JP2015118659A - Apparatus, method, and program for load distribution - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve load distribution among Web servers.SOLUTION: A session management unit 11 manages a session ID in a HTTP request by using a plurality of management tables with different session management time. A second session determination unit 14 determines whether or not a total number of session IDs stored in each of the plurality of management tables has reached a respective upper limit of the number of sessions processible by a Web server 3 in a session management time of each management table. A request control unit 15 either transfers or does not transfer the HTTP request from a client terminal 2 for the Web server 3 to the Web server 3 depending on a result of the determination.

Description

  The present invention relates to a technique for distributing a load on a server.

  In order to prevent a decrease in service due to an increase in requests, a load balancer is arranged in front of the Web server. The load balancer once receives an HTTP request from a client terminal, and if it is the first request, transfers it to an arbitrary Web server among a plurality of Web servers. On the other hand, if it is the second and subsequent requests, the request is transferred to the Web server that issued the session ID in the request in order to guarantee consistent communication between the client terminal and the Web server (see Non-Patent Document 1).

“Introduction to Load Balancing 6th Session Maintenance Function”, Fujitsu Limited, [online], [Searched on December 11, 2013], <URL: http://fenics.fujitsu.com/products/ipcom/catalog/ data / 1 / 6.html>

  However, since HTTP requests are always transferred in any case, there is a problem that it is difficult to reliably reduce the load on the Web server even if a load balancer is interposed. For example, when a message is registered from a client terminal to a disaster message dial in the event of a disaster, the message terminal that provides several levels of message dials provided in a certain wide area such as East Japan or West Japan will have tens of millions of client terminals. It is extremely difficult to accept and reliably process all HTTP requests from.

  The present invention has been made in view of the above circumstances, and an object thereof is to improve the load on a Web server.

  The load distribution apparatus according to claim 1, a management unit that manages session information in a request using a plurality of storage units having different session management times, and a reception unit that captures a request from the terminal device to the web providing device. And determining whether or not the total number of session information respectively stored in the plurality of storage means has reached each upper limit number of sessions that can be processed by the web providing device in each session management time of each storage means The gist of the invention is to include a determination unit and a control unit that transfers or does not transfer the capture request to the web providing device based on the determination result.

  According to the present invention, session information in a request is managed using a plurality of storage means having different session management times, and the total number of session information respectively stored in the plurality of storage means In order to determine whether each upper limit session number that can be processed by the web providing device in the management time has been reached, and to transfer or non-transfer the request from the terminal device to the web providing device based on the result of the determination, The load on the providing device can be reduced.

  The load distribution apparatus according to claim 2 is the load distribution apparatus according to claim 1, wherein the determination unit is configured to include the determination request in a case where any of the plurality of storage units has reached the upper limit number of sessions. If the session information is not in the storage means, the control means determines that the session information is not transferred to the terminal device without transferring the capture request. The gist is to send.

  According to the present invention, when any of the plurality of storage means has reached the upper limit number of sessions, it is determined whether the session information in the captured request is in the storage means with the longest session management time, If the request is not stored in the storage means, the load on the web providing apparatus can be reduced more reliably.

  According to a third aspect of the present invention, there is provided a load balancing method according to a third aspect of the present invention, comprising: managing steps for managing session information in a request using a plurality of storage units having different session management times; A reception step and whether or not the total number of session information respectively stored in the plurality of storage means has reached each upper limit number of sessions that can be processed by the web providing device in each session management time of each storage means The gist of the present invention is to include a determination step for determining each and a control step for transferring or not transferring the capture request to the web providing device based on the result of the determination.

  The load distribution method according to claim 4 is the load distribution method according to claim 3, wherein in the determination step, if any of the plurality of storage means has reached the upper limit number of sessions, Whether or not the session information is stored in the storage unit with the longest session management time. If the storage step does not exist in the storage unit, the non-transfer result information is not transferred to the terminal device without transferring the capture request. The gist is to send.

  The gist of a load distribution program according to a fifth aspect is to cause a computer to execute the load distribution method according to the third or fourth aspect.

  According to the present invention, the load on the Web server can be reduced.

It is a figure which shows the functional block structure of a load distribution apparatus. It is a figure which shows the processing flow of a load distribution apparatus. It is a figure which shows the sequence of the processing pattern A. It is a figure which shows the sequence of the process pattern B. FIG. It is a figure which shows the sequence of the process pattern C. FIG. It is a figure which shows the sequence of the process pattern D. FIG. It is a figure which shows the sequence of the process pattern E. FIG.

  The present invention (1) manages session IDs in HTTP requests using a plurality of management tables having different session management times, and (2) the total number of session IDs stored in the plurality of management tables It is determined whether or not the upper limit number of sessions that can be processed by the Web server has been reached in each session management time of the management table. (3) If any of the management tables has reached the upper limit number of sessions It is determined whether or not the session ID in the captured HTTP request is in the management table with the longest session management time. (4) If the session ID is not in the management table, the HTTP request is not transferred to the client terminal. By transmitting non-transfer result information, the load on the Web server is reduced. Hereinafter, an embodiment for carrying out the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram illustrating a functional block configuration of a load distribution apparatus 1 according to the present embodiment. The load distribution apparatus 1 includes a session management unit 11, a request reception unit 12, a first session determination unit 13, a second session determination unit 14, and a request control unit 15.

  The session management unit 11 is a functional unit that manages session IDs in HTTP requests received so far using a plurality of management tables (storage means) having different session management times. If there is a session ID in the received HTTP request, the session ID is registered in all the management tables, while the session management time in each management table is deleted after the session management time has elapsed. Note that the reception time of the HTTP request may be associated with the session ID and registered.

  The request receiving unit 12 is a functional unit that receives (captures) an HTTP request transmitted from the client terminal 2 to the Web server 3.

  The first session determination unit 13 is a functional unit that determines whether or not a session ID is included in the received HTTP request. Usually, if the HTTP request from the same client terminal 2 is the second time or later, the session ID is included.

  The second session determination unit 14 determines whether the total number of session IDs registered in each management table has reached the upper limit number of sessions that can be processed by the Web server 3 within each session management time of each management table. It is a functional unit for determining whether or not.

  The request control unit 15 is a functional unit that transfers or does not transfer the HTTP request received by the request reception unit 12 based on the determination results made by the first session determination unit 13 and the second session determination unit 14. .

  The above is the function of the load distribution apparatus 1. Note that the load distribution apparatus 1 is disposed between the client terminal 2 and the Web server 3 and is communicably connected to them.

  Next, the operation of the load distribution apparatus 1 according to this embodiment will be described. FIG. 2 is a diagram illustrating a processing flow of the load distribution apparatus 1. It is assumed that the session management unit 11 manages a 30-minute table and an IP table whose session management time is 30 minutes, and a 10-second table whose management time is 10 seconds. The 30-minute table is also used to manage session IDs that have not passed 30 minutes after receiving an HTTP request in the 10-second table. On the other hand, the IP table is for managing the session ID for each client terminal that has not passed 30 minutes after reception.

  First, in step S <b> 1, the request reception unit 12 receives (captures) an HTTP request transmitted from the client terminal 2 to the Web server 3.

  Next, in step S2, the first session determination unit 13 determines whether a session ID is included in the received HTTP request. As a result of the determination, if the session ID is not included, the HTTP request from the client terminal 2 is the first time, so the process proceeds to step S3.

  Next, in step S3, the second session determination unit 14 counts the number of session IDs currently registered in the 30-minute table, and reaches the upper limit number of sessions that can be processed in 30 minutes in the Web server 3. Judge whether it is. Similarly, it is determined whether the number of session IDs in the 10-second table has reached the upper limit number of sessions that can be processed in 10 seconds, and the number of session IDs in the IP table can be processed in 30 minutes. It is also determined whether the maximum number of sessions for each terminal has been reached.

  As a result of each determination, if any of the three types of management tables has reached the upper limit number of sessions, the current Web server 3 is in a high load state. The unit 15 displays “Sorry Page” (non-transfer result information) on the screen of the client terminal 2 that is the transmission source without transferring the HTTP request being received to the Web server 3. The processing pattern A at this time is shown in FIG.

  On the other hand, if none of the three types of management tables has reached the upper limit number of sessions or if there is a free space, the current web server 3 is not in a high load state. 15 transfers the HTTP request being received to the Web server 3. As shown in step S6, since there is no session ID for transferring an HTTP request, the session management unit 11 does not additionally register a session ID in each management table.

  The processing pattern B at this time is shown in FIG. After receiving the HTTP request, the Web server 3 issues a session start SDR (Subscriber Detail Record: access record created for each user) and a session ID if the internal processing for the request is successful, and the requesting client terminal 2 to send. Even if the HTTP request transfer process by the load balancer 1 is successful, if the internal process fails due to the state of the Web server 3, the Web server 3 notifies the client terminal 2 accordingly. .

  Steps S3 to S6 so far are operations of the load distribution apparatus 1 at the time of the initial request, and after the processing, this operation ends. On the other hand, if the result of determination in step S2 is that the session ID is included in the received HTTP request, it is the second and subsequent requests, so the process proceeds to step S7, where the second session determination unit 14 However, the same processing as step S3 is executed.

  As a result of each determination in step S7, if any of the three types of management tables has reached the upper limit number of sessions, the second session determination unit 14 subsequently proceeds to HTTP in step S8. It is determined whether the session ID in the request is registered in the 30-minute table with the longest session management time.

  As a result of the determination, if the session ID is not in the 30-minute table, the request is after the longest session management time has elapsed, and in step S9, the request control unit 15 receives the HTTP being received. Without transferring the request to the Web server 3, “Sorry Page” is displayed on the screen of the client terminal 2 that is the transmission source. Thereafter, the process proceeds to step S11, and the session management unit 11 overwrites the session ID in all the management tables (appends if there is a space). The processing pattern C at this time is shown in FIG.

  On the other hand, if the result of determination in step S8 is that the session ID is in the 30-minute table, even if any of the three types of management tables has reached the upper limit number of sessions, Since the request is a request, in step S 10, the request control unit 15 transfers the HTTP request being received to the Web server 3. Thereafter, the process proceeds to step S11, and the session management unit 11 overwrites all management tables with the session ID. A processing pattern D at this time is shown in FIG.

  Finally, as a result of the determination in step S7, if none of the three types of management tables has reached the upper limit number of sessions, the current web server 3 is not in a high load state, so that the process goes through step S8. In step S10, the request control unit 15 transfers the HTTP request being received to the Web server 3. Thereafter, the process proceeds to step S11, and the session management unit 11 overwrites all the management tables with the session ID in the HTTP request. A processing pattern E at this time is shown in FIG.

  The above is an example of the processing flow of the load distribution apparatus 1. Note that this processing flow is merely an example for easily understanding the operations of the processing patterns A to E. For example, steps S3 and S7 that are common operations are combined into one step, and steps S1 and S2 are performed. You may make it perform between. Further, each determination process of steps S2, S3, and S7 may be performed by one session determination unit.

  As described above, according to the present embodiment, the session management unit 11 manages the session ID in the HTTP request using a plurality of management tables having different session management times, and the second session determination unit 14 It is determined whether the total number of session IDs respectively stored in the management table has reached each upper limit number of sessions that can be processed by the Web server 3 in each session management time of each management table, and the request control unit 15 However, since the HTTP request from the client terminal 2 to the Web server 3 is transferred or not transferred to the Web server 3 based on the determination result, the load on the Web server 3 can be reduced.

  Further, the second session determination unit 14 determines whether the session ID in the captured HTTP request is in the management table with the longest session management time when any of the plurality of management tables has reached the upper limit number of sessions. Since the request control unit 15 does not transfer the HTTP request when it is not in the management table, the load on the Web server 3 can be more reliably reduced.

  As a result of the overload test performed using the load distribution apparatus 1 according to the present embodiment, the processing success rate in the Web server was 45% or more. In this test, since the number of HTTP requests from the client terminal is twice the peak time for the second, 90% or more succeed even if it is the peak time for the second. In other words, since almost all requests can be processed by using the load distribution apparatus 1 described in the present embodiment, the effect that the load on the Web server is appropriately improved is demonstrated.

  In this embodiment, the case where session IDs are managed using three management tables having different session management times has been described as an example. However, these management tables are included in one management table for management. Also good.

  Finally, the load distribution apparatus 1 described in the present embodiment can be realized by a computer having a memory and a CPU. The processing is executed by a program. Furthermore, each operation of the load distribution device 1 can be constructed as a program, installed in a computer and executed, or distributed via a communication network.

DESCRIPTION OF SYMBOLS 1 ... Load distribution apparatus 11 ... Session management part (management means)
12. Request receiving unit (receiving means)
13 ... 1st session determination part 14 ... 2nd session determination part (determination means)
15 ... Request control unit (control means)
2 ... Client terminal (terminal device)
3 ... Web server (Web providing device)
AE ... processing pattern S1-S11 ... step

Claims (5)

Management means for managing session information in the request using a plurality of storage means having different session management times;
Receiving means for capturing a request from the terminal device to the web providing device;
Judgment for respectively determining whether the total number of session information respectively stored in the plurality of storage means has reached each upper limit number of sessions that can be processed by the web providing device in each session management time of each storage means Means,
Control means for transferring or not transferring the capture request to the web providing device based on the result of the determination;
A load balancer comprising: The determination means includes
If any of the plurality of storage means has reached the upper limit number of sessions, determine whether the session information in the capture request is in the storage means with the longest session management time,
The control means includes
2. The load distribution apparatus according to claim 1, wherein if the storage unit is not present, the non-transfer result information is transmitted to the terminal device without transferring the capture request. By computer
A management step for managing session information in a request using a plurality of storage means having different session management times;
A receiving step for capturing a request from the terminal device to the web providing device;
Judgment for respectively determining whether the total number of session information respectively stored in the plurality of storage means has reached each upper limit number of sessions that can be processed by the web providing device in each session management time of each storage means Steps,
A control step of transferring or not transferring the capture request to the web providing device based on a result of the determination;
A load distribution method comprising: In the determination step,
If any of the plurality of storage means has reached the upper limit number of sessions, determine whether the session information in the capture request is in the storage means with the longest session management time,
In the control step,
4. The load distribution method according to claim 3, wherein if the storage means is not present, non-transfer result information is transmitted to the terminal device without transferring the capture request.   A load distribution program that causes a computer to execute the load distribution method according to claim 3.

Images