WO2013145629A1 - Information processing device for executing load evaluation and load evaluation method - Google Patents

Abstract

The present invention provides an information processing device that is capable of evaluating whether a desired load is being generated for application to a system that is a target for testing. The information processing device is provided with: a means that calculates the request rate per unit time on the basis of time series data for requests sent from a load generation device, and aggregates the request rates on the basis of predetermined class data in order to create frequency distribution data; a means that acquires expected frequency distribution data, said data being the frequency distribution data that is expected for a case in which it is assumed that the distribution of request rates follows a desired probability distribution; and a means that uses observed frequency distribution data that has been created and the expected frequency distribution data that has been acquired in order to carry out goodness-of-fit testing of the expected frequency distribution data with the probability distribution of the observed frequency distribution data.

Description

Information processing apparatus and load evaluation method for executing load evaluation

The present invention relates to a load test for a system or a server, and relates to a technique for evaluating whether a desired load can be generated for a test target.

A method called a load test is known as one of the methods for confirming the behavior of a system or server under high load. In the load test, a simulated load is applied to the target system or server.

In many cases, a load test is performed by a load test execution device using software called a load test tool. The load test execution device applies a load to the test target system by transmitting a request to the test target system. The test target system processes the request received from the load test execution device and responds to the load test execution device.

As the test target system, for example, a server of an online shopping site can be cited. Hereinafter, what kind of load is applied to the server during the actual operation of the online shopping site will be described.

There are an unspecified number of customers on the online shopping site. When each customer arrives at the online shopping site, the customer repeatedly sends a request and receives a response to the request. The request transmitted by the customer to the server is, for example, browsing or selecting a product, checking a cart, inputting a product delivery destination or a credit card number, and the like. The customer receives a response (for example, text describing the product, an image of the product, a video of the product, etc.) to the request transmitted by the customer. After that, the customer confirms the screen of the displayed product or thinks to select the product, and transmits the next request again.

The customer repeats the transmission of the request and the reception of the response, and completes the browsing of the online shopping site when the purpose of purchasing the target product is achieved.

In a system that receives requests from an unspecified number of customers, such as an online shopping site, it is known that the arrival interval of requests follows an exponential distribution. Here, the arrival interval is a time interval from the arrival of a certain request to the arrival of the next request. It is known that when the request arrival interval follows an exponential distribution, the request arrival rate (the number of request arrivals per unit time) follows a Poisson distribution.

Patent Document 1 discloses an example of a load test execution device. In the load test execution device disclosed in Patent Document 1, the operator of the load test execution device inputs the number of requests processed per unit time to be processed by the test target system as a performance target when performing the load test. The load test execution device automatically adjusts various parameters at the time of performing the load test so as to satisfy the performance target input by the operator, and transmits the request.

Patent Document 2 discloses a method for determining the load capacity of a server. The method disclosed in Patent Document 2 includes the following steps. In the first step, the method simulates the behavior of the virtual user and sends a request from the client to the server. In the second step, the method measures a metric that is a performance indicator of the server. In a third step, the method compares the measured metric with a predetermined value and changes the number of virtual users based on the difference between the metric and the predetermined value until the metric is within a predetermined range of the predetermined value.

Patent Document 3 discloses a distribution suitability test apparatus. The distribution suitability test apparatus disclosed in Patent Document 3 includes a totaling section determination unit, a totaling unit, and a suitability test unit. The tabulation section determining means determines the width of the tabulation section in which the measured data is tabulated so that the expected frequency has an equal probability. The totaling means totals the number of data of each determined total section. The fitness test means performs a fitness test between the aggregated data and the probability distribution.

Patent Document 4 discloses a technique for detecting an abnormality in the operating state using chi-square determination.

Patent Document 5 discloses a technique for generating a frequency distribution.

JP 2006-31178 A JP 2010-244528 A Japanese Patent No. 4349216 JP 2011-034208 A JP 2010-257227 A

∙ You cannot trust the results of a load test that does not accurately reproduce the load that is applied during production operations. Therefore, the load test execution device is required to accurately reproduce the load applied during the actual operation of the test target system.

Here, in order for the load test execution device to accurately reproduce the load applied during the actual operation of the test target system, it is necessary to satisfy at least the following two requirements and transmit the request. The first requirement is that the average number of requests transmitted per unit time satisfies a desired value. The second requirement is that requests are transmitted at time intervals according to a predetermined distribution (it can be said that the number of requests transmitted per unit time follows a predetermined distribution).

The load test execution device of Patent Document 1 has evaluated whether a request transmitted by the load test execution device satisfies the first requirement. However, the load test execution device has not evaluated whether the second requirement is satisfied.

Whether the request transmitted by the load test execution device satisfies the second requirement is important for the reliability of the load test result. This is because the response time differs between the case where the request arrival time interval is constant and the case where the request arrival time is not constant, even if the request processing performance of the system under test and the average number of request arrivals per unit time are equal. . Here, the response time is a time required from when the request arrives at the test target system until the load test execution device receives the response.

Generally, when the request arrival time interval is not constant, the response time tends to be longer than when the request arrival time interval is constant. This is because when a request is transmitted at a non-constant time interval and a request arrives at a short interval, a queue is generated due to a large number of requests arriving at the test target system.

In order not to underestimate the response time to the request, the load test execution device needs to transmit the request not at a constant time interval but at a time interval according to a predetermined distribution. However, until now, no device has been known for evaluating whether or not a request transmitted by the load test execution device satisfies a predetermined distribution.

The present invention provides an information processing apparatus, a load evaluation method, and a program for executing a load evaluation capable of evaluating whether a load test execution apparatus can generate a desired load for a test target system. With the goal.

A first invention for solving the above-described problem is based on time-series data storage means for storing time-series data of a request transmission time transmitted from a load generation device to a test target system, based on the time-series data. The request rate, which is the number of requests transmitted per unit time, is calculated, the request rate is aggregated based on the class data, which is the aggregate interval for creating the frequency distribution data, and the frequency distribution data of the request rate is obtained. Observation frequency distribution creation means to create, expected frequency distribution acquisition means to obtain expected frequency distribution data which is expected frequency distribution data assuming that the distribution of the request rate follows a desired probability distribution, and the observed frequency Using the distribution data and the expected frequency distribution data, the degree-of-fit test of the observed frequency distribution data with the desired probability distribution And goodness-of-fit test means for performing an information processing apparatus including a.

A second invention for solving the above-mentioned problem is based on time-series data storage means for storing time-series data of the transmission time of a request transmitted from the load generation device to the test target system, and based on the time-series data. , An observation frequency distribution creating means for calculating a request transmission time interval, totaling based on class data that is a total interval width for creating frequency distribution data, and creating frequency distribution data of a request transmission time interval; Expected frequency distribution acquisition means for acquiring expected frequency distribution data that is expected frequency distribution data assuming that the distribution of the transmission time intervals of the request follows a desired probability distribution, the observed frequency distribution data, and the expected frequency A fitness test means for performing a fitness test of the observed frequency distribution data with the desired probability distribution using the distribution data; It is obtain the information processing apparatus.

A third invention for solving the above-described problem is a method of controlling an apparatus including time-series data storage means for storing time-series data of a request transmission time transmitted from a load generation apparatus to a test target system. Based on the time series data, the request rate that is the number of requests transmitted per unit time is calculated, and the request rate is totaled based on the class data that is the total interval width for creating the frequency distribution data. Generating frequency distribution data of request rate, acquiring expected frequency distribution data which is frequency distribution data expected when the distribution of request rate follows a desired probability distribution, and the observed frequency Using the distribution data and the expected frequency distribution data, the degree of fitness of the observed frequency distribution data with the desired probability distribution A control method for the device for controlling to execute a step of performing a constant, a.

According to a fourth aspect of the present invention for solving the above-described problem, a computer comprising time-series data storage means for storing time-series data of the transmission time of a request transmitted from the load generation device to the test target system, Based on, the request rate that is the number of requests sent per unit time is calculated, the request rate is aggregated based on the class data that is the aggregate interval width for creating the frequency distribution data, and the frequency of the request rate Observation frequency distribution creation processing for creating distribution data, expected frequency distribution acquisition processing for obtaining expected frequency distribution data that is expected frequency distribution data assuming that the distribution of the request rate follows a desired probability distribution, the observation Using the frequency distribution data and the expected frequency distribution data, the desired probability distribution for the observed frequency distribution data. Goodness-of-fit test process for goodness-of-fit test with a non-volatile recording medium recording a program for executing the.

It is possible to provide an information processing apparatus, a load evaluation method, and a program for executing a load evaluation that can evaluate whether the load test execution apparatus can generate a desired load on the test target system. Here, “whether the load test execution device can generate a desired load for the test target system” includes “whether the request transmitted by the load test execution device satisfies a predetermined distribution”.

FIG. 1 is a network configuration diagram including a load evaluation device 1 according to the first embodiment. FIG. 2 is a block diagram illustrating a configuration of the load evaluation device 1 according to the first embodiment. FIG. 3 is an example of time-series data stored in the time-series data storage unit 111. FIG. 4 is an example of data stored in the chi-square distribution data storage unit 112. FIG. 5 is an example of class data acquired by the class acquisition unit 122. FIG. 6 is an example of observation frequency distribution data regarding the request rate. FIG. 7 is an example of expected frequency distribution data. FIG. 8 is a flowchart for explaining the operation of the load evaluation device 1 according to the first embodiment. FIG. 9 is an example in which observed frequencies and expected frequencies are arranged. FIG. 10 is an example of observation frequency distribution data regarding the request rate. FIG. 11 is a block diagram illustrating a configuration of the load test execution device 2 according to the second embodiment. FIG. 12 is a diagram illustrating typical behavior of a virtual user in a closed model. FIG. 13 is a diagram illustrating typical behavior of a virtual user in an open model. FIG. 14 is a block diagram illustrating a configuration of the load evaluation device 1 according to the second embodiment. FIG. 15 is a flowchart for explaining the operation of the load evaluation device 1 according to the second embodiment. FIG. 16 is a block diagram illustrating a configuration of the load evaluation device 1 according to the third embodiment. FIG. 17 is an example of the probability density function of the exponential distribution. FIG. 18 is an example of a Poisson distribution probability mass function. FIG. 19 is a block diagram illustrating a configuration of the load evaluation device 1 according to the fourth embodiment. FIG. 20 is a diagram illustrating a hardware configuration of a computer 700 that implements the load evaluation device 1 according to the first embodiment.

<< First Embodiment >>
<Description of configuration>
FIG. 1 is a network configuration diagram including a load evaluation device (also referred to as an information processing device) 1 according to the first embodiment.

The load evaluation device 1 is connected to the load test execution device 2 and the test target system 3 via the network 4 so as to communicate with each other.

The load test execution device 2 (also referred to as a load generation device) is a device that performs a load test on the test target system 3. The load test execution device 2 establishes a TCP (Transmission Control Protocol) connection with the test target system 3 and applies a load by continuing to transmit requests to the test target system 3.

The test target system 3 is a server or system that is a target of the load test. The test target system 3 processes the request received from the load test execution device 2 and responds to the load test execution device 2.

When the test target system 3 can process all the received requests without any problem, the transmission time interval of the request transmitted by the load test execution device 2 is equal to the arrival time interval of the request when viewed from the test target system 3. When the test target system 3 can process all the arrived requests without any problem, the average number of requests per unit time (request transmission rate) of the request transmitted by the load test execution device 2 is viewed from the test target system 3. And the average number of requests received per unit time (request arrival rate).

Hereafter, it is called the request interval without distinguishing between the request transmission interval and the arrival interval. Also, the request transmission rate and the request arrival rate are called request rates. The values of the request transmission rate and the request arrival rate are equal to the throughput value indicating the number of request processes per unit time.

The load evaluation device 1 acquires a log of communication data between the load test execution device 2 and the test target system 3. And the load evaluation apparatus 1 calculates a request interval or a request rate using the acquired log. The load evaluation device 1 evaluates whether or not the calculated request interval or request rate distribution matches the probability distribution desired by the operator of the load test execution device 2.

In the present embodiment, it is assumed that the test target system 3 is a system that receives requests from an unspecified number of customers, such as an online shopping site. It is known that during the actual operation of such a system, the request interval is a random number according to a probability distribution such as an exponential distribution or a Pareto distribution.

An example of the probability density function of the exponential distribution is shown in FIG. In FIG. 17, the horizontal axis represents the request interval, and the vertical axis represents the probability. The exponential distribution is a probability distribution having a parameter λ, and the expected value (average value) is 1 / λ. When the request interval is 1 / λ, its reciprocal λ represents the number of requests per second (request rate).

For example, when the average request interval is 1/20 (= 0.05) seconds, the average request rate is 20 [requests / second]. When the request interval follows an exponential distribution, the request rate is known to follow a Poisson distribution.

Fig. 18 shows an example of the Poisson distribution probability mass function. The horizontal axis in the figure represents the request rate, and the vertical axis represents the probability. A probability mass function is a probability density function for a discrete probability distribution. Since the request rate takes only an integer, the probability distribution is expressed as a probability mass function.

When the request interval and the request rate are not constant (constant) and follow the probability distribution as shown in FIG. 17 or 18, the request arrival interval is short or long, or the number of arrival requests per unit time is small. Time will be mixed. When a large number of requests arrive at the test target system, a queue is generated and the response time is likely to increase.

Generally, even if the average request interval is the same, the response time is longer when the request interval varies as shown in FIGS. 17 and 18 than when the request interval is constant.

As described above, it is empirically known that the request interval and the request rate vary during the actual operation of the test target system 3. Therefore, in order to accurately reproduce the load applied during the actual operation of the test target system 3, it is reasonable to express the load fluctuation using the probability distribution or the like for the request interval transmitted by the load test execution device 2 as well. is there.

Conversely, if the request interval transmitted by the load test execution device 2 is not as intended by the operator, the value of the throughput and the response time obtained as a result of the load test are not reliable. In order to guarantee the reliability of the result of the load test, it is necessary to evaluate that the request interval follows the distribution intended by the operator of the load test execution device 2.

FIG. 2 is a block diagram showing the configuration of the load evaluation device 1 according to the first embodiment of the present invention. The load evaluation device 1 is a computer that includes a storage unit 110, a processing unit 120, a communication unit 130, an input unit 140, and an output unit 150.

The storage unit 110 includes a time series data storage unit 111 and a chi-square distribution data storage unit 112.

The time series data storage unit 111 stores time series data related to communication between the load test execution device 2 and the test target system 3. The time series data includes at least data on the transmission time of the request transmitted from the load test execution device 2 to the test target system 3. The time series data may include a TCP connection request, a TCP disconnection request, and the like from the load test execution device 2 to the test target system 3.

FIG. 3 is a diagram illustrating an example of data stored in the time-series data storage unit 111. The data shown in FIG. 3 represents log information related to communication between the load test execution device 2 and the test target system 3.

(The leftmost column in Fig. 3 represents the row number.) The second column from the left represents the elapsed time (seconds) since the start of log collection. S and D in “S-> D” in the third to fifth columns from the left represent the IP (Internet Protocol) address of the host. “S そ の-> D” means that data is transmitted from S to D. The sixth column represents the communication protocol. “TCP” in this example indicates that the TCP protocol is used. The tenth column represents the type of the TCP packet, and the eleventh and subsequent columns represent the contents transmitted by the HTTP (HyperText Transfer Protocol) protocol.

In the example of the time-series data storage unit 111 shown in FIG. 3, for example, the line number 10050 is changed from the machine identified by the IP address 192.168.0.1 to /some.com on the machine identified by the IP address 192.168.0.2. This indicates that a GET request is transmitted at a time 21.256673 seconds after the start of log collection for the content “php”. Thus, the time-series data storage unit 111 stores time-series data of request transmission times.

The chi-square distribution data storage unit 112 stores a chi-square distribution table indicating the value of the chi-square distribution or a calculation formula for calculating the value of the chi-square distribution. The value of the chi-square distribution is used when performing a chi-square test described later. FIG. 4 shows an example of a chi-square distribution table stored in the chi-square distribution data storage unit 112. Details of the chi-square distribution table will be described later.

The processing unit 120 includes a communication data acquisition unit 121, a class acquisition unit 122, an observation frequency distribution creation unit 123, an expected frequency distribution acquisition unit 124, a superiority level acquisition unit 125, and a fitness test unit 126. Composed.

The communication data acquisition unit 121 acquires data related to communication between the load test execution device 2 and the test target server 3. Next, the communication data acquisition unit 121 stores the acquired data related to communication in the time-series data storage unit 111. Here, the data related to communication specifically refers to request transmission, TCP connection request, and TCP disconnection request.

The class acquisition unit 122 receives “class data” necessary for creating the “frequency distribution” from the operator of the load evaluation device 1.

The frequency distribution represents the distribution of the number of sample data for a certain property or variable. Frequency distribution is also called frequency distribution. At this time, a class of certain properties for classifying sample data and a range of variables (total section width) are called classes.

Class data is data that defines a class. In the load evaluation device 1 according to the present embodiment, the range of the interval related to the request interval or request rate transmitted by the load test execution device 2 is defined as a class.

FIG. 5 is a diagram illustrating an example of class data acquired by the class acquisition unit 122. The class data shown in FIG. 5 is an example of class data related to the request rate, and the contents thereof are as follows. In this example, seven classes are set. The request rate “6 to” represents “6 or more”.

Class 1: 0 (request / s) or more and less than 1 (request / s) Class 2: 1 (request / s) or more and less than 2 (requests / s) Class 3: 2 (requests / s) or more 3 (requests / s) Less than Class 4: 3 (request / second) or more and less than 4 (request / second) Class 5: 4 (request / second) or more and less than 5 (request / second) Class 6: 5 (request / second) or more 6 (request / second) ) Less than class 7: 6 (requests / second) or more The observation frequency distribution creating unit 123 calculates the request rate or the request interval from the time series data stored in the time series data storage unit 111. Next, the observation frequency distribution creation unit 123 aggregates the calculated request rate or request interval based on the class data received by the class acquisition unit 122 to create “observation frequency distribution data”.

The observed frequency distribution data is data in which the frequency corresponding to the classification of the class is counted for each class of the class data, and the class and the frequency are associated with each other.

FIG. 6 is a diagram illustrating an example of observation frequency distribution data regarding the request rate created by the observation frequency distribution creation unit 123. The observation frequency distribution data shown in FIG. 6 has a total of 100 observation frequencies (sample data numbers). The observation frequency distribution data is aggregated according to the class data acquired by the class acquisition unit 122.

The expected frequency distribution acquisition unit 124 acquires or calculates “expected frequency distribution data” for testing whether the observed frequency distribution is in line with the distribution intended by the operator.

FIG. 7 is a diagram illustrating an example of expected frequency distribution data acquired by the expected frequency distribution acquisition unit 124. The expected frequency distribution is a frequency distribution expected when the observed value distribution is assumed to follow a probability distribution desired by the operator. Here, the observation value is a value analyzed from the time series data stored in the time series data acquisition unit 111.

If the observed frequency distribution is a frequency distribution related to the request rate, the expected frequency distribution is a distribution according to the Poisson distribution. The expected frequency distribution shown in FIG. 7 is a distribution calculated based on a Poisson distribution with λ (expected value) of 2 so that the total frequency of the expected frequency distribution is the same as the total frequency of the observed frequency distribution. It is.

The superiority level acquisition unit 125 acquires the value of the significance level in the chi-square test from the operator. The value of the significance level is a probability value such as 5% or 1%, for example. The acquired value of significance level is used in the later-described chi-square test.

The goodness-of-fit test unit 126 performs a goodness-of-fit test (chi-square test) between the observed frequency distribution data created by the observed frequency distribution creation unit 123 and the expected frequency distribution data acquired by the expected frequency distribution acquisition unit 124. In this way, the suitability test unit 126 evaluates whether the request rate or request transmission interval of the request transmitted by the load test execution device 2 is in accordance with the distribution desired by the operator of the load test execution device 2.

The communication unit 300 communicates with the load test execution device 2 and the test target system 3.

The input unit 400 receives input from the operator of the load evaluation device 1.

The output unit 500 outputs information to the operator of the load evaluation device 1.
<Hardware configuration>
The storage unit 110 of the load evaluation device 1 can be realized using a RAM (Random Access Memory), an HDD (Hard Disk Drive), or the like. Each unit included in the processing unit 120 can be realized by a CPU (Central Processing Unit) included in the load evaluation device 1 executing a predetermined program or code. The communication unit 130 can be realized by an application program controlling a network interface card (NIC) using a function provided by an OS (Operating System), for example. The input unit 140 can be realized using, for example, a keyboard or a mouse. The output unit 150 can be realized using a display, for example.

FIG. 20 is a diagram illustrating a hardware configuration of a computer 700 that implements the load evaluation device 1 according to the present embodiment.

As shown in FIG. 20, a computer 700 includes a CPU (Central Processing Unit) 701, a storage unit 702, a storage device 703, an input unit 704, an output unit 705, and a communication unit 706. Furthermore, the computer 700 includes a recording medium (or storage medium) 707 supplied from the outside. The recording medium 707 may be a non-volatile recording medium that stores information non-temporarily.

The CPU 701 controls the overall operation of the computer 700 by operating an operating system (not shown). The CPU 701 reads a program and data from a recording medium 707 mounted on the storage device 703, for example, and writes the read program and data to the storage unit 702. Here, the program is, for example, a program that causes the computer 700 to execute operations of flowcharts shown in FIGS.

Then, the CPU 701 executes various processes as the processing unit 120 shown in FIG. 2 according to the read program and based on the read data.

Note that the CPU 701 may download a program or data to the storage unit 702 from an external computer (not shown) connected to a communication network (not shown).

The storage unit 702 may correspond to the storage unit 120 in FIG.

The storage device 703 is, for example, an optical disk, a flexible disk, a magnetic optical disk, an external hard disk, and a semiconductor memory, and includes a recording medium 707. The storage device 703 (recording medium 707) stores the program in a computer-readable manner. The storage device 703 may store data. The storage device 703 may correspond to the storage unit 120 in FIG.

The input unit 704 may correspond to the input unit 140 in FIG.

The output unit 705 may correspond to the output unit 150 in FIG.

The communication unit 706 may correspond to the communication unit 130 of FIG.

As described above, the functional unit block of the load evaluation apparatus 1 shown in FIG. 2 may be realized by the computer 700 having the hardware configuration shown in FIG. However, the means for realizing each unit included in the computer 700 is not limited to the above. In other words, the computer 700 may be realized by one physically coupled device, or may be realized by two or more physically separated devices connected by wire or wirelessly and by a plurality of these devices. .

A recording medium 707 in which the above-described program code is recorded may be supplied to the computer 700, and the CPU 701 may read and execute the program code stored in the recording medium 707. Alternatively, the CPU 701 may store the code of the program stored in the recording medium 707 in the storage unit 702, the storage device 703, or both. That is, the present embodiment includes an embodiment of a recording medium 707 that stores a program (software) executed by the computer 700 (CPU 701) temporarily or non-temporarily.
<Description of operation>
Next, the operation of the load evaluation device 1 according to the first embodiment will be described. FIG. 8 is a flowchart for explaining the operation of the load evaluation device 1.

First, the class acquisition unit 122 acquires class data necessary for creating the frequency distribution (S11). The class acquisition unit 122 acquires class data related to the request rate or the request interval as exemplified in FIG. Hereinafter, it is assumed that the class acquisition unit 122 has received class data related to the request rate.

Next, the observation frequency distribution creation unit 123 acquires class data related to the request rate from the class acquisition unit 122. Subsequently, the observation frequency distribution creation unit 123 calculates the request rate based on the time series data stored in the time series data storage unit 111. Subsequently, the observation frequency distribution creation unit 123 aggregates the calculated request rates based on the class data acquired from the class acquisition unit 122, and creates observation frequency distribution data regarding the request rate (S12). The observation frequency distribution creation unit 123 creates, for example, observation frequency distribution data having a request rate as shown in FIG.

On the other hand, the expected frequency distribution acquisition unit 124 acquires or creates expected frequency distribution data that is a frequency distribution that is expected when the observed value distribution follows the distribution desired by the operator. For example, expected frequency distribution data as shown in FIG. 7 is acquired or created (S13).

Here, for the sake of clarity, the observed frequency distribution and the expected frequency distribution are shown again as shown in FIG.

Next, the superiority level acquisition unit 125 acquires a significant level value from the operator. The superiority level acquisition unit 125 acquires a value of a significance level of “5%”, for example (S14).

The goodness-of-fit test unit 126 performs a goodness-of-fit test between the observation frequency distribution data created by the observation frequency distribution creation unit 123 and the expected frequency distribution data acquired by the expected frequency distribution acquisition unit 124 (S15).

The goodness-of-fit test unit 126 uses a chi-square test well known in the field of statistics for the goodness-of-fit test. The goodness-of-fit test based on the chi-square test is a technique for testing the hypothesis that “the frequency distribution of observed events matches the expected frequency distribution”. For example, if the significance level is 5% and the chi-square value calculated based on the difference between the observed event and the expected event is included in the danger zone of less than 5%, the assumption is rejected and may not be met. Is considered high. Conversely, if the chi-square value is not within the risk range of less than 5%, the assumption that it is met is adopted and it is considered likely that it is met.

Hereinafter, the chi-square test will be specifically described. In the chi-square (χ ² ) test, a chi-square value is first calculated from the distribution of the observed frequency and the expected frequency. When the observation frequency of class i is Oi and the expected frequency is Ei, the chi-square value is calculated by the following equation.

χ ² = Σi ((Oi−Ei) 2 / Ei)
In this example, the chi-square value ^{= {(15-13.5)} 2 /13.5}Tasu{(27-27.1) 2 ^{/27.1}Tasu{(29-27.1)} 2/27. 1} + {(13-18.0) ² /18.0}+{(6−9.02) ² /9.02}+{(5-3.61) ² /3.61} + {( 5-1.66) ² /1.66}≈9.96. Next, the probability that the chi-square value is 9.96 is calculated based on the chi-square distribution table. FIG. 4 shows a part of the chi-square distribution table. The degree of freedom in the chi-square distribution table corresponds to a number obtained by subtracting 1 from the number of classes. That is, the degree of freedom in this example is 7-1 = 6. This means that since the number of sample data is determined to be 100, if the frequencies for the six classes are determined, the remaining one expected frequency is determined. The chi-square distribution table of FIG. 4 shows the following regarding the chi-square value with six degrees of freedom.

・ Probability that chi-square value is 0.872 or less or 16.812 or more ・ 1% or less ・ Probability that chi-square value is 1.237 or less or 14.449 or more ・ 2.5% or less ・ chi-square value is 1 In this example, it is determined that the probability of obtaining the calculated chi-square value 9.96 is not in the danger zone of less than 5%. That is, the above-mentioned hypothesis that “the frequency distribution of observed events matches the expected frequency distribution” is adopted at a significance level of 5%. That is, the observed data may be considered to match the distribution intended by the user.

In addition, it is known that if there is a class whose expected frequency is less than 10 or 5, the error of the calculated chi-square value becomes large. In the case of this example, when calculating the chi-square value, the ranks of the request rates “5” and “6˜” may be combined into “5˜” as shown in FIG.

Finally, the output unit 150 outputs a determination result indicating conformity to an external output device such as a display, a log file, or the like (S16). In the case of the example described above, the determination result means that the request rate in the load generated by the load generator matches the distribution intended by the user.
<< Modification of First Embodiment >>
In the above example, the distribution with respect to the request rate transmitted by the load test execution device 2 is evaluated. However, the request interval transmitted by the load test execution device 2 can also be evaluated. When evaluating the request interval, it is assumed that the class data acquired by the class acquisition unit 122 is class data related to the request interval, and the expected degree distribution acquired or calculated by the expected frequency distribution acquisition unit 124 follows the exponential distribution. Good.

An example of class data related to the request interval is as follows.

Class 1: 0.0 (second) to less than 0.1 (second) Class 2: 0.1 (second) to less than 0.2 (second) Class 3: 0.2 (second) to 0.3 (second) Less than Class 4: 0.3 (second) or more and less than 0.4 (second) Class 5: 0.4 (second) or more and less than 0.5 (second) Class 6: 0.5 (second) or more and 0.6 (second) ) Less than Class 7: 0.6 (seconds) or more The observation frequency distribution creation unit 123 may create observation frequency distribution data related to the request interval when the class data received by the class acquisition unit 122 is data related to the request interval. . In addition, the observation frequency distribution creation unit 123 may create observation frequency distribution data related to the request rate when the class data received by the class acquisition unit 122 is data related to the request rate.

The load evaluation device 1 and the load test execution device 2 may be an integrated device. The class acquisition unit 122 may acquire class data from a storage unit (not shown) instead of acquiring class data from an operator. The expected frequency distribution acquisition unit 124 may acquire the expected frequency distribution data from an operator (not shown) instead of acquiring the expected frequency distribution data from the operator. The significance level acquisition unit 125 may acquire the expected frequency distribution data from a storage unit or the like (not shown) instead of acquiring the expected frequency distribution data from the operator.

As described above, according to the load evaluation device 1 according to the present embodiment, it is verified whether or not the transmission time interval or request transmission rate of the request transmitted by the load test execution device 2 follows a predetermined distribution. Can do. By referring to such information, it can be confirmed whether or not the load test has been accurately performed, and it can be determined whether or not the result of the load test is reliable.

This is because the following configuration is included. First, the time-series data storage unit stores the time-series data of the transmission time of the request transmitted from the load test execution device 2 to the test target system 3. Second, the observed frequency distribution creation unit 123 calculates a request rate that is the number of requests transmitted per unit time based on the time series data, and a class that is a total interval width for creating the frequency distribution data Based on the data, the request rate is aggregated, and observation frequency distribution data of the request rate is created. Third, the expected frequency distribution acquisition 124 acquires expected frequency distribution data that is expected frequency distribution data when the request rate distribution is assumed to follow a desired probability distribution. Fourth, the fitness test 126 uses the observed frequency distribution data and the expected frequency distribution data to perform a fitness test with the desired probability distribution for the observed frequency distribution data.
<< Second Embodiment >>
The load evaluation apparatus 1 according to the second embodiment uses a load test of a type executed by a test scenario as an evaluation target.

First, the concept of “test scenario” in the load test will be described. The “test scenario” is a script for transmitting a request to the test target system 3 and applying a load. The test scenario is executed by a “virtual user”. The “virtual user” is a function for transmitting a request to the test target system 3 and is generated by the load test execution device 2. Here, a configuration of the load test execution device 2 that is an evaluation target of the load evaluation device 1 according to the second embodiment will be described.
<Configuration of load test execution device 2>
FIG. 11 is a block diagram illustrating a configuration of the load test execution device 2 according to the second embodiment. The load test execution device 2 includes a test storage unit 210 and a generation unit 220.

The test storage unit 210 stores a test scenario.

The test scenario is a script for sending a request to the test target system 3 and applying a load. The test scenario includes a plurality of processes from sending a request to receiving a response to the request. The specific content of the test scenario is determined by the operator of the load test execution device 2 in consideration of typical processing content of the test target system 3.

For example, when the test target system 3 is a Web server, the request data is a URL (Uniform Resource Locator) of the Web server. For example, when the test target system 3 is an online shopping site, a series of requests in the test scenario are a series of requests such as product browsing and selection, cart confirmation, product delivery destination and credit card number input by customers of the online shopping site. It corresponds to the operation of.

The generation unit 220 generates a “virtual user” and causes the virtual user to execute a test scenario.

The virtual user is a function for transmitting a request to the test target system 3 in order to simulate access to the test target system 3 by the user, for example. Each virtual user applies a load by transmitting a request to the test target system 3.

The generation unit 220 generates OS level threads and processes for the number of virtual users so that the load test execution apparatus 2 operates a plurality of virtual users in parallel, and executes a test scenario in each thread or process. <Typical behavior of virtual users>
Next, typical behavior of a virtual user will be described. Typical behaviors of virtual users include “closed model” and “open model”. Whether the “closed model” or the “open model” is selected as the behavior of the virtual user is an item determined by the operator of the load test execution device 2 according to the nature of the load test to be executed.
<Description of closed model>
FIG. 12 is a diagram illustrating typical behavior of a virtual user in a closed model.

The virtual user first establishes a TCP connection in accordance with a TCP / IP (Transmission Control Protocol / Internet Protocol) protocol before sending a request to the test target system 3. Thus, a logical communication path for exchanging requests and responses is prepared between the load test execution device 1 on which the virtual user operates and the test target system 3.

After the TCP connection is established, the virtual user sends a request specified in the test scenario. First, the virtual user transmits a first request and receives a response to the first request. The virtual user pauses for a certain time (thinking time) from receiving a response to a certain request until transmitting the next request. When the thinking time elapses after receiving the response to the first request, the virtual user transmits the second request and receives the response to the second request.

The virtual user repeats these operations and disconnects the TCP connection when reception of the Nth request is completed for the predetermined number of requests (N) specified in the test scenario. The virtual user completes the processing for one user after the TCP connection is established until the TCP connection is disconnected. The virtual user pauses for a predetermined time (thinking time) after receiving the Nth request and then executes the test scenario again.

The virtual user repeats the test scenario execution a predetermined number of times or for a predetermined time. This is because each virtual user cannot continue to apply a load to the test target system 3 for a certain period if the process is completed by executing the test scenario once.

仕 組 み The mechanism in which a certain number of virtual users repeat similar request processing in this way is called a closed model.

<Thinking time> is a time that simulates "the time required for a user to select a request". For example, consider a case where the test target system 3 is an online shopping site. The customer of the online shopping site receives a response to the request (for example, text describing the product, product image, product video, etc.). Thereafter, the customer checks the screen of the displayed product, thinks to select a product, inputs a product destination, a credit card number, and the like. The customer sends the next request after these times have elapsed since receiving the response to the request. In this way, the time interval from when the user receives a response to a request until the next request is transmitted is simulated using the thinking time.

The length of the thinking time is a parameter that can be set by the operator of the load test execution device 2. In many cases, the operator of the load test execution device 2 sets an average value of the thinking time as a parameter. And the load test execution apparatus 2 determines the length of the thinking time until it transmits a request using the random number according to an exponential distribution based on the parameter. The load test execution device 2 adjusts the time interval between requests to be transmitted in this way, and reproduces the variation in the request transmission interval.

Here, when the virtual user transmits a request and receives a response to the request (response time) as RT, and the thinking time as TT, the request transmission time interval in the test scenario is the response time RT And the sum of the thinking time TT. Here, the response time RT is an observed value that is influenced by the load state of the test target system, and is a value that cannot be directly controlled by the operator of the load test execution device 1.

When the load on the test target system 3 increases, the response time RT increases, and the response time RT becomes larger than the thinking time TT, the response time RT becomes more dominant in determining the request transmission time interval. In such a case, although the thought time TT is set as a random number and the change in the request transmission interval is attempted to be reproduced, the effect is diminished. As a result, the request transmission time interval deviates from the transmission time interval intended by the operator.

In the closed model, the load on the system under test increases and the response time increases, so the time to complete the test scenario also increases, so the time to repeat the next test scenario (virtual user arrival) is delayed. There's a problem. In other words, the closed model has a drawback that the load generated by the load generator decreases as the load on the test target system increases.
<Description of open model>
The difference between a closed model and an open model is that a virtual user repeats the execution of a test scenario in the closed model, whereas a virtual user is dynamically generated in the open model and disappears after the test scenario is executed. It is in.

FIG. 13 is a diagram for explaining a typical behavior of a virtual user in the opened model.

The virtual user first establishes a TCP connection according to the TCP / IP protocol before sending a request to the test target system 3. Thus, a logical communication path for exchanging requests and responses is prepared between the load test execution device 1 on which the virtual user operates and the test target system 3.

After the TCP connection is established, the virtual user sends a request specified in the test scenario. First, the virtual user transmits a first request and receives a response to the first request.

Next, the virtual user transmits a second request and receives a response to the second request. The virtual user repeats these operations, and disconnects the TCP connection when the reception of the Nth request is completed for the predetermined number of requests (N) defined in the test scenario. The virtual user completes the processing for one user after the TCP connection is established until the TCP connection is disconnected.

In the open model, the virtual user arrives at the frequency intended by the user regardless of the load of the test target system, so that a more accurate load can be generated as compared with the closed model.

As described above, the outline of the load test by the execution of the test scenario has been explained, and the concept of the test scenario has been explained.
<Description of configuration>
FIG. 14 is a block diagram showing a configuration of the load evaluation device 1 according to the second exemplary embodiment of the present invention. Constituent elements similar to those in the first embodiment are given the same reference numerals, and the description thereof is omitted. The load evaluation device 1 according to the second embodiment includes a filtering unit 127 in addition to the configuration of the load evaluation device 1 in the first embodiment shown in FIG.

The filtering unit 127 extracts data related to the first request in the test scenario from the time series data stored in the time series data storage unit 111.

The time-series data storage unit 111 stores time-series data of transmission times of requests transmitted to the test target system 3 by repeatedly executing a script by the virtual user. The time-series data storage unit 111 further stores time-series data of the transmission time of the TCP connection request requested to the test target system 3 when the virtual user starts executing the script.

The filtering unit 127 extracts, from the time-series data storage unit 111, the data of the transmission time of the request transmitted first after the TCP test is sent from the load test execution device 2 to the test target system 3. This is because, as described in the description of the closed model and the open model described above, the request to be transmitted first after the TCP connection request is the first request in the test scenario.

The operation of the filtering unit 127 will be described using a specific example of data stored in the time-series data storage unit 111 shown in FIG. FIG. 3 shows log information related to communication between the load test execution device 2 and the test target system 3.

Lines 10047 to 10049 in FIG. 3 indicate connection connection requests from the virtual user (port 1855 at address 192.168.0.1) generated by the load test execution device 2 to the test target system 3 (port 80 at address 192.168.0.2). Indicates that the connection was successfully established. An application on the OS communicates through a TCP port. In preparation for this, the applications on the OS exchange [SYN], [SYN, ACK], and [ACK] packets to establish mutual connections.

The line 10050 in FIG. 3 represents that a GET request is transmitted from the virtual user to the test target system 3 for the content “/some.php” on the server by the HTTP protocol. Since the request described in the row 10050 is a request transmitted first after the TCP connection request described in the rows 10047 to 10049, the filtering unit 127 transmits the transmission time data of the request described in the row 10050. To extract.

The filtering unit 127 analyzes the data stored in the time-series data storage unit 111 in this way, and extracts data related to the first request in the test scenario.
<Description of operation>
Next, the operation of the load evaluation device 1 according to the second embodiment will be described. FIG. 15 is a flowchart for explaining the operation of the load evaluation device 1.

First, the filtering unit 127 extracts data related to the first request in the test scenario from the time series data stored in the time series data storage unit 111. Subsequently, the filtering unit 127 outputs the extracted data to the observation frequency distribution creation unit 123 (S21).

Next, the class acquisition unit 122 acquires class data necessary for creating the frequency distribution (S22).

Next, the observation frequency distribution creation unit 123 analyzes the request rate regarding the first request in the test scenario based on the data output from the filtering unit 127. The observation frequency distribution creation unit 123 further creates observation frequency distribution data related to the request rate based on the class data acquired from the class acquisition unit 122 (S12).

Since the subsequent operation is the same as the operation of the load evaluation system according to the first embodiment, the description thereof is omitted.

The load evaluation device 1 according to the second embodiment evaluates the request interval or the request rate for only the first request of the test scenario. Here, when the request interval follows an exponential distribution, or when the request rate follows a Poisson distribution, the evaluation regarding only the first request of the test scenario is evaluated to approximate the evaluation regarding the entire request transmitted by the load test execution device 2. Can be done automatically. This is because the exponential distribution or Poisson distribution has a property called “memorylessness”.

As described above, in the load evaluation device 1 according to the second embodiment, the filtering unit 127 extracts only a part from the time series data stored in the time series data storage unit 111. And the observation frequency distribution creation part 123 totals only the data which the filtering part 127 extracted. Thus, in the load evaluation apparatus 1 according to the second embodiment, since the data to be tabulated by the observation frequency distribution creation unit 123 is reduced, the series of processing of the observation frequency distribution creation unit 123 can be reduced in cost. It becomes possible.

In addition, the load evaluation device 1 according to the second embodiment evaluates the request interval or the request rate for only the first request of the test scenario extracted by the filtering unit 127. By doing so, the load evaluation device 1 evaluates the arrival interval and arrival rate of the virtual users. In the model opened as described above, the arrival interval and arrival rate of the virtual user are controlled on the load test execution device 2 side regardless of the load of the load target system 3. Therefore, the load evaluation device 1 according to the second embodiment is not influenced by the load of the test target system 3 by evaluating the arrival interval and arrival rate of the virtual users, and the virtual user generated by the load test execution device 2 It is possible to accurately evaluate the quality (request interval, request rate) of the transmitted request.
<< Modification of Second Embodiment >>
By the way, what the filtering unit 127 extracts from the time-series data storage unit 111 does not necessarily need to be only data related to the first request of the test scenario. For example, the filtering unit 127 may extract a TCP connection request from the load test execution device 2 to the test target system 3.

A request to a Web server is normally transmitted according to a protocol such as HTTP or HTTPS (Hypertext Transfer Protocol Protocol over Secure Secure Socket Layer). HTTP and HTTPS are protocols that operate on top of the TCP protocol, and analyzing TCP communication packets is less time-consuming than analyzing HTTP and HTTPS packets.

Therefore, the conformity test for the TCP connection request can be performed at a lower cost than the conformity test for the first request in the test scenario.

The filtering unit 127 extracts the transmission time data of the TCP connection request from the time series data storage unit 111 and outputs the data to the observation frequency distribution creation unit 123.

The observation frequency distribution creation unit 123 calculates the number of TCP connection requests per unit time (TCP connection request rate) based on the data extracted by the filtering unit 127. Next, the observation frequency distribution creation unit 123 aggregates the calculated TCP connection request rates based on the class data obtained from the class acquisition unit 122, and creates observation frequency distribution data of the TCP connection request rates.

Subsequent operations are the same as the operations of the load evaluation system according to the first embodiment, and a description thereof will be omitted.
<< Third Embodiment >>
FIG. 16 is a block diagram showing the configuration of the load evaluation device 1 according to the third embodiment of the present invention. Constituent elements similar to those in the first embodiment are given the same reference numerals, and the description thereof is omitted.

The load evaluation device 1 according to the third embodiment includes a filtering unit 127A instead of the filtering unit 127 as compared with the configuration of the load evaluation device 1 in the second embodiment shown in FIG. Further, a time series data storage unit 111 </ b> A is provided instead of the time series data storage unit 111.

The load evaluation device 1 according to the third embodiment can save the storage capacity of the time series data storage unit 111A as compared with the load evaluation device 1 according to the first embodiment.

That is, according to the load evaluation device 1 according to the third embodiment, the filtering unit 127A extracts only a part of the communication-related data acquired by the communication data acquisition unit 121 and stores it in the time-series data storage unit 111A. . Specifically, the filtering unit 127A extracts the data on the transmission time of the first request of the test scenario or the data on the transmission time of the TCP connection request from the data related to the communication acquired by the communication data acquisition unit 121, and sets the time series data. Store in the storage unit 111.

The operation of the filtering unit 127A to extract the data related to the first request of the test scenario or the TCP connection request is the same as that in the second embodiment, and thus the description thereof is omitted. Further, the operation of the load evaluation device 1 according to the third embodiment for performing the suitability test of the request interval or request rate distribution is the same as that of the first embodiment, and thus the description thereof is omitted.

As described above, in the load evaluation device 1 according to the third embodiment, the time-series data storage unit 111 stores only some data extracted by the filtering unit 127. Thereby, in the load evaluation device 1 according to the third embodiment, the storage capacity of the time-series data storage unit 111 can be saved.
<< Fourth Embodiment >>
A load evaluation apparatus 1 according to a fourth embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 19 is a configuration diagram of the load evaluation device 1 according to the fourth embodiment.

The load evaluation device 1 according to the fourth embodiment includes a time-series data storage unit 111, an observation frequency distribution creation unit 123, an expected frequency distribution acquisition unit 124, and a fitness test unit 126. .

The time-series data storage unit 111 stores time-series data of request transmission times transmitted from the load generation device to the test target system.

The observation frequency distribution creation unit 123 calculates a request rate that is the number of requests transmitted per unit time based on the time series data, and based on the class data that is the total section width for creating the frequency distribution data. The request rates are totaled, and frequency distribution data of the request rate is created.

The expected frequency distribution acquisition unit 124 acquires expected frequency distribution data which is frequency distribution data expected when the request rate distribution is based on a desired probability distribution.

The goodness-of-fit test unit 126 uses the observed frequency distribution data and the expected frequency distribution data to perform a goodness-of-fit test on the observed frequency distribution data with the desired probability distribution.

By including the above configuration, the load evaluation device 1 of the present embodiment can evaluate whether the load test execution device 2 can generate a desired load for the test target system. Here, “whether the load test execution device can generate a desired load for the test target system” includes “whether the request transmitted by the load test execution device satisfies a predetermined distribution”.

In the present invention, the load test execution device 2 is an example of the “load generation device” described in the claims. Alternatively, each virtual user generated by the load test execution device 2 can be regarded as a “load generation device” described in the claims. Note that the load test execution device 2 may separately provide an agent in which only the request transmission / reception function is mounted separately from the load test execution device 2. In this case, the agent that transmits / receives a request to / from the test target system 3 corresponds to the “load generation device” described in the claims.

As mentioned above, although the form for implementing this invention was demonstrated, the said embodiment is for making an understanding of this invention easy, and is not for limiting and interpreting this invention. The present invention can be changed and improved without departing from the gist thereof, and the present invention includes equivalents thereof.

This application claims priority based on Japanese Patent Application No. 2012-079419 filed on March 30, 2012, the entire disclosure of which is incorporated herein.

The present invention is suitable as an evaluation apparatus or system for a load test apparatus for verifying performance and behavior when a predetermined load is applied to a server.

DESCRIPTION OF SYMBOLS 1 Load evaluation apparatus 2 Load test execution apparatus 3 Test object system 4 Network 110 Storage part 111 Time series data storage part 112 Chi-square data storage part 120 Processing part 121 Communication data acquisition part 122 Class acquisition part 123 Observation frequency distribution preparation part 124 Expectation Frequency distribution acquisition unit 125 Dominance level acquisition unit 126 Goodness-of-fit test unit 127 Filtering unit 130 Communication unit 140 Input unit 150 Output unit 160 Test storage unit 220 Generation unit

Claims (10)

1. Time-series data storage means for storing time-series data of the transmission time of the request transmitted from the load generation device to the test target system;
   Based on the time series data, a request rate that is the number of requests transmitted per unit time is calculated, and the request rate is totaled based on class data that is a total interval width for creating frequency distribution data, Observation frequency distribution creation means for creating frequency distribution data of request rate,
   Expected frequency distribution acquisition means for acquiring expected frequency distribution data which is frequency distribution data expected when the request rate distribution is assumed to follow a desired probability distribution;
   Using the observed frequency distribution data and the expected frequency distribution data, a fitness test means for performing a fitness test with the desired probability distribution for the observed frequency distribution data,
   An information processing apparatus comprising:
2. The time-series data stored by the time-series data storage means is obtained by repeatedly executing a script including a process in which the load generation device receives a response to the request after transmitting the request a plurality of times. Is the time series data of the transmission time of the request sent to
   Filtering means for extracting data of the transmission time of the first request in the script from the time series data storage means, and outputting the extracted data to the observation frequency distribution creating means,
   The observation frequency distribution creating unit calculates a request rate that is the number of the first requests transmitted per unit time based on the data extracted by the filtering unit, totals the request rate, and observes the frequency distribution Create data,
   The information processing apparatus according to claim 1, wherein the expected frequency distribution acquisition unit acquires expected frequency distribution data that is frequency distribution data expected when the request rate distribution is assumed to follow a desired probability distribution.
3. The time-series data storage means is configured so that the load generation device transmits a request transmitted to the test target system by repeatedly executing a script including a process of receiving a response to the request after transmitting the request. Transmission time time series data, and when the load generation device starts executing the script, the TCP connection request transmission time time series data requested to the test target system,
   Filtering means for extracting from the time series data storage means data at the time of transmission of a TCP connection request from the load generation device to the load generation device, and outputting the extracted data to the observation frequency distribution creation means ,
   The observation frequency distribution creating unit calculates a TCP connection request rate that is the number of TCP connection requests transmitted per unit time based on the data extracted by the filtering unit, and tabulates the TCP connection request rate. Create observation frequency distribution data,
   The information processing apparatus according to claim 1, wherein the expected frequency distribution acquisition unit acquires expected frequency distribution data that is frequency distribution data expected when the distribution of the TCP connection request rate follows a desired probability distribution.
4. Time-series data storage means for storing time-series data of the transmission time of the request transmitted from the load generation device to the test target system;
   Based on the time-series data, the request transmission time interval is calculated, and the request transmission time interval is aggregated based on the class data that is the aggregation interval for creating the frequency distribution data, and the frequency distribution data of the request transmission time interval is created. Observation frequency distribution creation means,
   Expected frequency distribution acquisition means for acquiring expected frequency distribution data that is expected frequency distribution data when the distribution of transmission time intervals of the request follows a desired probability distribution, the observed frequency distribution data, and the expected frequency distribution A fitness test means for performing a fitness test with the desired probability distribution for the observed frequency distribution data using data,
   An information processing apparatus comprising:
5. The time-series data stored in the time-series data storage means is stored in the test target system by repeatedly executing a script including a process of receiving a response to the request after the load generation device transmits the request. Time-series data of the transmission time of the request sent to
   Filtering means for extracting data of the transmission time of the first request in the script from the time series data storage means, and outputting the extracted data to the observation frequency distribution creating means,
   The observation frequency distribution creation means calculates the transmission time interval of the first request based on the data extracted by the filtering means, and aggregates the transmission time intervals of the first request to create observation frequency distribution data. ,
   5. The information according to claim 4, wherein the expected frequency distribution acquisition means acquires expected frequency distribution data that is expected frequency distribution data when it is assumed that a transmission time interval distribution of the first request follows a desired probability distribution. Processing equipment.
6. The time-series data storage unit is configured to store the request transmitted to the test target system by repeatedly executing a script including a process of receiving a response to the request after the load generation device transmits the request. Time-series data of transmission time, and time-series data of transmission time of a TCP connection request requested to the test target system when the load generation device starts execution of the script,
   Filtering means for extracting from the time series data storage means data at the time of transmission of a TCP connection request from the load generation device to the load generation device, and outputting the extracted data to the observation frequency distribution creation means ,
   The observation frequency distribution creation means calculates a TCP connection request time interval based on the data extracted by the filtering means, totals the TCP connection request time interval, creates observation frequency distribution data,
   The information processing unit according to claim 4, wherein the expected frequency distribution acquisition unit acquires expected frequency distribution data that is expected frequency distribution data when it is assumed that a time interval distribution of the TCP connection request follows a desired probability distribution. apparatus.
7. The information processing apparatus according to claim 1, further comprising: a class acquisition unit that acquires the class data.
8. The information processing apparatus according to any one of claims 1 to 7,
   An information processing system including the load generation device.
9. A method for controlling an apparatus including time-series data storage means for storing time-series data of a transmission time of a request transmitted from a load generation apparatus to a test target system,
   Based on the time series data, a request rate that is the number of requests transmitted per unit time is calculated, and the request rate is totaled based on class data that is a total interval width for creating frequency distribution data, Creating frequency distribution data for request rates;
   Obtaining expected frequency distribution data which is frequency distribution data expected when the distribution of the request rate is assumed to follow a desired probability distribution;
   Using the observed frequency distribution data and the expected frequency distribution data, performing a fitness test with the desired probability distribution for the observed frequency distribution data;
   A method for controlling the apparatus for performing control so as to execute.
10. A computer including time-series data storage means for storing time-series data of the transmission time of a request transmitted from the load generation device to the test target system,
    Based on the time series data, a request rate that is the number of requests transmitted per unit time is calculated, and the request rate is totaled based on class data that is a total interval width for creating frequency distribution data, Observation frequency distribution creation process to create frequency distribution data of request rate,
    Expected frequency distribution acquisition processing for acquiring expected frequency distribution data which is frequency distribution data expected when the distribution of the request rate follows a desired probability distribution,
    Using the observed frequency distribution data and the expected frequency distribution data, a fitness test process for performing a fitness test with the desired probability distribution for the observed frequency distribution data,
    A non-volatile recording medium on which a program for executing is recorded.

Images