JP4460512B2 - Computer system and computer control method - Google Patents

Description

  The present invention relates to a computer system that improves response performance to a request from a client.

  Recent information systems are mainly web-based business systems (hereinafter referred to as “Web systems”) composed of a plurality of system device groups including network devices, a Web server group, an AP server group, and a DB server group. ing.

  Since a Web system is accessed from an unspecified number of clients via the Internet, a large amount of access may be concentrated. When the access to the Web system is concentrated, if the response to the user is delayed, the user will be away from the Web system, and the company that operates the Web system may lose a business opportunity.

Conventionally, in order to cope with this problem, for example, as a method for adding a new server group to the Web system and reducing the response time to the user, an extension such as adding a new server group to the Web system The load is distributed by.
It is generally known that the above load distribution method by extension is performed for each business application. That is, load distribution is performed by adding a server group having a business program group constituting a business application as a unit.

This load balancing method by adding business application units does not require changing parameters related to dependencies (configuration, flow rate control, etc.) between servers to be added and comparisons without stopping the Web system. Can be done easily.
However, since a server group in which a business program group with a low load is deployed is also an expansion target, a problem arises that a large amount of system resources (hardware) is required.

Furthermore, when the load is concentrated on specific Web content in the business program and affects other Web content processing in the same business program, the load balancing method by adding servers is not In some cases, the response time of the client to the Web content is not improved so much.
In addition, when a server group to be added is used between a plurality of Web systems, the capacity of a business program may be large, and at this time, the deployment time to the server also becomes long.

For this reason, the deployment time may become a bottleneck in an emergency that the operations manager did not expect, such as when a high load on the Web system suddenly occurs.
In addition, as an example of load distribution technology by extension, Patent Document 1 discloses a technique for calculating and determining the number of servers to be added based on performance information for a server group constituting a Web system. In this patent document 1, since the expansion is performed in units of servers, the problem of mass use of system resources can be alleviated as compared with load distribution by expansion of business applications.
Japanese Patent Laying-Open No. 2003-178040 (paragraph 009, FIG. 2)

  However, in Patent Document 1, when a load is applied to a specific Web content in a business program, there is a possibility that the response time to other Web content is not improved much even if load distribution by extension is executed. .

  Therefore, the present invention has been made in view of the above problems, and an object thereof is to efficiently distribute the load on a computer system by extracting high-load Web content and performing load distribution.

In order to solve the above problems, a computer system according to the present invention manages a load distribution device that responds to a request from a client, a plurality of computers having Web contents or application programs, and the load distribution device and the plurality of computers. A plurality of computers having either the Web content or the application program having a request transfer information table indicating correspondence between the URL pattern of the request and a transfer destination computer. The operation management computer includes a storage unit and a processing unit, the storage unit stores a target value of performance information related to the response time of the request for each URL pattern of the request, and the processing unit For each request URL pattern, The performance information related to the response time of the request is monitored, it is determined whether the monitored performance information satisfies the target value of the performance information read from the storage unit, and the request that does not satisfy the target value of the performance information the Web content corresponding to a URL pattern, by referring to the request transfer information table, the Web content, or computer with Web content, to identify whether it is deployed in Re computer or noise having an application program When it is not specified that the Web content is deployed in a computer having an application program , the Web server is load-balanced by moving the Web content to another computer.

  According to the present invention, it is possible to efficiently distribute the load on the computer system.

Hereinafter, load distribution of a business system according to an embodiment of the present invention will be described with reference to the drawings as appropriate.
First, the overall configuration of the present embodiment will be described with reference to FIG. FIG. 1 is an overall configuration diagram of the present embodiment.

  The Web system 141 as a business system is connected to the operation management terminal 101 and the operation management system (operation management computer) 111 via the operation management network 131, and is connected to the client 171 via the customer access network 161 and the Internet 181. .

The operation management terminal 101 is a device for an operation administrator to set and operate the operation management system 111, and includes the display unit 102 and the input unit 103, and can be realized by, for example, a personal computer.
The display unit 102 performs display, and is, for example, a liquid crystal display. The input unit 103 is input by an operation manager, and is, for example, a keyboard or a mouse.
The operation management terminal 101 is connected to the operation management system 111 via the operation management network 131.
The operation management terminal 101 may be configured as an apparatus integrated with the operation management system 111.

  The operation management system 111 includes a user interface unit 112, a load distribution execution unit 113, an SLO setting unit 114, an SLO monitoring unit 115, an operation system apparatus control unit 116, and a storage device (storage unit) 121. The operation management system 111 is connected to the operation management terminal 101, the load distribution device 142, the management target device (computer) 143, and the expansion candidate device (computer) 153 via the operation management network 131.

The user interface unit 112 includes a program (including modules etc.) that provides a user interface for setting a service level target (target value: hereinafter referred to as “SLO”) for the Web system 141 in the operation management terminal 101. The same applies to “program”).
Note that SLO is a target value of performance information related to response time for a request. For example, the time from when a load balancer receives a request from a client until it returns (transfers) the processing result by the server to the client. That is.

The load distribution execution unit 113 is a program having functions such as supervision of the SLO setting unit 114, the SLO monitoring unit 115, and the operation system apparatus control unit 116.
The SLO setting unit 114 is a program having a function of managing SLO setting information set through the user interface unit 112.
The SLO monitoring unit 115 is a program having a function of managing and monitoring performance information related to SLO setting information.
The operation system apparatus control unit 116 is a program having a function of controlling the Web system 141 and the expansion candidate apparatus group 151.
Further, the load distribution execution unit 113 can cope with an event notification from the load distribution device 142 (a notification regarding a request from the client 171).

The storage device 121 includes an SLO performance information table 122, an SLO setting information table 123, and an extracted Web content information table 124.
The SLO performance information table 122 is a table for the SLO monitoring unit 115 to store data.
The SLO setting information table 123 is a table for the SLO setting unit 114 to store data.
The extracted Web content information table 124 is a table for the load distribution execution unit 113 to store data.

  The storage device 121 may be of any type as long as it can store data such as a hard disk. Further, the SLO performance information table 122, the SLO setting information table 123, and the extracted Web content information table 124 may be stored not in the operation management system 111 but in other devices connected to the operation management network 131.

The Web system 141 includes a load distribution device 142 and a plurality of management target devices 143 (hereinafter also referred to as “hosts”), which are connected to the customer access network 161 and the operation management network 131, respectively.
The Web system 141 is accessed from a plurality of clients 171 via the Internet 181.

  The load balancer 142 stores a URL distribution information table (request distribution table) 191 that stores information in which URL pattern / distribution destination host pairs are grouped as a policy, and a policy that stores pair information of a virtual server IP address and an applied policy. The application information table 192 and the SLO setting information table 193 which is a table having the same contents as the SLO setting information table 123 in the storage device 121 are included.

  The operation system device control unit 116 of the operation management system 111 performs transmission / reception of instruction sequences by Telnet, etc., and transmission / reception of setting files by FTP (File Transfer Protocol), TFTP (Trivial File Transfer Protocol), etc. The load balancer 142 is controlled.

The management target device 143 includes a processor, a memory, a disk storage, a network interface, and the like, and an operating system (hereinafter referred to as OS) and various processes including a server operate on the OS.
The management target device 143 has a management agent 144 that is a program for managing a server group in the same device, and a management target Web server 145.

The management target Web server 145 is a server (software) controlled by the management agent 144 that has received a command from the operation system apparatus control unit 116.
The management target Web server 145 includes a management target business program 146 that is one element of the business application, a request transfer information table 148 that stores pair information between the URL pattern and the transfer destination AP server, and pair information between the URL pattern and the Web content storage folder. It has a stored URL mapping information table 149 and an SLO setting information table 194 which is a table having the same contents as the SLO setting information table 123 in the storage device 121.
The management target business program 146 has a plurality of management target Web contents 147.

  A preferable example of the management target Web server 145 is an HTTP server in Apache Foundation Software (hereinafter referred to as “Apache”), and in this case, the URL mapping information table 149 partially includes the contents of the setting file of the HTTP server. become.

Further, as a suitable example of the request transfer information table 148, the contents of a setting file of a Jakarta Tomcat Connection, which is a redirect function between the Web server and the AP server of The Apache Jakarta Project, correspond.
In the present embodiment, it is assumed that the Web system 141 is based on J2EE (registered trademark: Java (registered trademark) 2 Platform, Enterprise Edition), but is not limited to J2EE (registered trademark). .

The expansion candidate device group 151 includes a plurality of expansion candidate devices 153 to be added at the time of load distribution execution.
Like the management target device 143, the expansion candidate device 153 includes a processor, memory, disk storage, a network interface, and the like, and various processes including a server operate on the OS and the OS.

Further, the expansion candidate device 153 holds a management agent 154 that is a program for managing a server group in the same device and an expansion candidate Web server 155.
The operation system apparatus control unit 116 of the operation management system 111 controls the expansion candidate Web server 155 by communicating with the management agent 154.

  The expansion candidate Web server 155 is a server (software) having a request transfer information table 158 and a URL mapping information table 159. Further, the expansion candidate Web server 155 collectively deploys a plurality of extracted Web contents 157 as an extracted business program 156 when executing load distribution.

Next, the hardware configuration of each device in FIG. 1 will be described with reference to FIG. 2 (see FIG. 1 as appropriate). FIG. 2 is a hardware configuration diagram of each device in FIG.
The information processing device 201 is a computer device corresponding to the operation management system 111, the load distribution device 142, the management target device 143, and the expansion candidate device 153, and includes a storage unit 202, a CPU (Central Processing Unit) 203, and a communication unit. 204 is comprised.

The storage unit 202 stores information such as various programs, various tables, and various contents, and can be realized by a memory, a ROM (Read Only Memory), a hard disk, or the like.
The CPU 203 performs various arithmetic processes based on information (such as the user interface unit 112 in FIG. 1) such as various programs stored in the storage unit 202.
The communication unit 204 is a communication interface used when communicating with other devices.

Next, a rough processing flow of the present embodiment will be described with reference to FIG.
First, the operation manager of the Web system 141 accesses the operation management system 111 via the operation management network 131 by operating the input unit 103 of the operation management terminal 101.

The operation management system 111 prompts the operation administrator to set the SLO by displaying on the display unit 102 of the operation management terminal 101 using the user interface unit 112. The operation manager inputs each setting item (see FIG. 3) of the SLO using the input unit 103, and sends an end notification to the operation management system 111.
After the end notification, the load distribution execution unit 113 updates the SLO setting information table 123 using the SLO setting unit 114. At the same time, the load distribution execution unit 113 uses the operation system apparatus control unit 116 to notify the load distribution apparatus 142 of the contents of the SLO setting information and to the management target Web server 145 via the management agent 144. Notify the contents of.

Next, when receiving a request from the client 171, the load distribution apparatus 142 refers to a URL (hereinafter referred to as “request URL”) included in the request and the SLO setting information table 123 of the operation management system 111. If the request belongs to the URL pattern to be monitored by SLO, the load distribution apparatus 142 transfers the request to the management target Web server 145 and measures the response time for the request URL.
Note that the response time is the time from when the load balancer 142 receives a request from the client 171 until it returns a response to the client 171 (transfer of the processing result by the management target device 143).

  The managed Web server 145 to which the request has been transferred refers to the request URL and the SLO setting information table 123 in the same manner as the load balancer 142. If the request belongs to the URL pattern to be monitored by SLO, the management target Web server 145 measures the server processing time and sends the server processing time and the processing result for the request URL to the load balancer 142.

  After returning the processing result for the request to the client 171, the load distribution apparatus 142 notifies the load distribution execution unit 113 of the request URL, response time, and server processing time. Upon receiving the notification, the load distribution execution unit 113 updates the SLO performance information table 122 using the SLO monitoring unit 115.

The SLO monitoring unit 115 periodically checks the SLO performance information table 122 for the presence of a request URL that exceeds (does not satisfy) the SLO, and if there is a request URL that exceeds the SLO, The storage location of the corresponding Web content is stored in the extracted Web content information table 124.
Further, the SLO monitoring unit 115 notifies the operation manager of the presence of the request URL exceeding the SLO from the contents of the SLO setting information table 123 and the extracted Web content information table 124 by the display unit 102 of the operation management terminal 101. Or perform load balancing automatically.

When performing load distribution automatically, the load distribution execution unit 113 converts the Web content extracted from the management target Web content 147 into the extracted Web content 157 of the extraction work program 156 based on the information in the extracted Web content information table 124. Deploy them together.
Furthermore, the load distribution execution unit 113 updates the request transfer information table 158 and the URL mapping information table 159 of the expansion candidate Web server 155, and activates the expansion candidate Web server 155 if the expansion candidate Web server 155 is not activated. .

Thereafter, the load distribution execution unit 113 updates the URL distribution information table 191 and the policy application information table 192 of the load distribution apparatus 142 and incorporates the expansion candidate apparatus 153 into the Web system 141.
When load distribution is not automatically performed, the operation manager manually performs load distribution by operating the input unit 103.

Next, each screen and each table will be described with reference to FIGS. 3 to 11 (see FIG. 1 as appropriate).
FIG. 3 is an example of an SLO setting screen displayed on the display unit of the operation management terminal. The SLO setting screen 301 includes an add button 311, a delete button 312, a text field 321, a text field 322, a check box 323, a table 341, an OK button 351, and a cancel button 352.
Hereinafter, the operation manager performs each setting using the input unit 103.

The text field 321 is for inputting the URL pattern of the management target request, and the text field 322 is for inputting the target value of the average response time corresponding to the URL pattern. This average response time may be appropriately determined by the operation manager from various circumstances such as the type (text data, image data, etc.) and size of the Web content.
The check box 323 is used to select whether or not to automatically perform load balancing when the management target request exceeds the average response time target value input in the text field 322. The operation manager can appropriately select the presence / absence of automation in consideration of various circumstances by using the check box 323.

When the add button 311 is pressed after these inputs and selections, they are reflected in the URL pattern 332, the average response time 333, and the automatic load distribution 334, which are the columns of the table 341.
The row 342 shows an example of the setting contents of the SLO. The URL pattern is “/ shopping /”, the target value of the average response time is 3 seconds, and automatic load balancing is ON. In this case, the actual average When the response time exceeds 3 seconds, load balancing is automatically executed.

Thereafter, when the OK button 351 is pressed, the load distribution execution unit 113 stores the contents input / selected on the SLO setting screen 301 in the SLO setting information table 123 using the SLO setting unit 114.
When the cancel button 352 is pressed, the input / selected content is cancelled.

  FIG. 4 is a detailed view of an SLO performance information table for storing data by the SLO monitoring unit. The SLO performance information table 122 includes a request URL 411, an average response time 412, an average server processing time 413, and an access count 414. In addition, lines 421 and 422 show examples of the contents of the SLO performance information table 122.

The request URL 411 represents the URL for each request received from the client 171.
The average response time 412 represents the average response time for each request URL.
The average server processing time 413 represents the average server processing time for each request URL.
The number of accesses 414 represents the number of accesses for each request URL.

  FIG. 5 is a detailed view of an SLO setting information table for storing data by the SLO setting unit. The SLO setting information table 123 includes a URL pattern 512, an average response time 513, and an automatic load distribution 514. A row 521 shows an example of SLO setting information, and indicates that load distribution is automatically executed when the average response time of the URL pattern “/ shopping /” exceeds 3 seconds.

Returning to FIG. 4, the contents of the row 421 and the row 422 will be described (see FIG. 5 as appropriate).
The request URL 411 in the row 421 matches (matches) the URL pattern “/ shopping /” in the row 521 which is an example of the contents of the SLO setting information table 123, and the average response time is 3.897 seconds, and the average server processing time. In 0.804 seconds, the number of accesses is 35444783 times.

  Further, the request URL 411 in the row 422 matches the URL pattern “/ shopping /” in the row 521 which is an example of the contents of the SLO setting information table 123, and the average response time is 2.523 seconds, and the average server processing time. 2. In 304 seconds, the number of accesses is 3,546,782 times.

FIG. 6 is a detailed view of the extracted Web content information table. The extracted Web content information table 124 includes a URL 602, a host IP 603, and a Web content storage location 604.
A row 611 is an example of the contents of the extracted Web content information table 124. When the URL is “http://a.com/shopping/img/item1.jpg”, the corresponding Web content has an IP address “192. This indicates that the host of “168.10.24” exists in “C: \ somewhere1 \ shopping \ img \ item1.jpg”.

  The URL pattern means a type of URL including specific information. In the present embodiment, for example, it is used to mean a URL having “/ shopping /”, but in addition, “http: //A.com/shopping/img/item1.jpg "may mean the whole address.

FIG. 7 is a screen example of the extracted Web content information table. On the display unit 102 of the operation management terminal 101, the URL 602, host IP 603, and Web content storage location 604 information of the extracted Web content information table 124 of FIG. 6 are displayed in the URL 702, host IP 703, and Web content storage location 704, respectively. .
By looking at this screen, the operation manager can recognize information related to high-load Web content.

FIG. 8 is a detailed diagram of a URL distribution information table existing in the load balancer.
The URL distribution information table 191 includes a policy name 801, a URL pattern 803, a host group 805, and a host 807.
Lines 811 and 813 show an example of contents after execution of load distribution in the URL distribution information table 191. The line 811 is set before execution of load distribution. When the URL pattern is “/ shopping /”, the load distribution apparatus 142 exists on the host whose IP address is “192.168.10.24”. The content of entrusting the processing to the management target Web server is shown. A row 813 shows the contents of the setting added after execution of load distribution.
The host group 805 represents a host group.

In this case, when the URL pattern of the request is “/shopping/img/item1.jpg”, the load balancer 142 newly adds (adds) a host (IP address is “192.168.10. 26 ″), the processing is entrusted to the management target Web server.
In addition, a policy name (URL distribution pattern name) is assigned to each setting in the rows 811 and 813, a policy name “Policy1” is assigned to the row 811, and a policy name “Policy2” is assigned to the row 813. Yes.

FIG. 9 is a detailed diagram of a policy application information table existing in the load balancer.
The policy application information table 192 includes a virtual IP address 901 and an application policy order 903.
In a row 911, an example of contents after executing load distribution in the policy application information table 192 is shown.

A virtual IP address 901 (a virtual IP address 925 here) means an IP address (a virtual IP address of the load balancer 142) that the client 171 accesses through the Internet 181.
In the applied policy order 903, the order of applied policies is registered in order from the top.

  Here, when the client 171 accesses the virtual IP address 925, the application policy 921 is applied first, and then the application policy 923 is applied. Detailed setting contents for the application policy 921 and the application policy 923 are stored in the URL distribution information table 191.

For example, when a request including a request URL “http://a.com/shopping/img/item1.jpg” comes from the client 171, the load balancer 142 attempts to apply the application policy 921. Detailed settings of the application policy 921 are in the row 813 of the URL distribution information table 191. Since the URL pattern of the line 813 is “/Shopping/img/item1.jpg” and matches the request URL of the client 171, the load balancer 142 is on the host whose IP address is “192.168.10.26”. Outsources the processing to the management target Web server existing in the server.
This makes it possible to distribute the processing load in units of Web content.

FIG. 10 is a detailed diagram of a request transfer information table held by the management target Web server.
The request transfer information table 148 includes a URL pattern 1002, a host group 1003, and a host IP 1004.
A row 1011 is a content example of the request transfer information table 148. Here, when the URL pattern is “/ purchase /”, the IP address is “192.168.1.21” or “192.168.1.22” without being processed by the management target Web server 145. Or the processing is entrusted to the AP server existing on the host “192.168.1.23”.

FIG. 11 is a detailed diagram of a URL mapping information table held by the management target Web server.
The URL mapping information table 149 includes a URL pattern 1102 and a Web content storage folder 1103.

A row 1111 is an example of the contents of the URL mapping information table 149. When the URL pattern is “/ shopping /”, the Web content specified by the URL below “/ shopping /” is “C: \ somewhere1 \ shopping”. "I mean it's in a folder."
That is, for example, when the request URL is “http://a.com/shopping/img/item1.jpg”, the corresponding Web content exists in “C: ¥ somewhere1 ¥ shopping ¥ img ¥ item1.jpg”. To do.

Hereinafter, each operation in the present embodiment will be described with reference to FIGS. 12 to 18 (see FIG. 1 and the like as appropriate).
FIG. 12 is a flowchart when the load balancer performs performance monitoring based on the SLO setting information.
First, the load balancer 142 receives a request from the client 171 via the Internet 181 and the customer access network 161 (step S1201).

  Next, based on the SLO setting information table 193, the load balancer 142 determines whether the request URL matches the URL pattern of the SLO setting information, that is, whether the request URL is a monitoring target (step). S1211).

If the request URL is a monitoring target (Yes in step S1211), the load balancer 142 starts measuring the response time for the request and records it as the start time T1 (step S1221).
Next, the load balancer 142 sends the request to the management target Web server 145 and entrusts the processing (step S1231).

Subsequently, the load balancer 142 acquires the processing result corresponding to the request URL and the server processing time from the management target Web server 145, and transmits the processing result to the client 171 (step S1241).
Next, the load balancer 142 ends the response time measurement and records it as the end time T2 (step S1251).

Thereafter, the load balancer 142 obtains a response time by calculating T2-T1 from the start time T1 obtained in step S1221 and the end time T2 obtained in step S1251, and obtains the request URL, response time, and server processing time. Is notified to the load distribution execution unit 113 (step S1261).
Then, the load distribution execution unit 113 uses the SLO monitoring unit 115 to update the SLO performance information table 122 based on the received request URL, response time, and server processing time.
In this way, the SLO performance information table 122 can be updated.

  If the request URL is not a monitoring target (No in step S1211), the load distribution apparatus 142 sends the request to the management target web server 145 and entrusts the processing (step S1235), and the processing result for the request URL is managed by the management target web server 145. And the processing result is transmitted to the client 171 (step S1245).

FIG. 13 is a flowchart when the management target Web server measures the server processing time based on the SLO setting information. The server processing time measured here is used in step S1241 in the flowchart of FIG.
First, the management target web server 145 receives a request from the load balancer 142 (step S1301).

  Next, the management target Web server 145 determines, based on the SLO setting information table 194, whether the request URL matches the URL pattern of the SLO setting information, that is, whether the request URL is a monitoring target ( Step S1311).

If the request URL is a monitoring target (Yes in step S1311), the management target Web server 145 starts measuring the server processing time and records it as the start time T1 (step S1321).
Next, the management target web server 145 executes processing for the request URL (step S1331).

Thereafter, the management target web server 145 ends the measurement of the server processing time and records it as the end time T2 (step S1341).
Subsequently, the management target Web server 145 obtains the server processing time by calculating T2-T1 from the start time T1 obtained in step S1321 and the end time T2 obtained in step S1341, and obtains the request URL and the response to it. The processing result and server processing time to be transmitted are transmitted to the load balancer 142 (step S1351).
In this way, the management target Web server 145 can acquire the server processing time.

  If the request URL is not a monitoring target (No in step S1311), the management target Web server 145 performs processing for the request URL (step S1335), and transmits the request URL and the processing result to the load balancer 142 (step S1355).

FIG. 14 is a flowchart when the SLO monitoring unit updates the SLO performance information table. The process according to the flowchart of FIG. 14 is executed after the process of step S1261 in the flowchart of FIG.
First, the SLO monitoring unit 115 acquires a request URL, response time, and server processing time from the load distribution execution unit 113 (step S1401).
Next, the SLO monitoring unit 115 determines whether the request URL acquired in step S1401 exists in the SLO performance information table 122 (step S1405).

When the request URL exists in the SLO performance information table 122 (Yes in step S1405), the SLO monitoring unit 115 uses the request URL acquired in step S1401 from the SLO performance information table 122 as a key, and the corresponding average response time 412, The average server processing time 413 and the number of accesses 414 are acquired (step S1411).
Subsequently, the SLO monitoring unit 115 calculates (average response time 412 × access count 414 + response time acquired in step S1401) ÷ (access count 414 + 1), and updates the average response time 412 (step S1421).

Next, the SLO monitoring unit 115 calculates (average server processing time 413 × access number 414 + server processing time acquired in step S1401) ÷ (access number 414 + 1), and updates the average server processing time 413 (step S1431). ).
Thereafter, the SLO monitoring unit 115 calculates (access count 414 + 1) and updates the access count 414 (step S1441).

When the request URL does not exist in the SLO performance information table 122 (No in step S1405), the SLO monitoring unit 115 adds a row to the SLO performance information table 122 and inserts the request URL acquired in step S1401 into the request URL 411 ( Step S1415).
Next, the SLO monitoring unit 115 inserts the response time acquired in step S1401 into the average response time 412 (step S1425) and adds the server processing time acquired in step S1401 to the average server processing time 413 in step S1415. It is inserted (step S1435).

Thereafter, the SLO monitoring unit 115 inserts 1 into the access count 414 in the same row in step S1415 (step S1445).
In this way, the SLO performance information table 122 can be updated.

15 and 16 are flowcharts when the SLO monitoring unit generates the extracted Web content information table from the SLO performance information table.
In the generation of the extracted Web content information table 124 according to the flowcharts of FIGS. 15 and 16, the SLO monitoring unit 115 executes the following steps for each row of the SLO performance information table 122.

First, the SLO monitoring unit 115 searches for the corresponding URL pattern 512 from the SLO setting information table 123 based on the request URL 411 of the SLO performance information table 122 (step S1501).
Next, the SLO monitoring unit 115 searches the SLO setting information table 123 for an average response time 513 corresponding to the URL pattern 512 (step S1511).
Subsequently, the SLO monitoring unit 115 determines whether the average response time 412 in the SLO performance information table 122 exceeds the average response time 513 searched in step S1511 (step S1521).

If the average response time 412 exceeds the average response time 513 (Yes in step S1521), the SLO monitoring unit 115 searches all rows corresponding to the URL pattern 512 from the SLO performance information table 122, and performs average server processing. A request URL 411 having a long time 413 is detected (step S1531).
Note that the request URL 411 having a long average server processing time 413 corresponds to, for example, a request URL 411 having a higher average server processing time 413 of 10% or a request URL 411 having an average server processing time 413 exceeding 2 seconds. To do.

Next, the SLO monitoring unit 115 searches the extracted Web content information table 124 for a row corresponding to the request URL 411 obtained in step S1531 (step S1541).
If the corresponding line does not exist in the extracted Web content information table 124 (No in step S1541), the request transfer information table 148 is searched for a URL pattern corresponding to the request URL 411 obtained in step S1531 (step S1551).
If the extracted Web content information table 124 includes the corresponding row (Yes in step S1541), the extracted Web content information table 124 is updated because the row corresponding to the request URL 411 already exists in the extracted Web content information table 124. do not have to.

Note that step S1551 is processing unique to a J2EE (registered trademark) -based Web system. If the corresponding URL pattern exists as a result of the search in step S1551, it means that the corresponding Web content exists on the business program deployed in the AP server.
In other words, when high-load Web content on the AP server is extracted and load distribution is performed, the setting of the Web server associated with the AP server that is the load distribution target must be changed and restarted. It will be paused.

  In this embodiment, load distribution for a Web server that does not need to be restarted is assumed assuming 24-hour non-stop, so load distribution for a Web server will be described. Since it is determined in step S1551 whether the Web server or the AP server is used, if this embodiment is applied, it is possible to deal with load balancing for the AP server.

If there is no corresponding row in step S1551 (No), the SLO monitoring unit 115 searches the URL mapping information table 149 for the URL pattern 1102 corresponding to the request URL 411 obtained in step S1531 (step S1561).
Next, the SLO monitoring unit 115 searches the URL mapping information table 149 for the Web content storage folder 1103 corresponding to the URL pattern 1102 obtained in step S1561 (step S1571).

Subsequently, the SLO monitoring unit 115 calculates the full path of the web content storage location from the request URL 411 obtained in step S1531 and the web content storage folder 1103 obtained in step S1571 (step S1581).
Next, the SLO monitoring unit 115 communicates with the management agent 144 existing in the management target device 143 using the operation system device control unit 116 held by the load distribution execution unit 113, and maintains the URL mapping information table 149. The IP address of the target Web server 145 is acquired (step S1591).

Subsequently, the SLO monitoring unit 115 stores the request URL 411 obtained in step S1531, the IP address of the management target web server 145 obtained in step S1591, and the web content storage location calculated in step S1581 in the extracted web content information table 124. Are stored (step S1592).
In this way, the SLO monitoring unit 115 can generate the extracted Web content information table 124.

Thereafter, the SLO monitoring unit 115 performs notification processing to the operation manager or automatic load distribution processing (step S1593).
Here, the process of step S1593 is executed after step S1592, but the load distribution execution unit 113 executes the process while referring to the extracted Web content information table 124 as shown in the flowcharts of FIGS. You may do it. Further, the processing in step S1593 may be executed in accordance with the operation manager's specification at the time of SLO setting.

FIGS. 17 and 18 are flowcharts showing a method for notifying the operation manager of the extracted high-load Web content based on the presence or absence of automatic load distribution in the SLO setting information table and a method for automatically distributing the load. It is.
First, the load distribution execution unit 113 selects an expansion candidate device 153 from the expansion candidate device group 151 (step S1601), and holds the information in a memory (corresponding to the storage unit 202 in FIG. 2).

Next, the load distribution execution unit 113 turns off the copied flag, which is a flag indicating whether or not copy processing has been performed on the extracted Web content (step S1602), and stores the information in the memory (the storage unit in FIG. 2). 202).
Thereafter, the load distribution execution unit 113 repeats the processing from step S1603 to step S1641 for the number of rows in the extracted Web content information table 124.

The load distribution execution unit 113 acquires the average response time 412 from the SLO performance information table 122 and the average response time 513 from the SLO setting information table 123 based on the URL 602 of the extracted Web content information table 124 (step S1603). It is determined whether the response time 412 exceeds the average response time 513 (step S1604).
Note that the Web content stored in the extracted Web content information table 124 is the Web content extracted as the average response time 412 exceeds the average response time 513 according to the flowchart of FIG. Therefore, in step S1604, it is determined again whether the average response time 412 of the Web content exceeds the average response time 513.

When the average response time 412 exceeds the average response time 513 (Yes in step S1604), the load distribution execution unit 113 displays the URL pattern 512 corresponding to the URL 602 of the extracted Web content information table 124 from the SLO setting information table 123. Search is performed (step S1611).
Next, the load distribution execution unit 113 determines whether the automatic load distribution 514 corresponding to the URL pattern 512 searched in step S1611 is ON (step S1621).

  When the automatic load distribution 514 is ON (Yes in step S1621), the load distribution execution unit 113 extracts the extraction work program 156 to the expansion candidate device 153 based on the host IP 603 and the Web content storage location 604 of the extracted Web content information table 124. As the extracted Web content 157, a copy process is performed (step S1631), and a copy (process) completed flag is turned ON (step S1641).

  When the automatic load distribution 514 is OFF (No in step S1621), the load distribution execution unit 113 uses the user interface unit 112 to correspond to the contents of the extracted Web content information table 124 on the display unit 102 of the operation management terminal 101. By displaying the contents of the SLO performance information table 122 and the contents of the corresponding SLO setting information table 123, they are notified to the operation manager (step S1633).

Next, the load distribution execution unit 113 determines whether the copied flag is ON, that is, whether the process of step S1641 has been performed once (step S1651).
In the case of Yes in step S1651, the load distribution execution unit 113 selects the contents of the request transfer information table 148 and the URL mapping information table 149 of the management target Web server 145 existing in the host IP 603 in step S1631 in step S1601. The request transfer information table 158 and the URL mapping information table 159 existing in the added candidate device 153 are stored (step S1661).

  After that, the load distribution execution unit 113 uses the operation system apparatus control unit 116 to activate the expansion candidate Web server 155 existing in the expansion candidate apparatus 153 (step S1671), and updates the URL distribution information table 191 of the load distribution apparatus 142 ( Step S1681) and the policy application information table are updated (Step S1691), and the load distribution is completed.

In the case of No in step S <b> 1651, the operation manager can manually perform load distribution using the load distribution execution unit 113 by operating the input unit 103 of the operation management terminal 101.
In this way, load distribution regarding high-load Web content can be performed manually or automatically.

  As described above, according to the present embodiment, only the extracted high-load specific Web content is deployed on the target server. Therefore, the number of servers (or hosts) is smaller than the conventional load distribution by adding business application units. )), And it is possible to cope with load distribution in an emergency by shortening the deployment time.

In this embodiment, the explanation has been made assuming that the expansion candidate device 153 exists in the expansion candidate device group 151. However, the expansion is possible because the number of devices (computers) to be operated is limited. There may be no candidate device (computer). Even in this case, the operation manager registers the operation candidate system in the operation management system 111 in consideration of the load status of the management target apparatus (computer) existing in the Web system 141. It is possible to perform load balancing in the manner described.
Furthermore, since requests for high-load specific Web content are distributed to the added server, response time can be improved as compared with the conventional high-load Web content in the business program.

This is the end of the description of the embodiments, but the aspects of the present invention are not limited to these. For example, the present invention can be applied to a case where load distribution is performed by a round robin method instead of using a load distribution apparatus.
In addition, the specific configuration of each device, each program, and each flowchart can be appropriately changed (integrated, separated, etc.) without departing from the spirit of the present invention.

It is a whole block diagram of this embodiment. It is a hardware block diagram of each apparatus in FIG. It is an example of the setting screen of SLO displayed on the display part of an operation management terminal. It is a detailed view of an SLO performance information table. It is a detailed view of an SLO setting information table. It is a detailed view of an extracted Web content information table. It is an example of a screen of an extraction Web content information table. It is a detailed view of a URL distribution information table. It is a detailed view of a policy application information table existing in the load balancer. It is a detailed view of a request transfer information table held by a management target Web server. It is a detailed view of a URL mapping information table held by a management target Web server. It is a flowchart when a load distribution apparatus performs performance monitoring based on SLO setting information. It is a flowchart when a management object web server measures server processing time based on SLO setting information. It is a flowchart when an SLO monitoring part updates a SLO performance information table. It is the first half of a flowchart when an SLO monitoring part produces | generates an extraction Web content information table from an SLO performance information table. It is the latter half of the flowchart when an SLO monitoring part produces | generates an extraction Web content information table from an SLO performance information table. It is the first half of the flowchart which showed the method of notifying an operation manager of the extracted high load web content, and the method of carrying out load distribution automatically. It is the second half of the flowchart showing a method for notifying the operation manager of the extracted high-load Web content and a method for automatically distributing the load.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Operation management terminal 111 Operation management system 113 Load distribution execution part 114 SLO setting part 115 SLO monitoring part 116 Operation system apparatus control part 121 Storage device 122 SLO performance information table 123 SLO setting information table 124 Extraction Web content information table 141 Web system 142 Load balancer 143 Managed device 145 Managed web server 146 Managed business program 147 Managed web content 151 Additional candidate device group 153 Additional candidate device 171 Client 201 Information processing device 202 Storage unit 203 CPU

Claims (8)

A computer system comprising a load balancer responding to a request from a client, a plurality of computers having Web contents or application programs, and an operation management computer for managing the load balancer and the plurality of computers,
A plurality of computers having either the Web content or the application program has a request transfer information table that indicates the correspondence between the URL pattern of the request and the transfer destination computer,
The operation management computer includes a storage unit and a processing unit,
The storage unit
For each URL pattern of the request, store a target value of performance information related to the response time of the request,
The processor is
For each URL pattern of the request, monitor performance information regarding the response time of the request;
It is determined whether the monitored performance information satisfies a target value of the performance information read from the storage unit, and Web content corresponding to a URL pattern of a request that does not satisfy the target value of the performance information is transferred to the request by referring to the information table, the Web content, or computer with Web content, to identify whether it is deployed in Re computer or noise having an application program, deployed to the computer to which the Web content has an application program A computer system that distributes the load of a Web server by moving the Web content to another computer when it is not specified that the Web content has been specified . The computer system according to claim 1 ,
The storage unit
For Web content corresponding to a URL pattern of a request that does not satisfy the target value of the performance information,
By referring to the URL mapping information table that indicates the correspondence between the URL pattern and the storage location of the Web content corresponding to the URL pattern, held by the computer having the Web content or application program,
Specify the storage location of the Web content corresponding to the URL pattern of the request that does not satisfy the target value of the performance information,
Storing a URL pattern of a request that does not satisfy the target value of the performance information and an extracted Web content information table indicating a correspondence between a storage location of the Web content corresponding to the URL pattern of the request;
The processor is
A computer system, characterized in that, using the extracted Web content information table, setting is made for another computer that is a destination of the Web content corresponding to the URL pattern of the request. The computer system according to claim 1 ,
A load balancer that distributes requests from clients to the plurality of computers;
The load balancer has a request distribution table that indicates the correspondence between the URL pattern of the request and the computer that is the distribution destination,
The processor is
A computer system, wherein when performing load distribution, the load distribution apparatus is set using the request distribution table. The computer system according to claim 1 ,
The target value of the performance information includes either an average response time for the request or an average processing time of the computer for the request. A computer control method in a computer system comprising a load balancer responding to a request from a client, a plurality of computers having Web contents or application programs, and an operation management computer for managing the load balancer and the plurality of computers. ,
The computer system is
A plurality of computers having either the Web content or the application program has a request transfer information table that indicates the correspondence between the URL pattern of the request and the transfer destination computer,
The operation management computer includes a storage unit and a processing unit,
The storage unit
For each URL pattern of the request, store a target value of performance information related to the response time of the request,
The processor is
For each URL pattern of the request, monitor the performance information related to the response time of the request, determine whether the monitored performance information satisfies the target value of the performance information read from the storage unit,
By referring to the request transfer information table for the Web content corresponding to the URL pattern of the request that does not satisfy the target value of the performance information , the Web content is a computer having the Web content or a computer having an application program. and determine whether is deployed in Re noise, if the Web content is not identified as being deployed to the computer having an application program, by moving the Web content to another computer, Web server A computer control method characterized by performing load balancing. A computer control method according to claim 5 , comprising:
The storage unit
For Web content corresponding to a URL pattern of a request that does not satisfy the target value of the performance information,
By referring to the URL mapping information table that indicates the correspondence between the URL pattern and the storage location of the Web content corresponding to the URL pattern, held by the computer having the Web content or application program,
Specify the storage location of the Web content corresponding to the URL pattern of the request that does not satisfy the target value of the performance information,
Storing a URL pattern of a request that does not satisfy the target value of the performance information and an extracted Web content information table indicating a correspondence between a storage location of the Web content corresponding to the URL pattern of the request;
The processor is
A computer control method characterized in that, using the extracted Web content information table, setting is made for another computer to which the Web content corresponding to the URL pattern of the request is moved. A computer control method according to claim 5 , comprising:
The computer system further includes a load balancer that distributes requests from clients to the plurality of computers,
The load balancer has a request distribution table that indicates the correspondence between the URL pattern of the request and the computer that is the distribution destination,
The processor is
A computer control method, comprising: setting the load balancer using the request distribution table when performing load balancing. A computer control method according to claim 5 , comprising:
The target value of the performance information includes either an average response time for the request or an average processing time of the computer for the request.

Images