JP2007328417A - Request limiting device, server device, request limiting method and request limiting program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a request limiting method for selectively limiting a specific request pattern whose load reduction effect is determined to be the highest according to the load state of a server. <P>SOLUTION: A Web server 2 is provided with a load information acquiring means 22 for acquiring load information about a plurality of load items; a request classification means 23 for classifying requests transmitted from a client device 3 to the Web server into a plurality of request patterns, and for acquiring a load index value at each request pattern, an influence index value calculation means 24 for calculating an influence index value showing the level of an influence given by a request to load information at each request pattern and each load item based on the load index value; and a request limiting means 25 for determining a limited request pattern as a request limiting object based on the influence index value, and for limiting a portion or the whole part of the request belonging to the limited request pattern. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a server request restriction technique, and more particularly to a technique for selectively restricting a specific request pattern that is considered to be most effective in reducing the load on the server.

In a web server system that provides a service to an unspecified number of users, there are many cases in which a significant load fluctuation occurs due to a large number of accesses due to the service provision mechanism. Patent Document 1 describes a technique for restricting access from a terminal when a load on a server increases and a service cannot be normally provided.
The access control device described in this document does not transmit a request from a terminal to a server when a load measuring unit that measures the load on the server and the server load measured by the load measuring unit exceeds a specified value. In addition, when a predetermined result is transmitted to the terminal and does not exceed, a reception processing means for transmitting a request to the server is provided, and this configuration does not exponentially deteriorate processing efficiency due to concentration of access to the server. , Always keep the server throughput close to the maximum processing capacity.

JP 2002-259234 A

  As shown in FIG. 9, the load that a request from a client places on the server varies depending on the request pattern, such as an increase in CPU (Central Processing Unit) utilization rate and an increase in I / O load. For example, the “A pattern” shown in FIG. 9 has a significant influence on the I / O load, and the “C pattern” has a significant influence on the CPU load. The “B pattern” has a small influence on the above three types of loads.

The reason why the load on the server fluctuates is that the access to all the contents does not fluctuate evenly, but it is mostly caused by concentrated requests from specific contents or concentrated requests from specific addresses. In other words, the server load becomes excessive when requests of a specific pattern are concentrated.
In such cases, the conventional traffic control mechanism limits requests uniformly regardless of the server load status (I / O load, CPU load, number of simultaneous connections, etc.). It was.

FIG. 10 and FIG. 11 are diagrams schematically showing such a limiting method. Now, as shown in FIG. 11, it is assumed that the CPU load is excessive, but the I / O load and the number of simultaneous connections have a margin as shown in FIG. Even in this case, according to the conventional method, as shown in FIG. 10, the requests for the A pattern and the B pattern that do not greatly affect the CPU load are limited.
For this reason, a specific depleted resource has become a bottleneck, and requests that mainly occupy non-depleted resources are limited, making it impossible to effectively use the processing capacity of the server.

  Therefore, an object of the present invention is to provide a request limiting method for selectively limiting a specific request pattern considered to have the highest load reduction effect according to the load state of a server.

  The access restriction device of the present invention classifies requests transmitted from a client device to a server device into a plurality of request patterns, and obtains load information of the server device for a plurality of load items. Request classification means for acquiring a load index value indicating the processing result by the server device of the request belonging to the request pattern, and the degree of influence of the request on the load information for each request pattern and for each load item based on the load index value An influence index value calculation means for calculating an influence index value, a restriction request pattern to be subject to request restriction based on the influence index value, and a restriction on a part or all of the requests belonging to the restriction request pattern Request limiting means. Motomeko 6).

According to the above access restriction device, the request restriction means determines the restriction request pattern based on the load influence value calculated for each request pattern and for each load information, and restricts the request included in the pattern.
Therefore, it is possible to restrict only requests that have a large influence on the load information that is excessive, and to effectively use the resources of the server device.

In the request restriction device, the request classification unit may perform classification based on a network address of the client device or information specifying content to be accessed by the client device included in the request (claim 2).
In this way, when a large number of requests are transmitted from a specific client device, or when access concentrates on specific content provided by the server device, an access group to be restricted is appropriately selected. be able to.

In the request limiting device, the load item includes a CPU load indicating the CPU usage rate of the server device, an I / O load indicating the usage rate of the input / output interface of the server device, and a logical connection between the server device and the client device. It is also possible to include at least one of the number of simultaneous connections indicating the number of connections (claim 3).
In this way, load information can be collected for each hardware resource included in the server device, and the hardware load that is the bottleneck of processing can be reduced.

In the above request limiting device, the load index value includes the number of accesses indicating the number of received requests, the average response time that is the average time required to receive a response after receiving the request, and the request. Average number of bytes transferred, which is the average amount of information transferred between the client device and the server device for processing, and the average value, which is the average execution time of the CGI program executed to process the request At least one of the execution times may be included (claim 4).
In this way, it is possible to appropriately calculate the influence index value according to the processing capacity of the request server device, the amount of requests, and their properties.

In the request limiting device, the influence index value calculating means may calculate the influence index value based on a numerical value obtained by multiplying the load index value by a coefficient determined for each load item (claim 5).
In this way, it is possible to appropriately determine the coefficient according to the hardware of the server device and the characteristics of the program operating on the server device, and to perform the most effective request restriction for each server device.

In the request limiting device, when there is a load information that exceeds a predetermined threshold, the request limiting unit sets the request group having the maximum influence index value corresponding to the load information as a limiting pattern. (Claim 6).
In this way, the most effective request restriction can be performed in order to reduce the excessive load information.

  The server device according to the present invention classifies requests into a plurality of request patterns, and obtains processing results of requests belonging to the request pattern for each request pattern. A request classifying unit that obtains a load index value to indicate, an impact index value calculating unit that calculates an impact index value indicating a degree of influence of a request on load information for each request pattern and for each load item based on the load index value; Request restriction means for determining a restriction request pattern subject to request restriction based on the influence index value and restricting a part or all of the requests belonging to the restriction request pattern is provided.

According to the server device, the request restriction unit determines the restriction request pattern based on the load influence value calculated for each request pattern and for each load information, and restricts the request included in the pattern.
Therefore, it is possible to restrict only requests that have a large influence on the load information that is excessive, and to effectively use the resources of the server device.

  The request restriction method according to the present invention includes a load information acquisition step for acquiring load information of a server device for a plurality of load items, and classifies requests transmitted from a client device to a server device into a plurality of request patterns. A request classification step for obtaining a load index value indicating the processing result of the server device of the request belonging to the request pattern, and a degree of influence of the request on the load information for each request pattern and for each load item based on the load index value An impact index value calculation step for calculating an impact index value, a limit request pattern that is subject to request restriction based on the impact index value, and a restriction on a part or all of the requests that belong to the limit request pattern Request restriction by request restriction process Cormorant (claim 8).

According to the request restriction method, in the request restriction step, the restriction request pattern is determined based on the load influence value calculated for each request pattern and for each load information, and restriction is performed on the requests included in the pattern.
Therefore, it is possible to restrict only requests that have a large influence on the load information that is excessive, and to effectively use the resources of the server device.

In the request restriction method, in the request classification step, classification may be performed on the basis of the network address of the client device or information specifying content to be accessed by the client device included in the request.
In this way, when a large number of requests are transmitted from a specific client device, or when access concentrates on specific content provided by the server device, an access group to be restricted is appropriately selected. be able to.

In the request limiting method, the load item includes a CPU load indicating the CPU usage rate of the server device, an I / O load indicating the usage rate of the input / output interface of the server device, and a logical connection between the server device and the client device. It is also possible to include at least one of the number of simultaneous connections indicating the number of claims.
In this way, load information can be collected for each hardware resource included in the server device, and the hardware load that is the bottleneck of processing can be reduced.

In the above request limiting method, the load index value includes the number of accesses indicating the number of received requests, the average response time that is the average time required to receive a response after receiving the request, and the request. The average number of bytes transferred, which is the average value of the amount of information transferred between the client device and the server device for processing, and the average value, which is the average value of the execution time of the CGI program executed to process the request At least one of the execution times may be included (claim 11).
In this way, it is possible to appropriately calculate the influence index value according to the processing capacity of the request server device, the amount of requests, and their properties.

In the request limiting method, in the influence index value calculating step, the influence index value may be calculated based on a numerical value obtained by multiplying the load index value by a coefficient determined for each load item (claim 12).
In this way, it is possible to appropriately determine the coefficient according to the hardware of the server device and the characteristics of the program operating on the server device, and to perform the most effective request restriction for each server device.

In the request restriction method, in the request restriction process, when there is a load information that exceeds a predetermined threshold, the request group having the maximum influence index value corresponding to the load information is set as a restriction pattern. (Claim 13).
In this way, the most effective request restriction can be performed in order to reduce the excessive load information.

  The request restriction program of the present invention classifies a request transmitted from a client device to a server device into a plurality of request patterns, a load information acquisition process for acquiring load information of the server device for a plurality of load items, and requests Request classification processing for obtaining the load index value indicating the processing result by the server device of the request belonging to the request pattern for each pattern, and the influence of the request on the load information for each request pattern and for each load item based on the load index value An impact index value calculation process that calculates an impact index value that indicates the degree, and a limit request pattern that is subject to request restriction is determined based on the impact index value, and for some or all of the requests that belong to this limit request pattern Request limit processing to limit and Make a request limit by executing (claims 14 to 19).

According to the request restriction program, in the request restriction process, the restriction request pattern is determined based on the load influence value calculated for each request pattern and for each load information, and restriction is performed on the requests included in the pattern.
Therefore, it is possible to restrict only requests that have a large influence on the load information that is excessive, and to effectively use the resources of the server device.

In the request restriction program, the request classification process may be classified based on a network address of the client device or information specifying content to be accessed by the client device included in the request (claim 15).
In this way, when a large number of requests are transmitted from a specific client device, or when access concentrates on specific content provided by the server device, an access group to be restricted is appropriately selected. be able to.

In the request restriction program, the load item includes a CPU load indicating the CPU usage rate of the server device, an I / O load indicating the usage rate of the input / output interface of the server device, and a logical connection between the server device and the client device. It is also possible to include at least one of the number of simultaneous connections indicating the number of connections (claim 16).
In this way, load information can be collected for each hardware resource included in the server device, and the hardware load that is the bottleneck of processing can be reduced.

In the above request restriction program, the load index value includes the number of accesses indicating the number of received requests, the average response time that is the average time required to receive a response after receiving the request, and the request. The average number of bytes transferred, which is the average value of the amount of information transferred between the client device and the server device for processing, and the average value, which is the average value of the execution time of the CGI program executed to process the request At least one of the execution times may be included (claim 17).
In this way, it is possible to appropriately calculate the influence index value according to the processing capacity of the request server device, the amount of requests, and their properties.

In the request restriction program, the influence index value calculation process may calculate the influence index value based on a numerical value obtained by multiplying the load index value by a coefficient determined for each load item (claim 18).
In this way, it is possible to appropriately determine the coefficient according to the hardware of the server device and the characteristics of the program operating on the server device, and to perform the most effective request restriction for each server device.

In the above request restriction program, when there is a request restriction process that exceeds a predetermined threshold in the load information, the request group having the maximum influence index value corresponding to the load information may be used as the restriction pattern. Good (claim 19).
In this way, the most effective request restriction can be performed in order to reduce the excessive load information.

According to the present invention, a restriction request pattern is determined based on a load influence value calculated for each request pattern and for each load information, and restriction is performed on requests included in the pattern.
Therefore, it is possible to restrict only requests that have a large influence on the load information that is excessive, and to effectively use the resources of the server device.

Hereinafter, the configuration and operation of the traffic restriction system 1 according to an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing a configuration of the request restriction system 1.
The request restriction system 1 includes a web server 2 and a plurality of clients 3a, 3b, 3c, etc. that receive services by connecting to the web server 2 via the Internet 4 (hereinafter, unless there is a particular need to distinguish them) Generically called “client 3”).

  The web server 2 (an example of a server device) is a server computer that has a communication function and executes an HTTP (HyperText Transfer Protocol) server program. The HTML (HyperText Markup Language) stored in an internal storage device or an external storage device A content such as a format is transmitted in response to a request from the client 3.

  A client 3 (an example of a client device) is a terminal device that has a communication function such as a personal computer, a mobile phone, and a smartphone and executes a web browser program, and transmits a request to the web server 2 in accordance with an instruction input by the user. The response is received, and the content included in the response is displayed on the display device.

The web server 2 collects information (load index value) for estimating the degree of influence on each load item in advance for each request pattern, so as to reduce the load on items exceeding the threshold. Limit only valid request patterns.
FIG. 2 is a diagram schematically showing such a function of the web server 2. Assume that CPU loads, I / O (input / output) loads, and the number of simultaneous connections are considered as load items, and requests from the client 3 are classified into three patterns of A, B, and C according to contents to access.
Now, assuming that the load of the web server 2 is as shown in the graph 100, the request is limited because the CPU load is larger than the predetermined threshold value 102. Since it is known from the load index value that the request of the C pattern has a large influence on the CPU load, the requests of the A pattern and the B pattern are not particularly limited, and responses are returned for all requests (reference numeral 104). ). Access restriction is performed by not returning a response to a part of the C pattern access (reference numeral 103) (reference numeral 105). As a result of such a restriction, the load status is as shown in the graph 101, and the CPU load is within the threshold value.

FIG. 3A is a block diagram illustrating a hardware configuration of the web server 2.
The web server 2 includes a CPU 11, for example, a main storage device 12 constituted by a RAM (Random Access Memory), a hard disk device 13 and a communication interface 14 as external storage devices, and these are connected to a bus 19.
The CPU 11 has control and calculation functions, and executes programs loaded in the main storage device 12 to control each hardware of the web server 2.
The hard disk device 13 stores a program to be executed by the CPU 11 and is loaded into the main storage device 12 when an execution instruction is given or when the power is turned on. The hard disk device 13 also stores content to be transmitted in response to a request from the client 3.
The communication interface 14 is configured by a LAN (Local Area Network) adapter, for example, and performs data transmission / reception with the client 3 under the control of the CPU 11.

FIG. 3B is a functional block diagram of the web server 2.
The web server 2 collects and calculates various information for request restriction, executes a request restriction, a request restriction unit 21, a web server unit 29 that provides a service by returning a response to a request from the client 3, And a storage unit 30 that stores various types of information referred to by the request restriction unit 21.

The request restriction unit 21 includes a load information acquisition unit 22, a request classification unit 23, an influence index value calculation unit 24, and a request restriction unit 25.
The load information acquisition unit 22 periodically acquires load information indicating the magnitude of the load of the web server 2 and stores the load information as load information 31 in the storage unit 30. In the present embodiment, the load information is obtained by dividing the load information into three load items: CPU load, I / O load, and the number of simultaneous connections.

The request classification means 23 classifies the request into a plurality of request patterns based on the address of the client 3 that transmitted the request or information for specifying the content of the request destination, for example, a URI (Uniform Resource Identifier), and the load index value for each request pattern. As a result, the following values for the past fixed time are acquired and held.
1) Number of accesses in the past fixed time (hereinafter referred to as “N”)
2) Average response time in the past fixed time (hereinafter referred to as “RSP”)
3) Average number of bytes transferred in the past certain time (hereinafter referred to as “TXB”)
4) If the content is CGI (Common Gateway Interface), the average execution time of the CGI (hereinafter referred to as “EXT”)
These pieces of information are updated and held each time a request is processed. When there is a request with an unprecedented pattern, each of the above information regarding the pattern is newly added to the load information 31. If there is no request for a certain pattern for a certain period, the data related to that pattern is discarded.

The influence index value calculation unit 24 calculates an influence index value for the load for each request pattern. Specifically, the load index value for each request pattern is converted into an influence index value indicating the degree of influence on the load of the web server 2 for each load characteristic.
The influence index value of each request pattern for each load characteristic item can be generally expressed by the following formula.
Impact index value = N (F (e) EXT + F (r) RSP + F (t) TXB + F (p))
Here, F is a coefficient indicating the following.
F (e) is a coefficient indicating the impact of CPU time on the load
F (r) is a coefficient indicating the degree of influence of response time on the load
F (t) is a coefficient indicating the degree of influence of the number of bytes transferred to the load
F (p) is a coefficient indicating the correlation with the number of requests for the load. These coefficients depend on the type of content provided by the web server 2 and the characteristics of the OS (Operating System) running on the web server 2, so use it. An appropriate value can be set according to the environment. For example, a coefficient is determined as follows and converted into an influence index value.
CPU load impact index value = N (0.7 EXT + 0.1RSP + 0.1 TXB + 0.1 F (p))
I / O load influence index value = N (0.1 EXT + 0.1RSP + 0.7 TXB + 0.1 F (p))
Impact index value of the number of simultaneous connections = N (0.1 EXT + 0.4RSP + 0.1 TXB + 0.4
F (p))
The influence index value calculation unit 24 generates a pattern list 35 that is a list in which request patterns are arranged in the order of the influence index value for each load item calculated as described above, and requests having the greatest influence on each load item. Identify the pattern.

The request restriction unit 25 sets a threshold value for each load item, and when the load information acquired by the load information acquisition unit 22 exceeds the threshold value, the request restriction unit 25 sets the upper pattern of the pattern list created by the influence index value calculation unit 24. Limit processing to a request at a certain ratio (limit ratio).
The limit ratio is determined for the request pattern, for example, by the following method.
Increase the ratio to 2%, 4%, 8%, 16%, etc. and the factorial percentage of 2 up to 100% for the first (= maximum influence) pattern in the pattern list for load items that exceed the threshold Implement processing restrictions. If the load information falls below the threshold due to the addition of a restriction, the restriction ratio is sequentially reduced (rule 1).
As a result of the restriction, a request that is not normally processed is also reflected in the load index value for each request pattern according to the cost of the restricted process. For this reason, if the restriction is limited to some extent, the request pattern having the second highest degree of influence becomes the head of the list and is newly restricted (rule 2).
When the load information does not fall below the threshold even if the limit ratio is 64%, the load is limited to 100%. However, the request pattern that reaches the 100% limit is 98% if the load information falls below the threshold and the limit ratio is reduced. Reduce the limit ratio to% and 92% by 2 factorial percentages (Rule 3).
The initial limit ratio when the load item exceeds the threshold value is increased to 4%, 8%, 16%, and 32% according to the degree of excess from the threshold value of the load item (Rule 4).
When there are a plurality of load items exceeding the threshold, the above-mentioned algorithm is used to limit the load item having the maximum threshold excess (Rule 5).
When the load item value falls below the threshold, the limit ratio is reduced to 64% → 32% → 16% → 8% → 4% → 2% at regular intervals regardless of the size of the load information. . However, the case where rule 3 is applied is excluded (rule 6).

FIG. 7 is a flowchart showing determination of the limit ratio according to the above rule.
When any load information is equal to or greater than the threshold value and request restriction is started, the first restriction ratio is determined according to rule 4 (S301).
If the load information does not fall below the threshold even when the restriction is made (No in S302), the restriction ratio is increased according to the rules 1 and 3. That is, if the current limit ratio is 64%, the limit ratio is set to 100% (Yes in S303, S304), and if not, the limit ratio is increased by one rank (No in S303, S305). .
As a result of the restriction, if the load information falls below the threshold value (Yes in S302), the restriction ratio is reduced according to rule 1, rule 3, and rule 6.
That is, when the current limit ratio is 100% (Yes in S306), the current limit ratio is decreased according to the value of the load information (S307). For example, when the threshold value is 80%, if the load information is 70% or more, the rank is lowered by one, and if the load information is 50% or more and less than 70%, the rank is lowered by two.
If the current limit ratio is not 100% (No in S306), the limit ratio is lowered by one rank every fixed time regardless of the value of the load information.
The limiting ratio gradually decreases, and when it reaches 0% (when the determination in S308 or S310 is YES), the limiting process ends.

As specific contents of the process restriction for the request, for example, the following three kinds of restriction methods are determined in advance, and the request restriction unit 25 selects a process to be executed according to the load situation and instructs the web server unit 20. To do.
1) Return of error message This process returns a message to the client 3 that sent the request that the web server 2 is overloaded and cannot respond to the request.
2) Return of redirect message This process returns to the client 3 of the request source a message indicating that the processing has been transferred to another web server because the load on the web server 2 is excessive.
3) Interrupt the process and disconnect the connection This process interrupts the process and disconnects the TCP connection without returning a message to the client 3 that sent the request.

  Each of the above means is realized by the CPU 11 executing a request restriction program to control each hardware of the web server 2.

Next, an example of the content of information stored in the storage unit 30 is shown.
FIG. 4A is an example of load information. The load information 31 is information indicating the magnitude of the load of the web server 2 for each load item acquired by the load information acquisition unit 22. In this example, three items of CPU load, I / O load, and the number of simultaneous connections are stored.
The CPU load is information indicating the usage rate of the CPU 11 in FIG. 3A, and is information indicating the percentage of time during which the CPU 11 was performing some processing other than the idle process.
The I / O load is information indicating a ratio of time that an I / O port (not shown) of the web server 2 is used for input / output processing between the CPU 11 and peripheral devices such as the communication interface 14.
The number of simultaneous connections is the number of logical communication connections that the web server 2 maintains with the client 3, for example, the number of active TCP connections with the client 3.
These pieces of information are updated every information acquisition interval (for example, 1 minute) of the load information acquisition unit 22, and only the latest information is stored.

FIG. 4B is an example of the load index value. In this example, the access count, average response time, average transfer byte count, and average execution time acquired by the request classification means 23 are stored for each request pattern.
The number of accesses is information indicating the number of requests belonging to each access pattern during a predetermined period (for example, the past one minute based on the data update time).
The average response time is information indicating the time required from when the web server 2 receives a request until it starts to return a response to the client 3.
The average transfer byte number is information indicating the amount of data transmitted and received between the web server 2 and the client 3 as a request and a response.
The average execution time is information indicating the execution time of the CGI executed for request processing.
As for the average response time, the average number of transfer bytes, and the average execution time, the average values during the predetermined period are stored.

  FIG. 4C is an example of the influence index value. The influence index value 33 includes an influence index value calculated by the influence index value calculating unit 24 based on the load index value 32 for each access pattern for each load item, for example, CPU load, I / O load, and the number of simultaneous connections. It is remembered.

  FIG. 4D is an example of a pattern list. The pattern list 35 stores the order based on the magnitude of the influence index value of the request pattern for each load item. For example, the first column of the figure shows that the CPU load influence index value is that the pattern C is the first largest, the pattern A is the second, and the pattern B is the third.

  The threshold value 35 stores a threshold value predetermined for each load item. For example, the threshold is set to 70% for the CPU load, 60% for the I / O load, and 50 for the number of simultaneous connections.

Next, the operation of the web server 2 will be described.
FIG. 5 is a diagram for explaining the flow of processing when the load information 31 is within the threshold value for any load item.
The load information acquisition unit 22 periodically acquires load information (S101). This process is executed independently of other processes.
When the web server unit 29 receives a request from the client 3 (S102), if no load item exceeds the threshold in the load information acquired in S101, the web server unit 29 performs normal request processing (S103), and sends a response to the client. 3 is returned (S104).
The request classification unit 24 collects the load index value of the request from the request processing result, and updates the load index value 32 stored in the storage unit 30. The influence index value calculation means 24 recalculates and updates the influence index value 32 and the pattern list 35 in preparation for processing when the threshold value is exceeded (S105).

FIG. 6 is a diagram for explaining the flow of processing when the load information is greater than or equal to the threshold value.
The load information acquisition unit 22 periodically acquires load information (S201).
When there is a load item that exceeds the threshold, the request restriction unit 25 refers to the pattern list 35 for the load item having the maximum excess and identifies the request pattern having the maximum impact (S202).
The request restriction unit 25 calculates a restriction execution ratio from the degree of excess load and the restriction processing continuity state (S203).
When a request from the client 3 is received after the above processing is completed (S204), the request restriction unit 25 compares the received request with the request pattern specified in S202, and if the request matches, the restriction execution calculated in S203 is performed. Request restriction processing is performed according to the ratio. The web server unit 29 performs processing such as returning a response and returning an error message according to the specified restriction (S206).
The request classification unit 23 collects the load information of the request from the request processing result, and stores it in the load index value 32. Further, the influence index value calculation unit 24 recalculates and updates the influence index value 33 and the request pattern 35 in preparation for the next request processing (S207).

In the above description, as shown in FIG. 2, the web server 2 is implemented with all functions as a web server and functions for restricting requests. However, each function is distributed and implemented in a plurality of computers. You can also
FIG. 8 shows an example of such an embodiment.
The request restriction system 1a includes a client 3, a request restriction device 40, and a web server 2a.
The web server 2 a includes the web server unit 29 of FIG. 2B and processes a request from the client 3. In other words, it is an ordinary web server.
The request restriction device 40 installed between the web server 2a and the client 3 includes the request restriction unit 21 and the storage unit 30 in FIG.

According to the traffic restriction system 1, the request classification unit 23 classifies the request transmitted from the client 3 into a request pattern, and calculates a load index value. The influence index value calculation unit 24 calculates an influence index value 33 and generates a pattern list 35. The request restriction unit 25 refers to the pattern list 35 when any load item is equal to or greater than the threshold value, and restricts access to a request corresponding to the request pattern at the head of the list. In other words, only requests having an access pattern that is most effective in recovering the insufficient resource are limited by the load characteristics of the server.
Therefore, it is possible to provide a service that efficiently uses limited resources of the web server 2.

It is a block diagram of the request restriction system which is one Embodiment of this invention. It is a figure explaining the principle of the request restriction | limiting in the web server of FIG. FIG. 3A is a diagram illustrating a hardware configuration of the web server. FIG. 3B is a functional block diagram of the web server 2. 4A to 4D are diagrams illustrating data stored in the storage unit of FIG. It is a figure explaining operation | movement of a web server when load information is less than a threshold value. It is a figure explaining operation | movement of a web server when load information is more than a threshold value. It is a figure explaining the determination method of a limiting ratio. It is a block diagram of the modification of a request restriction | limiting system. It is a figure explaining the conventional request limiting method. It is a figure explaining the conventional request limiting method. It is a figure explaining the conventional request limiting method.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Request restriction system 2 Web server 3 Client 21 Request restriction part 22 Load information acquisition means 23 Request classification means 24 Influence index value calculation means 25 Request restriction means 29 Web server part 30 Storage part 31 Load information 32 Load index value 33 Influence index value 34 Threshold 35 Pattern list

Claims (19)

Load information acquisition means for acquiring the load information of the server device for a plurality of load items;
Request classification means for classifying a request transmitted from a client device to the server device into a plurality of request patterns and obtaining a load index value indicating a processing result by the server device of the request belonging to the request pattern for each request pattern When,
An influence index value calculating means for calculating an influence index value indicating a degree of influence of the request on the load information for each request pattern and for each load item based on the load index value;
A request restriction device comprising: a request restriction unit that determines a restriction request pattern that is a request restriction target based on the influence index value, and that restricts a part or all of the requests belonging to the restriction request pattern.   2. The request according to claim 1, wherein the request classification unit performs the classification on the basis of a network address of the client device or information specifying content to be accessed by the client device included in the request. Restriction device.   The load item includes a CPU load indicating a CPU usage rate of the server device, an I / O load indicating a usage rate of an input / output interface of the server device, and a logical connection between the server device and the client device. The request limiting device according to claim 1, wherein the request limiting device includes at least one of the number of simultaneous connections indicating the number.   The load index value is the number of accesses indicating the number of received requests, the average response time that is the average time taken from receiving the request to transmitting a response to the client device, and processing the request. The average number of bytes transferred, which is the average value of the amount of information transferred between the client device and the server device, and the average value of the execution time of the CGI program executed to process the request 4. The request limiting device according to claim 1, wherein at least one of the average execution times is included.   The said influence index value calculation means calculates the said influence index value based on the numerical value which multiplied the coefficient defined for every said load item to the said load index value, The said any one of Claim 1 thru | or 4 characterized by the above-mentioned. The request restriction device according to one.   The request restriction means, when there is a piece of load information that exceeds a predetermined threshold, sets the request group having the maximum influence index value corresponding to the load information as the restriction pattern. The request restriction device according to any one of claims 1 to 5. In a server device that receives a request from a client device and transmits a response to the request to the client device,
Load information acquisition means for acquiring load information of the server device for a plurality of load items;
Request classification means for classifying the request into a plurality of request patterns and obtaining a load index value indicating a processing result of the request belonging to the request pattern for each request pattern;
An influence index value calculating means for calculating an influence index value indicating a degree of influence of the request on the load information for each request pattern and for each load item based on the load index value;
Request restriction means for determining a restriction request pattern subject to request restriction based on the influence index value, and restricting a part or all of the requests belonging to the restriction request pattern;
A server device comprising: A load information acquisition step of acquiring load information of the server device for a plurality of load items;
Request classification step of classifying requests transmitted from the client device to the server device into a plurality of request patterns, and obtaining a load index value indicating a processing result by the server device of the request belonging to the request pattern for each request pattern When,
An impact index value calculating step for calculating an impact index value indicating a degree of influence of the request on the load information for each request pattern and for each load item based on the load index value;
A request restriction method comprising: a request restriction step of determining a restriction request pattern subject to request restriction based on the influence index value, and restricting a part or all of the requests belonging to the restriction request pattern.   9. The request according to claim 8, wherein, in the request classification step, the classification is performed based on a network address of the client device or information specifying content to be accessed by the client device included in the request. Restriction method.   The load item includes a CPU load indicating a CPU usage rate of the server device, an I / O load indicating a usage rate of an input / output interface of the server device, and a logical connection between the server device and the client device. The request limiting method according to claim 8 or 9, comprising at least one of the number of simultaneous connections indicating the number.   The load index value is the number of accesses indicating the number of received requests, the average response time that is the average time taken from receiving the request to transmitting a response to the client device, and processing the request. The average number of bytes transferred, which is the average value of the amount of information transferred between the client device and the server device, and the average value of the execution time of the CGI program executed to process the request 11. The request limiting method according to claim 8, further comprising at least one of a certain average execution time.   12. The influence index value calculating step, wherein the influence index value is calculated based on a numerical value obtained by multiplying the load index value by a coefficient determined for each load item. The request restriction method described in one.   In the request restriction step, when there is a piece of load information that exceeds a predetermined threshold, the request group having the maximum influence index value corresponding to the load information is set as the restriction pattern. The request restriction method according to any one of claims 8 to 12. On the computer,
Load information acquisition processing for acquiring server device load information for a plurality of load items;
Request classification processing for classifying requests transmitted from a client device to the server device into a plurality of request patterns, and obtaining a load index value indicating a processing result by the server device of the request belonging to the request pattern for each request pattern When,
An influence index value calculation process for calculating an influence index value indicating a degree of influence of the request on the load information for each request pattern and for each load item based on the load index value;
A request limiting process for determining a request restriction target based on the influence index value and performing a request restriction process for restricting a part or all of the requests belonging to the restriction request pattern is executed. Request restriction program.   15. The request classification process according to claim 14, wherein the classification is performed based on a network address of the client device or information that is included in the request and specifies content to be accessed by the client device. Request restriction program as described.   The load item includes a CPU load indicating a CPU usage rate of the server device, an I / O load indicating a usage rate of an input / output interface of the server device, and a logical connection between the server device and the client device. 16. The request restriction program according to claim 14 or 15, comprising at least one of the number of simultaneous connections indicating the number.   The load index value is the number of accesses indicating the number of received requests, the average response time that is the average time taken from receiving the request to transmitting a response to the client device, and processing the request. The average number of bytes transferred, which is the average value of the amount of information transferred between the client device and the server device, and the average value of the execution time of the CGI program executed to process the request The request restriction program according to any one of claims 14 to 16, comprising at least one of a certain average execution time.   18. The influence index value calculation process calculates the influence index value based on a numerical value obtained by multiplying the load index value by a coefficient determined for each load item. The request restriction program described in any one of.   In the request restriction process, when there is a piece of load information that exceeds a predetermined threshold, the request group having the maximum influence index value corresponding to the load information is used as the restriction pattern. The request restriction program according to any one of claims 14 to 18, characterized by:

Images