on the assumption that the disk cache device is used.

In order to perform a simulation, data that is the basis of calculations during the simulation is required. When acquiring the data by actual measurement, how to determine the period of the acquisition operation is extremely important.

[Conventional technology]

2. Description of the Related Art Conventionally, there are the following methods for sampling and acquiring measured data.

First, data acquisition operation is performed until A = elapsed time / unit time reaches a certain reference value S1. This is referred to as an A method.

For example, if the unit time is 1 second and the reference value S1 is 1, the acquisition operation is performed during 1 second, and the data generated during that time is sampled.

The other is to perform the data acquisition operation until B = the number of samplings / the number of units reaches a certain reference value S2. This is referred to as the B method.

For example, when the unit number is set to 10 and the reference value S2 is set to 1, the acquisition operation is performed until the sampling number reaches 10.

[Problems to be solved by the invention]

However, in the above-described conventional A method, sampling is performed only at a predetermined fixed elapsed time. Therefore, if the fluctuation period of the data generation frequency is longer than the elapsed time, the number of samplings, that is, The number of measured data to be measured greatly differs depending on the timing at which the acquisition operation is performed.

For example, the number of times of sampling may be extremely small, or the acquisition operation may end without any sampling being performed, so that the reliability of the measured data is extremely low.

In addition, in the above-described B method, the acquisition operation is continued indefinitely until a predetermined number of samplings is reached, so that when there is no data for an extremely long time,
The acquisition operation time becomes very long, and it is not possible to cope with a change in the frequency of data generation during the acquisition operation.

For example, if the number of samplings due to many data that does not reach the unit number occurs in a very short time and no data occurs before and after that, many data generated in a short time will be averaged over a long time. Will be done.

When performing a simulation using actual measurement data including time data, an important factor is the usage rate of the apparatus.

The usage rate of the device is a value obtained by dividing the usage time during which the device is used within a certain reference time by the reference time.For example, to determine the usage rate of the disk device,
Each time an access instruction (IO instruction) to the disk device is issued, the usage time of the disk device is measured.

However, the processing speed of the disk device is slow because a mechanical seek operation is performed. For this reason, a relatively long time (waiting time) may occur while the processing by the previous IO instruction is being executed and the access to an accessible state is being performed even though the IO instruction has been issued. Often there.

When the waiting time occurs, since no IO instruction is issued during that time, a state in which data sampling is not performed for a relatively long time continues.

Therefore, when the actual measurement data on the usage time is obtained by using the above-mentioned method A or B, the drawbacks of the respective methods appear, and a large error is included in the usage rate obtained based on this. Become.

The present invention has been made in view of the above-described problems, and in a case where a waiting time occurs when measuring the use time of a disk device,
It is an object of the present invention to accurately and flexibly obtain actual measurement data when a large change is expected in the frequency of occurrence of data.

[Means for solving the problem]

In order to solve the above-mentioned problem, the method according to the first aspect of the present invention, as shown in FIGS. 1 to 4, obtains measured data 31 by (elapsed time / unit time) + (IO count) Data is sampled during a period MT until (÷ unit number) reaches a certain reference value.

In the method according to the second aspect of the present invention, the unit time and the unit number are set according to the weight of the elapsed time and the IO number.

According to a third aspect of the present invention, the reference value is set to one.

(Operation)

The period MT is determined by a combination of time and the number of times.

In other words, until the elapsed time reaches the unit time, or until the IO count reaches the unit count, the maximum value of the period MT is reached. When the sum of the ratios to the maximum value reaches the reference value, one data acquisition operation ends.

〔Example〕

 Hereinafter, embodiments of the present invention will be described with reference to the drawings.

In the present embodiment, a case will be described in which the actually measured data 31 is acquired for the simulation of the disk cache device 13.

In the simulation of the disk cache device 13, various data on the disk device 12 currently used and the disk cache device 13 to be used are used, and the average access time required to execute an IO instruction issued from the processing device 1 is used. However, it is required that the use of the disk cache device 13 reduce the time.

FIG. 2 is a block diagram showing the processing system 2 before the disk cache device 13 is introduced, and FIG. 3 is a cache processing system 3 after the disk cache device 13 is introduced.
FIG.

The processing system 2 includes a processing device 11 such as a host computer, and a disk device 12 connected to the processing device 11 and storing a plurality of files (data sets). Device 12
Read / write (access) to or from a predetermined file or track. Usually, a plurality of disk devices 12
Is connected.

In the cache processing system 3, a disk cache device 13 is connected between a processing device 11 and a disk device 12. The disk cache device 13 includes a controller 21, a cache memory 22, an NVS (non-volatile memory) 23
And the IO instruction from the processing device 11 is
The decrypted data is decrypted by 21 and a predetermined access is made to the cache memory 22, the NVS 23, or the disk device 12.

In the current processing system 2, in order to predict the effect of the cache processing system 3 using the disk cache device 13, an access operation simulation of the processing device 11 in a state where the disk cache device 13 is connected is performed. .

FIG. 4 is a diagram for explaining a simulation of the disk cache device.

As shown in FIG. 4, measured data 31, basic data 32,
The simulator 36 receives the disk cache device configuration specification data 33 and the cache target specification data 34 as input data.
Is performed, and the result is output as a report 41.

The measured data 31 is stored in the disk cache device in the processing system 2 while the trace program is running.
It is obtained by executing a processing operation to be actually performed after the introduction of the storage device 13 and recording (storing) the access status to the disk device 12 at that time.

As the actual measurement data 31, the type of the IO instruction issued from the processing device 11, the time at which the IO instruction was issued, the number or name of the disk device 12 accessed by the IO instruction, the address (CCHHR) on the disk device 12, the access Data amount, number of records, number of IOs (number of IO instructions issued),
There are waiting time, data transfer time, access time, and the like.

The actual measurement data 31 is recorded for each IO command issued from the processing device 11 within a fixed period MT in which a fixed unit time and a unit IO count are combined. The period MT will be described later in detail.

The types of the IO instruction include a R / W instruction, a search instruction, a sense instruction, and other instructions for accessing the disk device 12.

The basic data 32 is data relating to the time required for processing in the disk cache device 13.

As the basic data 32, the device transfer speed of the disk cache device 13, the channel transfer speed, the read time, the write time, the time required for decoding the IO instruction, the time required for determining the cache target, the time required for hit determination, and other processing. Time, the rotation speed of the disk device 12, and the like.

The disk cache device configuration designation data 33 is data for designating the configuration of the disk cache device 13 to be simulated.

The number of channel adapters in the controller 21 (the processing device 11
Number of adapters for connecting to the disk device 12, number of device adapters (number of adapters for connecting to the disk device 12), capacity (size) of the cache memory 22, capacity of the NVS 23, number of channel switches (for a plurality of Number of switches to connect).

The cache target designation data 34 is stored in the cache memory 22.
This is the data for specifying the target to be placed in.

Disk device 12 as cache target designation data 34
Number, name, file name, address (CCHHR) on the disk device 12, and the like.

The simulator 36 is a program executed on the processing system 2. The simulator 36 assumes a state in which the disk cache device 13 is connected in the processing system 2, that is, assumes the configuration of the cache processing system 3,
With respect to the instruction included in 31, the access operation to the disk device 12 and the disk cache device 13 is simulated.

The report 41 includes a current status report for the current processing system 2 in which the disk cache device 13 is not installed, and a simulation report for the cache processing system 3 in which the disk cache device 13 is installed.

FIG. 1 is a diagram showing periods divided according to the number of IOs.

FIG. 1A is a diagram showing the state of the number of IOs with respect to the time axis.

According to FIG. 1A, it can be seen that the number of IOs issued from the processing device 11 greatly varies depending on the time (time).

FIG. 1 (b) is a diagram showing periods MT divided by the acquisition method according to the present invention.

In FIG. 1 (b), every time the condition shown by the following expression (1) is no longer satisfied with respect to the issue status of the IO instruction shown in FIG. MT is separated.

(Elapsed time ÷ 4) + (IO count ÷ 200) <1 (1) That is, the maximum elapsed time is 4 seconds, the maximum IO count is 200, and the data acquisition operation is started. After that, when the sum of the ratios to the respective maximum values reaches 1, between the respective maximum values, the data acquisition operation is ended once.

During each acquisition operation period MT1,2,3 ...
Each time an instruction is issued, an operation of acquiring the number of IOs and various other data associated with the IO instruction is performed.

As shown in FIG. 1 (b), when the period MT of the acquisition operation is determined based on the above equation (1), the period MT of the acquisition operation does not become unnecessarily long even in the period where the number of IOs is small. On the contrary, even during the period in which the number of IOs is large, the period MT of the acquisition operation is not shortened unnecessarily, so that data acquisition is performed accurately.

On the other hand, FIG. 1 (c) also shows FIG. 1 (a)
FIG. 11 is a diagram showing the issuance status of the IO instruction shown in FIG. 7 in which the acquisition operation period mt is divided by the conventional A method shown in the following equation (2).

(Elapsed time ÷ 2) <1 (2) FIG. 1 (d) shows a conventional B shown in the following equation (3).
FIG. 11 is a diagram showing a period mt of an acquisition operation divided according to a method.

(IO count ÷ 100) <1 (3) As shown in FIGS. 1 (c) and (d), in the A method, since the periods mt are all the same, while the IO count is small, In, the number of times of data sampling becomes extremely small. In the B method, the number of samplings is
Since each period mt is extended until it reaches 100 times, the period mt becomes extremely long while the number of IO times is small.

On the other hand, in the method according to the present invention, the method A and the method B
Since the period MT is determined by assigning a weight of 0.5 to each method, these disadvantages are compensated for each other,
The number of times of data sampling is not extremely reduced, and the period MT is not extremely long.

FIG. 5 is a diagram showing an example of a report 41a showing the result of the simulation.

The report 41a includes the number (volume number) VOL of the disk unit 12, the average utilization ratio UCB of the disk cache unit 13 (including the disk unit 12) viewed from the processing unit 11, the actual average utilization ratio DEB of the disk unit 12, Total number of IO instructions issued from the processing unit 11 (IO count) SIO, hit rate HT
R, BFR before use and AT after use of disk cache device 13
For each of R, the ratio RSE at which the actual seek operation occurred,
An average seek distance ASE (the unit is the number of tracks) in one seek operation, an average response time RTM for one IO command, and the like are shown.

The average usage rates UCB and DEB are values obtained by dividing the time actually required for executing the IO instruction in the period MT by the period MT.

The time required to execute the IO instruction is the time required to decode the instruction, the time to determine whether or not to hit, the time to read data from the cache memory 22 or the disk device 12,
This is the total time of transferring the read data to the processing device 11, waiting time for reading the data from the disk device 12, and the like. However, write back from the NVS 23 to the disk unit 12 and cache memory from the disk unit 12
The time required for prefetching to 22 is excluded.

The ratio RSE is a value obtained by dividing the number of IO instructions in which a seek operation has occurred by the total number of IO instructions (IO times) SIO, and the response time RTM is obtained by dividing the total access time by the total number of IO instructions SIO. Value.

According to report 41a, the ratio RSE at which the actual seek operation occurs
From 1.9% to 0.3%, response time RTM from 16ms to 1
In this case, ms is greatly reduced, and it can be seen that there is a great effect in shortening the access time to the disk device 12 having the number VOL of “00”.

FIG. 6 is a diagram showing another example of the report 41b showing the result of the simulation.

This report 41b is for the disk device 12 whose volume number VOL is "00", as indicated in the title.

The report 41b includes the file name FNM, the number of tracks TRA indicating the size of data, the total number of IO instructions (IO count) SIO for each file, and the number of hits R when reading.
HT, hit rate HTR, response time BFRRTM before use of the disk cache device 13, response time ATRRTM after use, and the like are shown.

According to the report 41b, especially for the file name FNM “F4”, the number of IOs is large, the hit rate HTR is high, and the response time RTM is also shortened.

According to the above-described embodiment, the actual measurement data 31 is acquired for the currently used processing system 2 and the processing contents, and the simulation is performed based on the actual measurement data 31. Therefore, when the disk cache device 13 is introduced and the business is actually performed, it is possible to easily and accurately know the effect of shortening the access time and the like.

Further, the period MT of the acquisition operation is determined based on the above equation (1).
However, even during the period where the number of IOs is large, the period MT is not shortened unnecessarily, and the data is accurately acquired. Since the simulation is performed based on the reliable measurement data 31 obtained in this way, the simulation results such as the average utilization UCB and DEB are accurate,
It will be reliable.

In the above embodiment, the unit time is 4, and the unit number is
Although 200 and the reference value are 1, these numerical values can be variously changed. For example, if the weight of the elapsed time is to be increased, the unit time is reduced or the unit number is increased, and conversely, if the weight of the IO number is to be increased, the unit time is increased or the unit number is increased. What is necessary is just to make it small. Since these weights change depending on the issue status of the IO instruction, the processing system 2 actually used
May be flexibly determined according to the issue status of the IO instruction.

The present invention can also be applied to acquisition of actual measurement data related to IO instructions for an external storage device other than the disk device 12, for example, a floppy disk device, a magnetic tape device, and the like.

[Effects of the Invention] According to the present invention, it is possible to accurately acquire measured data when large fluctuations in data generation frequency are expected, such as when a wait time occurs when measuring the use time of a disk device. It can be done flexibly.

[Brief description of the drawings]

FIG. 1 is a diagram showing periods divided according to the number of IOs, FIG. 2 is a block diagram showing a processing system before introducing a disk cache device, and FIG. 3 is a cache process after introducing a disk cache device. FIG. 4 is a block diagram showing a system, FIG. 4 is a diagram for explaining a simulation of a disk cache device, FIG. 5 is a diagram showing an example of a report showing a result of the simulation, and FIG. 6 is a report showing a result of the simulation. It is a figure showing other examples of. In the figure, 31 is actual measurement data, and MT is a period.

Claims (3)

(57) [Claims] In order to acquire actual measurement data (31), data is sampled in a period (MT) until (elapsed time / unit time) + (IO count / unit count) reaches a predetermined reference value. Acquisition method of measured data characterized by the following. 2. The method according to claim 1, wherein the unit time and the unit number are set in accordance with the weight of the elapsed time and the number of IOs. 3. The method according to claim 1, wherein the reference value is set to 1.