JP2009026020A - Load test system, load test data creation method, and program therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a load test system and a load test data creation method, which create data for load tests properly incorporating fluctuation of a machine load due to usage by a user. <P>SOLUTION: A load test system includes: a client terminal 100 used for normal operation; a GW server 120 for analyzing request information from the client terminal 100 to send back a reply; and a retrieval server 130, wherein normal data usable as load test data is only stored in a storage device 123 on the GW server 120. The GW server 120 has a means 125 which performs matching between the request information and information for managing a user-field, creates load test data suitable for a load test, and distributes the load test data 142 to a load test terminal 140. In a load test terminal 140, a virtual client is created for each of users designated on the load test terminal 140 by using the load test data 142 distributed from the GW server 120; and to the GW server 120, a request is made which is identical to a request transmitted from the client terminal 100 during the normal operation. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention performs a server load test by simulating various access patterns and generating loads on various network server systems existing on the network. A load test system that can generate a test in a single test execution terminal and perform a load test on the target server by performing various load tests from the virtual client, and records transactions in actual operation And load test data creation method and program capable of automatically changing the content of the load test operation executed by the virtual client to a state close to the actual operation by using the access pattern simulated by the virtual client. About.

  In recent years, with the development of network technology, applications implemented as server / client systems on the network are increasing. Testing such an application is very difficult because, unlike traditional stand-alone applications, the entire system, including the network, must be examined to measure its performance and detect bottlenecks. It has become.

In a conventional load test system for a network server, a plurality of client terminals are actually prepared, a client application is executed, and the server is actually accessed to generate a load.
As a solution, for example, as described in JP-A-2002-7232 (see Patent Document 1), a plurality of client processes are generated on one terminal to generate access to the server. That is, the test target server and the server test apparatus are interconnected by a dedicated data communication path instead of the public data communication network, and the server test apparatus recognizes a large number of HTTP responses disguised as many user agents individually, It is determined whether or not the object specified in each HTTP request is included in the response without error, and the determination result is output in a predetermined report format.
Moreover, the load test data was obtained by obtaining data at the time of actual operation and using it as test data to perform a test assuming a state close to actual operation.

  Also, for example, as described in JP-A-2004-21523 (see Patent Document 2), a plurality of designated virtual clients are generated in one test execution terminal, and a plurality of test execution terminals are managed. Some manager terminals are prepared, and various load test operations are performed on a plurality of virtual clients generated from the test manager terminal. That is, the setting means includes means for distributing the set / instructed load test operation information, the test client module and the test data to each test execution terminal, and the client module distributed in accordance with the distributed load test operation information The load access can be generated for the test target server that is operated and instructed using the distributed test data.

JP 2002-7232 A JP 2004-21523 A

In the conventional technology, when creating load test data, it is possible to test in a state closer to the actual operation by using the actual operation data, but the actual operation data is extracted at random. Therefore, there is a problem that the bias of the difference in the usage method by the user cannot be considered.
Specifically, for each user who uses the system, there are users who perform a search with a small amount of machine load and users who perform a large number of searches with a small load on the machine, and consider how to use various users. It is important to conduct a load test in In particular, in a system that provides various functions, a change in machine load due to a user's usage is very large and cannot be ignored. In other words, “Even if the same user function is used, the content of the request varies greatly from user to user.”, “The flow of a series of operations (combination of functions to be used) differs from user to user when performing user duties” Therefore, it is desirable to create test data.

(the purpose)
An object of the present invention is to provide a load test system, a load test data creation method, and a program therefor, which can create load test data that appropriately incorporates machine load fluctuations according to the usage method of the user.

  The load test system according to the present invention includes a client terminal used during normal operation, a gateway (GW) server that analyzes request information from the client and returns a response, a search server, and normal data that can be used as load test data. Only save to the storage device on the GW server. The GW server includes means for matching the request information with information managing users and fields, creating load test data suitable for a load test, and distributing the load test data to a load test terminal. It is characterized by.

  In addition, the load test terminal generates virtual clients for the number of users specified on the load test terminal using the load test data distributed from the GW server, and performs normal operation on the GW server. It is characterized by comprising a normal business means by making a request similar to the request transmitted from the client terminal at the time.

  The load test data creation method of the present invention is the same as that described in Patent Document 1 in that the test data is extracted from the actual operation data, but the usage method bias by the user is similar to the usage method. Creates load test data that appropriately incorporates machine load fluctuations according to the user's usage by grouping users, extracting actual operation data in consideration of the bias of each group, and using it as test data Is the method.

  According to the present invention, it is possible to automatically create appropriate load test data by effectively using actual operation data in a system in which the machine load greatly fluctuates due to an operation for each user. .

Hereinafter, embodiments of the means for collecting requests and the test data creating means of the present invention will be described in detail with reference to the drawings. Here, a high load test is described as an example.
FIG. 1 is a block diagram of a load test system according to an embodiment of the present invention.
In the GW server 120 having a request collection unit and a load test data creation unit, a data collection unit 121 for collecting requests from the client application 111, a request information management DB unit 123 for storing the requests, A user / category management DB unit 124 for managing the user IDs included and the classification for grouping the user IDs, and a test data creation unit 125 for creating load test data are provided.

The test data creation (generation) unit 125 has the following functions.
Generation target operation start log determination function and load test data generation function. First, the user operation table is generated by the user operation table creation function in the log determination function, and the ratio is acquired by the ratio acquisition function for each field with respect to the total number of users. The extraction user operation is determined by the extraction user operation determination function. Next, the user operation table / request information management DB unit matching function of the load test data generation function is used to match the two, and finally, test data is generated using the test data generation function.

  The GW server 120 first edits the request from the client application 111 by the data collection unit 121 and transmits the request to the back-end server 130. Next, it is determined whether or not the response from the back-end server 130 is in a normal state. In this manner, by collecting requests from the client 140, data used in actual operation can be used when performing a load test.

  Next, the test data creation unit 125 collates the user ID possessed by the request information management DB unit 123 with the user ID possessed by the user / field management DB unit 124, so that the client for each field Aggregate the number of requests, extract the number of request information according to the ratio for each field from the request information management DB unit 123, and create the data for load test. To create data for

Hereinafter, details of each component of the GW server 120 will be described.
First, the data collection unit 121 edits a request from the client application 111 and transmits the request to the backend server 130. Next, it is determined whether or not the response from the backend server 130 is in a normal state. If it is determined that the response is normal, the request from the client is registered in the request information management DB unit 123. Here, the request requested by the user includes data including an error such as an invalid search expression, a search expression in which the number of hits exceeding the upper limit of the system occurs, and a reference request for nonexistent data. The return code included in the response includes a distinction between a normally processed request and an illegally processed request.

Therefore, in determining whether the response from the back-end server 130 is in a normal state, the determination based on the return code from the back-end server 130 is performed. Thereby, unnecessary data due to incomplete user input data can be excluded. The determination procedure of the normal state and the contents registered in the request information management DB unit 123 will be described in detail with reference to FIG.
The test data creation unit 125 receives, as parameters, the time range and the number of extracted users as data creation parameters used in the load test, and extracts the user operation content that matches the conditions from the request information management DB unit 123 for the load test. Create data for. This extraction method and data creation contents will be described in detail with reference to FIG.

FIG. 2 is a flowchart showing a first processing operation example according to the present invention of the GW server in FIG. FIG. 3 is a data relation diagram showing details of the processing contents in FIG.
The purpose of the data collection unit 121 in FIG. 1 is to collect request information of clients in a normal state at the back-end server 130. As a process for that, the following steps are executed. First, data is read from the request sent from the client (step 202). Next, the request data is edited and transferred to the back-end server 130 (step 203). The edited content is edited from the client request 32 in FIG. 3 into a request 37 to the back-end server 130. Next, a response is received from the backend server 130 (step 204). Next, a return from the backend server 130 is determined (step 205).

If the return code of the response 38 from the back-end server 130 in FIG. 3 is 0, the request data is registered in the request information management DB unit 123 in FIG. 1 (step 206). This registered content includes the time when the return code of the response 38 is determined as the time stamp of the request information management DB 35, the content of the <User Name> tag of the client request 32 as the user ID of the request information management DB 35, and the client request 32. <Ns0. ...> The contents of the tag are stored in the service of the request information management DB 35, and the contents of the client request 32 are stored in the request data of the request information management DB 35, respectively.
Next, the response from the backend server 130 is edited and the response is made to the client (step 207). This editing content is to edit the back-end server response 38 of FIG.

FIG. 4 is a flowchart showing a second processing operation example of the GW server in FIG. 1 according to the present invention.
In the following, the test data creation program will be described in detail, but the concept of user operation will be supplemented. As described above, when creating test data, it is important to consider the point that “the request contents differ greatly from user to user” and “the flow of a series of operations differ from user to user”.
The point that “the content of the request varies greatly from user to user” will be described in detail below, but this can be satisfied by extracting test data in consideration of the bias for each field.
On the other hand, regarding the point that “the flow of a series of operations differs for each user”, it is important in what order and what functions the user has used. As a series of user operations, it begins with login, and after using search and other functions, it logs off. For this reason, the difference in the flow of a series of operations for each user can be satisfied by following the history of operation of a specific user and extracting it as test data.

Here, the request information management DB 123 of FIG. 1 stores information related to all user operations. Therefore, the following two processes are performed as means for specifying a series of user operations. First, the user operation information is deduplicated for each user, and the first record is determined in the flow of a series of user operations by adopting the first record at the time of deduplication.
Secondly, subsequent operation information regarding the corresponding user is acquired for each head record of the determined user operation.

The purpose of the load test data creation program in FIG. 1 is to specify the actual operation time zone used as load test data and the number of user operations to be extracted. Is to create data.
In FIG. 4, first, a creation target operation first log determination process is performed (step 402). The details of this processing will be described in detail with reference to FIGS. Next, load test data generation processing is performed (step 403). The details of this processing will be described in detail with reference to FIGS.

(Create target operation start log determination process)
FIG. 5 and FIG. 6 are a flowchart and data related diagrams showing an example of processing operation related to the creation target operation first log determination processing (step 402) in FIG.
First, a user ID and a time stamp are extracted (step 502). The acquisition target is the time stamp and user ID of the request information management DB 601 in FIG. Next, the data extracted in step 502 is deduplicated with the user ID, and the user operation table 604 in FIG. 6 is created (step 503). When deduplication is performed with the user ID, the first record in the duplicate data is adopted. Next, the user ID in the user ID / field management DB 605 in FIG. 6 is searched with the user ID in the user operation table 604 in FIG. 6, the field is acquired, and the field in the user operation table 604 in FIG. 6 is updated (step 504). ).

Next, the ratio for each field of the user operation table 604 is acquired using the user operation table 604 as a base (step 505). Next, the number of extracted user operations for each field is determined from the ratio for each field acquired in step 505 using the number of extracted users specified as a parameter at the time of the load test data creation process as a base (step 506).
Next, user operations corresponding to the number of user operations determined in step 506 are extracted from the user operation table 604 in FIG. 6 (step 507). At this time, the user operations are extracted from the same field on the average.

(Load test data generation processing)
7 and 8 are a flowchart and a data relation diagram showing an example of processing operation related to the load test data generation process (step 403) in FIG.
First, the process described below is repeated for the records in the user operation table 801 in FIG. 8 (step 701). Then, the user operation table 801 in FIG. 8 is read (step 702). The time stamp and user ID in the user information management DB 802 in FIG. 8 are searched using the time stamp and user ID in the user operation table 801 in FIG. 8 (step 703). Next, load test data is created (step 704). As the load test data, information acquired from the request data of the user information management DB 802 in FIG. 8 is output. Next, the next record of the user information management DB 802 of FIG. 8 is read (step 705).

  Next, it is determined whether or not the time stamp of the read record is within the time range specified as the parameter at the time of the load test data creation process. If so, the process proceeds to the next step (step 706). Next, it is compared whether or not the user ID of the read record matches the user ID read from the user operation table. If they are the same, the process proceeds to the post-reading step, and if they are different, the process returns to step 705 (step 707). Next, load test data is created (step 708). As the load test data, information acquired from the request data of the user information management DB 802 in FIG. 8 is output. Next, it is determined whether or not the service acquired from the request information management DB 802 in FIG. 8 is “Logoff”. If they match, the process returns to the beginning of the loop in step 701. If they do not match, the process returns to step 705 (step 709).

As described above with reference to FIGS. 1 to 8, in the GW server of the present invention, responses from clients to be collected are collected during actual operation. Also, by designating the time range and the number of user operations from the collected data during actual operation, the same load test data as in actual operation is created. This makes it possible to use data used in the most recent actual operation when performing the load test, and to perform a load test assuming actual operation even if the user usage status changes.
In addition, this time, the test data creation of the load test was explained as the subject, but by selecting the time zone to extract the time zone where the load is low, the number of extraction users is reduced, etc. It can also be used for tests at low loads.

  2, FIG. 4, FIG. 5 and FIG. 7 converted into the program are stored in a recording medium such as a CD-ROM, and this recording medium is stored in the computer of the load test system of the present invention. The present invention can be easily realized if it is installed, installed, and executed. If the program is downloaded from the computer to another computer via a network, the program can be generalized.

1 is an overall block diagram of a load test system according to an embodiment of the present invention. It is a flowchart which shows the 1st processing operation example which concerns on this invention of the GW server in FIG. It is a data related figure which shows the processing content in FIG. It is a flowchart which shows the 2nd processing operation example which concerns on this invention of the GW server in FIG. It is a flowchart which shows the 1st processing operation example which concerns on the test data preparation part in FIG. FIG. 6 is a data relation diagram showing processing contents in FIG. 5. It is a flowchart which shows the 2nd processing operation example which concerns on the test data preparation part in FIG. FIG. 8 is a data relation diagram showing processing contents in FIG. 7.

Explanation of symbols

31 Client 32 Client Request 33 Client Response 34 Data Editing Section 35 Request Information Management DB
36 Request data 37 Request to backend server 38 Response from backend server 39 Backend server 100 Client 111 Client application 120 GW server 121 Data collection unit 123 Request information management DB unit 124 User / field management DB unit 125 Test data creation Unit 130 Back-end server 131 Search unit 132 Search DB unit 140 Client 141 Client load test simulator 142 Test data 601 Request information management DB
602 Specific example of request data 603 Generation target operation first log determination processing 604 User operation table 605 User ID / field management DB
606 User operation table after first log determination processing 801 User operation table 802 User information management DB
803 Information acquired from request data in the user information management DB 804 Load test data generation 805 Data generated by the load test data generation unit

Claims (8)

In a client / server system,
A client terminal used during normal operation;
A back-end server that includes a search database and that searches and collates the request data stored in the search database with respect to the received request to determine whether the request is a normal request;
Receives a request from the client terminal, edits the request, forwards the edited request to the backend server, receives a response from the backend server, edits the response, and sends it to the client terminal A data collection unit that responds to the response and collects a request from the client terminal, and determines a normal state from a return code of a response from the backend server. And a gateway server provided with a test data creation means for creating test data from the registered request data. The load test system according to claim 1,
The test data creation means creates a load test request information from the collected request information when the response from the back-end server is determined to be normal, and uses a predefined client terminal and classification combination A load test system that analyzes trends for each category and creates load test data for performing load tests with data close to normal operation. In the data collection / edit processing method for creating load test data,
The gateway server reads the data from the request sent from the client terminal, edits the request data, transfers it to the backend server, receives a response from the backend server, determines the return from the backend server, To register the request data, the time when the return code of the response is determined is the time stamp of the request information management DB, the content of the user name of the client request is the user ID of the request information management DB, and the client request The contents of the ns tag are stored in the service of the request information management DB, the contents of the client request are stored in the request data of the request information management DB, the response from the backend server is edited, and the response to the client terminal Features to do Load test data generation method for. In the generation target operation first log determination processing method for creating load test data,
The gateway server extracts the user ID and time stamp from the request information management DB, deduplicates the extracted data with the user ID, creates a user operation table, and uses the user ID / field of the user operation table with the user ID Search the user ID of the management DB, acquire the field, update the field of the user operation table, acquire the ratio for each field of the user operation table using the user operation table as a base, and create the load test data Based on the number of extracted users specified as a parameter at the time of processing, the number of extracted user operations for each field is determined from the ratio for each acquired field, and user operations corresponding to the number of user operations previously determined from the user operation table are performed. A load test data creation method characterized by extracting. In the load test data creation method,
The gateway server reads the user operation table, searches the time stamp and user ID of the user information management DB with the time stamp and user ID of the user operation table, creates load test data, and reads the time of the read record It is determined whether or not the stamp is within the time range specified as the parameter at the time of load test data creation processing. If it is out of the range, it returns to the original state. Compare whether or not the ID matches the user ID read from the user operation table. If the ID is the same, proceed to the next step. If the ID is different, return to the original state and then create load test data. Load test data generation method. A data collection / editing program for creating load test data,
A procedure for reading data from a request sent from a client terminal to a computer in a gateway server, a procedure for editing request data, a procedure for transferring to a back-end server, a procedure for receiving a response from the back-end server, from the back-end server A program for data / editing processing for executing a procedure for determining a return code, a procedure for registering request data, a procedure for editing a response from the back-end server, and a procedure for responding to a client terminal. A generation target operation log determination program for creating load test data,
A procedure for extracting a user ID and a time stamp to a computer in the gateway server, a procedure for performing deduplication with the user ID for the extracted data, a procedure for creating a user operation table, and a user ID / user ID with a user ID in the user operation table A procedure for searching for a user ID in the field management DB, a procedure for acquiring a field and updating the field of the user operation table, a procedure for acquiring a ratio for each field of the user operation table using the user operation table as a base, Procedure for determining the number of extracted user operations for each field from the ratio for each acquired field, based on the number of extracted users specified as a parameter at the time of load test data creation processing, the number of user operations previously determined from the user operation table Generation target operation start log decision to execute the procedure for extracting user operations for each minute Processing program. A program for creating load test data,
A procedure for reading the user operation table into the computer in the gateway server, a procedure for searching the time stamp and user ID in the user information management DB with the time stamp and user ID of the user operation table, a procedure for creating load test data, Procedure to determine whether the time stamp of the read record is within the time range specified as a parameter during the load test data creation process, if it is out of range, return to the top of the loop, if within the range, Next procedure, procedure for comparing whether or not the user ID of the read record matches the user ID read from the user operation table, if it is the same, proceed to the post-reading step, if different, return to the original The procedure for returning and the load test data are the procedure for outputting the information acquired from the request data in the user information management DB. A load test for executing a procedure for determining whether or not the service acquired from the information management DB is “Logoff”, if it matches, returning to the top of the loop; if not, returning to the original procedure Data creation program.

Images