JP2007200349A - Server-client system, load distribution device, load distribution method, and load distribution program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To efficiently execute a process. <P>SOLUTION: The load distribution device 101a comprises a process information receiving part 401 receiving information for the process from a client 102, a determination part 402 determining, based on the received information for the process, a server which is to process the process from among a plurality of process processing servers 101b-101n, and a server information transmitting part 403 transmitting information for the determined server to the client 102. Further, the client 102 includes a server information receiving part 412 receiving the transmitted information for the server, and a process request transmitting part 413 transmitting information for a process processing request to the process processing server 101b-101n related to the received information. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a server / client system, a load distribution device, a load distribution method, and a load distribution program for distributing access to servers, and in particular, by monitoring and evaluating a server operating state, it is optimal among a plurality of servers. The present invention relates to a server / client system, a load distribution apparatus, a load distribution method, and a load distribution program for selecting a secure server.

  2. Description of the Related Art Conventionally, a server / client type system in which an application program on a client requests a server to execute a server program via a network has been widely used. In this system, in response to a program execution request from a client, a server that provides a service that is a program execution result starts a process that is a processing unit of the program for providing the service. When a process such as a Web content distribution program, business software, or game software is started on the server, services such as text data and image data are provided to the requested client.

  When performing such server / client system processing using multiple servers, there is a technology that distributes the load of request processing sent from clients in order to prevent some servers from being overloaded. . For example, in the case of a Web server that distributes Web content, a round robin method in which processing is distributed in order to a server that provides multiple access from clients, or a least connection that selects a server with the smallest number of sessions from a plurality of servers A server for assigning a process is selected by a method or the like.

  When load distribution is performed using the round robin method or the least connection method, the consumption of resources consumed by processes to be executed (for example, resources provided in personal computers such as CPU and memory), and the amount of resources provided in the server Servers are assigned regardless of

  FIG. 19 is an explanatory diagram illustrating an example of the relationship between the resource amount of the server and the resource amount consumed by the process. In FIG. 19, the CPU consumption of the server for executing the server program is the X coordinate, and the memory consumption is the Y coordinate. In FIG. 19, Xmax on the X coordinate represents 100%, which is the maximum value of CPU consumption, and Ymax on the Y coordinate represents 100%, which is the maximum value of memory consumption. A coordinate 1902 indicates an ideal position where 100% of the CPU and the memory are consumed. When the process 1901 that consumes the CPU and the memory is executed on this server, the memory consumption becomes 100% even though the CPU consumption is lower than 100%.

  FIG. 20 is an explanatory diagram showing another example of the relationship between the resource amount of the server and the resource amount consumed by the process. In FIG. 20, similarly to FIG. 19, the CPU consumption of the server for executing the server program is the X coordinate, and the memory consumption is the Y coordinate. Xmax on the X coordinate represents 100%, which is the maximum value of CPU consumption, and Ymax on the Y coordinate represents 100%, which is the maximum value of memory consumption. A coordinate 2002 indicates an ideal position where the CPU and the memory are 100% consumed. When the process 2001 that consumes the CPU and the memory is executed on this server, the consumption of the CPU becomes 100% although the consumption of the memory is lower than 100%, contrary to FIG.

  In addition, the load balancer grasps the number of sessions of each of the plurality of servers, and selects an appropriate server from the plurality of servers by the number of sessions and the weight determined in advance by the machine performance of each server. There is a system for assigning (see, for example, Patent Document 1 below).

JP 2002-269061 A

  However, in the above-described load distribution using the round robin method or the least connection method, the load is distributed depending on the number of sessions with the client and the like regardless of the resources provided by the server and the resources consumed by the process. As a result, when the process requested to execute is started on the server, it is determined whether there is an excess or shortage of resources, so there is a problem that a delay occurs in the response to the request and a waiting time occurs for the client. It was.

  On the other hand, because the amount of resources that each server can have varies depending on the performance of the device used as the server, when using a method that distributes load only by the number of sessions, such as the round robin method or the least connection method In addition, there is a problem that a process is allocated to a server with a small remaining amount of resources or a server with a bias in resource consumption, without considering the resource usage after load distribution.

  Also, if a process is executed on a server with a biased consumption of either CPU or memory, only the resource area with the higher consumption is available even though the resource with the lower consumption has free space Due to the shortage, it becomes impossible to execute a new process. Such a server with uneven resource consumption has a problem in that the number of processes that can be executed is smaller than a server in which resources are evenly consumed.

  In order to solve the above-described problems caused by the conventional technology, the present invention is optimally selected from a plurality of servers deployed in one or a plurality of locations by numerically evaluating the resources and operating states of the servers to which the process execution is allocated. It is an object of the present invention to provide a server client system, a load distribution device, a load distribution method, and a load distribution program capable of selecting a server and efficiently executing a process on each server.

  In order to solve the above problems, a server / client system according to the present invention is configured such that a plurality of servers and a plurality of clients are connected via a network, and the server performs processing based on a process request from the client. In the server-client system for transmitting a processing result to the client, at least one of the servers receives process information receiving means from the client via the network, and process information receiving means Based on the information on the process received by the information receiving means, a determination means for determining a server that processes the process from the plurality of servers, and transmits information on the server determined by the determination means to the client. Sir An information transmission unit, wherein the client receives information about the server transmitted by the server information transmission unit via the network, and the information received by the server information reception unit Process request transmission means for transmitting information related to the process processing request to the server.

  In the load distribution apparatus according to the present invention, a plurality of servers and a plurality of clients are connected via a network, the server performs processing based on a process request from the clients, and transmits a processing result to the clients. A load distribution device for distributing a load on the server in a server / client system, the process information receiving unit receiving information about the process from the client via the network, and the process information receiving unit A determination unit that determines a server that processes the process from the plurality of servers based on information about the process; and a server information transmission unit that transmits information about the server determined by the determination unit to the client. It is characterized by having

  Further, according to the load distribution method of the present invention, a plurality of servers and a plurality of clients are connected via a network, the server performs processing based on a process request from the client, and transmits a processing result to the client. A load distribution method for distributing the load on the server in a server / client system, the process information receiving step receiving information about the process from the client via the network, and the process information receiving step received by the process information receiving step A determination step for determining a server to process the process from the plurality of servers based on information on the process, and a process request transmission step for transmitting information on the process processing request to the server determined by the determination step And To.

  Also, according to the load distribution program of the present invention, a plurality of servers and a plurality of clients are connected via a network, the server performs processing based on a process request from the client, and transmits a processing result to the client. A load distribution program for distributing the load on the server in a server / client system, the process information receiving step for receiving information on the process from the client via the network, and the process information receiving step A determination step of determining a server for processing the process from the plurality of servers based on information on the process is executed by the server.

  Here, the determining means (process) is the resource consumption obtained by adding the resource consumption of the process to the point representing the current resource consumption of each server in the space centered on the parameter of the resource in each server. A first distance calculating means (step) for calculating a first distance between an origin and a straight line connecting the maximum usable capacity of the parameter, and a space on the basis of the parameter of the resource in each server The second distance calculating means for calculating the second distance from the origin of the expected point of the resource consumption obtained by adding the resource consumption of the process to the point representing the current resource consumption of each server (Step), and includes (including) a process for processing the process based on one or both values of the first distance and the second distance. It is also possible to determine the server.

  The parameter may include at least one of the CPU load of the server, the system memory load, the graphic processing unit load, the video memory load, and the network interface card load. Good.

  According to the present invention, an optimal server is selected from a plurality of servers deployed at one or a plurality of locations by numerically evaluating the resources and operating states of the servers to which the process execution is allocated, and a process is performed on each server. There is an effect that it is possible to efficiently execute.

  Exemplary embodiments of a server / client system, a load distribution apparatus, a load distribution method, and a load distribution program according to the present invention will be explained below in detail with reference to the accompanying drawings.

(System configuration)
First, a system configuration of a server / client system including a load distribution apparatus according to the embodiment of the present invention will be described. FIG. 1 is a schematic diagram showing a system configuration of a server / client system according to this embodiment of the present invention. In FIG. 1, servers 101a to 101n are configured by connecting to a terminal device (client) 102 or a telephone station 105 via a network 100 such as the Internet. A radio base station 104 is connected to the telephone station 105, and the mobile phone 103 is connected to the servers 101 a to 101 n via the radio base station 104.

  The servers 101a to 101n are personal computers used for server purposes called PC servers, for example, and are managed and operated by a service provider that distributes Web contents and network games. Then, by installing dedicated application software on the PC server, a load distribution device having a load distribution function is obtained. One or a plurality of servers having a load distribution function can exist in the entire system. Here, the server 101a will be described as a load balancer.

  Therefore, the servers (process processing servers) 101b to 101n other than the server 101a can also be a load distribution apparatus having a load distribution function by installing dedicated application software in the same manner, and are activated as the load distribution apparatus 101a. When the active server 101a is stopped due to a failure, one of the servers 101b to 101n is switched as the load balancer 101a and can act as a proxy.

  When there is a process execution request from a terminal device 102 or a mobile phone 103 used by a user, which will be described later, the load distribution apparatus 101a allocates the execution of the process to the servers 101b to 101n. The servers 101b to 101n execute the requested process and provide a service as a result of the execution to the terminal device 102 and the mobile phone 103 to be described later via the network 100.

  The terminal device 102 is an information terminal device such as a personal computer, and is used by individuals and companies. The mobile phone 103 can be connected to at least the Internet, such as “i-mode (registered trademark)” of NTT Docomo or “ezweb (registered trademark)” of au, and can be connected to a browser or Java (registered). This is a mobile phone having an information communication function for accessing the network 100 using a trademark application. The radio base station 104 transmits the communication data received from the telephone station 105 as a radio wave to the mobile phone 103, and further transmits the radio wave received from the mobile phone 103 to the telephone station 105. The telephone station 105 performs line switching when data communication is performed among the mobile phone 103, the load balancer 101a, and the servers 101b to 101n.

  The system for the mobile phone 103 is a notebook PC or PDA with a wireless LAN adapter mounted / built-in instead of the mobile phone, and a wireless LAN base station instead of the mobile phone wireless base station 104. This is a mode in which the present invention operates effectively.

  Further, in communication via the network 100 performed between the load balancer 101a, the servers 101b to 101n, and the terminal device 102 or the mobile phone 103, a security function using an SSL method or the like, an encryption technique, etc. Use it to ensure confidentiality.

(Hardware configuration)
Next, the hardware configuration of the terminal device and the server according to this embodiment of the present invention will be described. FIG. 2 is a block diagram showing a hardware configuration of the load distribution device, the terminal device, and the server of the server / client system according to the embodiment of the present invention.

  In FIG. 2, a terminal device 102, a load balancer 101a, and servers 101b to 101n are a CPU 201, a ROM 202, a RAM 203, an HDD (hard disk drive) 204, an HD (hard disk) 205, and an FDD (flexible disk drive). 206, FD (flexible disk) 207 as an example of a removable recording medium, display 208, network I / F (interface) 209, keyboard 210, mouse 211, printer 212, and CD-ROM drive 213, a CD-ROM 214 as an example of a removable recording medium, and a speaker 215. Each component is connected by a bus 200.

  Here, the CPU 201 governs overall control of each of the terminal device 102, the load distribution device 101a, and the servers 101b to 101n. The ROM 202 stores programs such as a basic input / output program and a boot program. The RAM 203 is used as a work area for the CPU 201. The HDD 204 controls data read / write with respect to the HD 205 according to the control of the CPU 201. The HD 205 stores data written under the control of the HDD 204.

  The FDD 206 controls reading / writing of data with respect to the FD 207 according to the control of the CPU 201. The FD 207 stores data written under the control of the FDD 206. As a removable recording medium, in addition to the FD 207, a CD-RW, MO, DVD (Digital Versatile Disk), or the like may be used. The display 208 displays a cursor (menu) or a window (browser) relating to data such as documents, images, and function information. For example, CRT, TFT liquid crystal display, plasma display and the like.

  The network I / F 209 is connected to the network 100 and is connected to the terminal device 102 via the network 100. The network I / F 209 serves as an interface between the network 100 and the inside, and controls input / output of data from the load balancer 101a, the servers 101b to 101n, and the terminal device 102. The network I / F 209 is, for example, a modem or a LAN adapter.

  The keyboard 210 includes keys for inputting characters, numerical values, various instructions, and the like, and performs data input. The mouse 211 performs cursor movement, range selection, window movement, size change, and the like. A track ball, a joystick, a game pad, or the like may be used as long as it has a similar function as a pointing device. The printer 212 prints document data. For example, a laser printer or an ink jet printer. The CD-ROM drive 213 controls reading of data from the CD-ROM 214 according to the control of the CPU 201. The CD-ROM 214 is a detachable recording medium. A speaker 215 (including headphones and earphones) outputs voice, music, and the like.

  Next, a hardware configuration of the mobile phone according to the embodiment of the present invention will be described. FIG. 3 is a block diagram showing a hardware configuration of the cellular phone of the server client system according to the embodiment of the present invention. 3, the mobile phone 103 includes a CPU 301, a ROM 302, a RAM 303, a display 304, an operation key 305, a microphone 306, a speaker 307, a communication control unit 308, an antenna 309, and an external connection terminal 310. And an external storage device 311, and the antenna 309 is connected to the radio base station 104 for communication. Each component is connected by a bus 300.

  Here, the CPU 301 controls the entire mobile phone 103. The ROM 302 stores programs such as a basic input / output program and a boot program. The RAM 303 is used as a work area for the CPU 301. A display 304 is a liquid crystal display and displays a window (browser) related to data such as documents, images, and function information. The operation key 305 inputs characters, numbers, various instructions, and the like. The microphone 306 converts the input sound into an electrical signal. The speaker 307 converts the input electrical signal into sound and outputs it. The communication control unit 308 transmits and receives radio waves with the radio base station 104 via the antenna 309 and performs control. The external connection terminal 310 serves as a connection port with an external storage device 311 such as a flash memory.

(Functional configuration)
Next, a functional configuration of the server / client system according to the embodiment of the present invention will be described. FIG. 4 is a block diagram showing a functional configuration of the server client system according to the embodiment of the present invention. In FIG. 4, the load distribution apparatus 101 a constituting the server / client system includes a process information receiving unit 401, a determining unit 402, and a server information transmitting unit 403, and the determining unit 402 further calculates a distance. A part 404 is included. The client 102 constituting the server / client system includes a process information transmission unit 411, a server information reception unit 412, and a process request transmission unit 413.

  The process information receiving unit 401 receives information about a process from the client 102 via the network 100. Further, the determination unit 402 determines a server to process a process from among the plurality of process processing servers 101b to 101n based on the information regarding the process received by the process information reception unit 401. Here, the distance calculation unit 404 calculates the distance of the resource consumption of the process between the origin and a straight line connecting the maximum usable capacity of the parameter in a space with the resource parameter as an axis. Then, based on the distance calculated by the distance calculation unit 404, a server for processing is determined. In addition, the server information transmission unit 403 transmits information regarding the server determined by the determination unit 402 to the client 102.

  On the other hand, the process information transmission unit 411 transmits information on the process to the load distribution apparatus 101a via the network 100 before making a process request. Further, the server information receiving unit 412 receives information about the server transmitted by the server information transmitting unit 403 of the load distribution apparatus 101a via the network 100. In addition, the process request transmission unit 413 transmits information to the server received by the server information reception unit 412, that is, the server determined by the determination unit 402 of the load distribution apparatus 101a (any one of the process processing servers 101b to 101n). Send information about process processing requests. Then, the server performs processing based on this process request, and transmits the processing result to the client 102.

(Load balancing procedure)
Next, an outline of the processing procedure of the load distribution method using the load distribution apparatus 101a according to this embodiment of the present invention will be described. FIG. 5 is an explanatory diagram showing an outline of the processing procedure of the load distribution method using the load distribution apparatus according to this embodiment of the present invention. In FIG. 5, the load distribution apparatus 101 a distributes the load of execution of a process requested by the terminal apparatuses 102 a to 102 n (for example, a game process such as an online game) and allocates servers 101 b to 101 n.

  When playing an online game using the terminal device 102a, the terminal device 102a requests the load distribution device 101a to execute a game process indicated by an arrow 501. The load distribution apparatus 101a requested to execute the process evaluates a game server for executing the game process from the servers 101b to 101n. The game server is evaluated based on the resources provided in each of the servers 101b to 101n. As a result of the evaluation by the load distribution apparatus 101a, the server 101b is determined as the load distribution destination as indicated by an arrow 502. When the game server is determined, a game process execution is requested from the load balancer 101a, and the game process is executed in the server 101b. Thereby, the use of the game can be started from the terminal device 102a.

  As described above, the load distribution apparatus 101a selects an optimum one for executing this game process from the servers 101b to 101n, triggered by a request for execution of the game process from the terminal apparatuses 102a to 102n. Further, not only the online game as described above, but also when executing a process by other application software, the load can be distributed using the load distribution device 101a.

  Next, the relationship between the resource amount of the server and the resource amount consumed by the process according to this embodiment of the present invention will be described. 6 and 7 are explanatory diagrams showing an example of ideal process allocation according to the embodiment of the present invention. In FIG. 6, the CPU consumption of the servers 101b to 101n is shown as an X coordinate, and the memory consumption is shown as a Y coordinate. This graph represents a consumption trend of resources used by processes executed on the servers 101b to 101n.

  The characteristics of the graph shown in FIG. 6 will be described below assuming that the server 101b is provided. In this graph, the amount corresponding to 100%, which is the maximum value of CPU consumption, is represented by a numerical value of 20. This is a numerical value representing the ratio of the CPU consumption of the server 101c shown in FIG. 7 described later (the CPU consumption of the server 101c is 10), and the server 101b has twice the CPU of the server 101c. It represents having the performance that can be consumed.

  Next, an amount corresponding to 100%, which is the maximum value of memory consumption, is represented by a numerical value of 10. As in the case of the CPU, this is a numerical value representing the ratio of the memory consumption of the server 101c shown in FIG. 7 described later (in FIG. 7, the memory consumption is 20). This means that it has a performance capable of consuming 1/2 times as much memory as 101c.

  A process 601 shown in the graph of FIG. 6 represents the CPU and memory consumption of a process executed on the server 101b. The CPU and memory consumed by executing this process 601 are 10 CPUs and 5 memory, respectively, and occupy half of the resources of the server 101b.

  The state in which the first process 601 is activated can be expressed as a rectangle from the origin to the position 10 of CPU consumption and the position of 5 memory consumption, and the maximum values of CPU and memory consumed. Is represented as coordinates 602 (hereinafter, the coordinates consisting of the maximum CPU and memory consumed by one process are referred to as “consumed resource coordinates”). Furthermore, when the second process 601 is executed, the coordinates can be expressed in a step shape with the coordinates 602 of the first process 601 as a starting point. By assigning the second process 601, the CPU and memory consumption become the maximum values of the respective resources of the server 101 b, and coordinates indicating the maximum values are represented as coordinates 603.

  As described above, when two processes 601 are executed, the CPU and memory consumption in the server 101b are both maximized, and the remaining resources for allocating other processes are eliminated in both the CPU and the memory, so that the resources are ideally consumed. It will be in the state.

  In FIG. 7, the CPU consumption of the servers 101b to 101n is shown as an X coordinate, and the memory consumption is shown as a Y coordinate. This graph represents a consumption trend of resources used by processes executed by the servers 101b to 101n.

  The characteristics of the graph shown in FIG. 7 will be described below assuming that the server 101c is provided. In this graph, an amount corresponding to 100%, which is the maximum value of CPU consumption, is represented by a numerical value of 10. This represents ½ times the amount of CPU that can be consumed by the server 101b of FIG. 6 described above, and represents that the CPU 101 has a performance that can consume ½ times the amount of CPU consumed by the server 101b.

  Next, an amount corresponding to 100%, which is the maximum value of the memory consumption, is represented by a numerical value of 20. As in the case of the CPU, this represents twice the amount of memory that can be consumed by the server 101b of FIG. 6 described above, and has the capability of consuming twice as much memory as the server 101b. To express.

  A process 701 shown in the graph of FIG. 7 represents the CPU and memory consumption of a process executed on the server 101c. The CPU and memory consumed by executing this process 701 are 5 for CPU and 10 for memory, respectively, and occupy half of the resources of the server 101c.

  The state in which the first process 701 is activated can be represented as a rectangle from the origin to the position of 5 of CPU consumption and the position of 10 of memory consumption, and the maximum values of CPU and memory consumed. Is represented as coordinates 702. Furthermore, when the second process 701 is executed, the coordinates can be represented in a step shape with the coordinates 702 of the first process 701 as a starting point. By assigning the second process 701, the CPU and memory consumption become the maximum values of the respective resources of the server 101c, and coordinates indicating the maximum values are represented as coordinates 703.

  As described above, when two processes 701 are executed, the CPU and memory consumption in the server 101c are both maximized, and the remaining resources for allocating other processes are eliminated in both the CPU and the memory, and the resources are ideally consumed. It will be in the state.

  Next, optimal resource consumption in the server according to the embodiment of the present invention will be described. 8 to 10 are explanatory diagrams showing an example of optimum resource consumption in the server of the server / client system according to the present embodiment. In FIG. 8, the CPU consumption is shown as an X coordinate, and the memory consumption is shown as a Y coordinate. 8 and FIGS. 9 and 10 to be described later indicate that when there are servers 101d to 101f, the amount of resources that can be consumed by each server is different.

  First, the characteristics of the graph of FIG. 8 will be described assuming that the server 101d is provided. In this graph, an amount corresponding to 100%, which is the maximum value of CPU consumption, is represented by a maximum value 801. This represents 80% of the maximum value of CPU consumption in FIG. 9 described later, and further represents 50% of the maximum value of CPU consumption in FIG. An amount corresponding to 100%, which is the maximum value of memory consumption, is represented as a maximum value 802. This represents the same amount as the maximum value of the memory consumption of FIG. 9 described later, and further represents 200%, which is twice the maximum value of the memory consumption of FIG.

  Then, as shown in the graph of FIG. 8, the process 803 that is started first is described from the position of the origin, and the processes that are started following the process 803 are described in a time-series manner from processes 804 to 808. Eventually, a coordinate 809 that is the intersection of the maximum value 801 of CPU consumption and the maximum value 802 of memory consumption is reached. Thus, the server 101d is in a state where the CPU and the memory are ideally consumed.

  In FIG. 9, the graph shows the CPU consumption as the X coordinate and the memory consumption as the Y coordinate. FIG. 9 shows that when there are servers 101d to 101f as in FIG. 8, there is a difference in the amount of resources that can be consumed by each server.

  The characteristics of the graph shown in FIG. 9 will be described below assuming that the server 101e is provided. In this graph, an amount corresponding to 100%, which is the maximum value of CPU consumption, is represented as a maximum value 901. This represents 125% of the maximum value of CPU consumption in FIG. 8 described above, and further represents 62.5% of the maximum value of CPU consumption in FIG. 10 described later.

  Next, an amount corresponding to 100%, which is the maximum value of memory consumption, is represented as a maximum value 902. This represents the same amount as the maximum value of the memory consumption of FIG. 8 described above, and is twice the maximum value of the memory consumption of FIG. 10 described later.

  Then, as shown in the graph of FIG. 9, the process 903 that is started first is described from the position of the origin, and the processes that are started following the process 903 are described in a time-series manner from processes 904 to 908. Eventually, a coordinate 909 that is the intersection of the maximum CPU consumption value 901 and the maximum memory consumption value 902 is reached. Thus, the server 101e is in a state where the CPU and the memory are ideally consumed.

  In FIG. 10, the CPU consumption is shown as an X coordinate, and the memory consumption is shown as a Y coordinate. FIG. 10 shows that when there are servers 101d to 101f as in FIGS. 8 and 9, there is a difference in the amount of resources that can be consumed by each server.

  The characteristics of the graph shown in FIG. 10 will be described below assuming that the server 101f is provided. In this graph, an amount corresponding to 100%, which is the maximum value of CPU consumption, is represented as a maximum value 1001. This represents 200% of the maximum value of CPU consumption in FIG. 8 described above, and further represents 160% of the maximum value of CPU consumption in FIG. An amount corresponding to 100%, which is the maximum value of the memory consumption, is shown as a maximum value 1002, and this maximum value 1002 is 50% of the maximum value of the memory consumption of FIGS. 8 and 9 described above.

  Then, as shown in the graph of FIG. 10, the process 1003 that is started first is described from the position of the origin, and the processes that are started following the process 1003 are described in a time series from process 1004 to 1008. Eventually, a coordinate 1009 that is the intersection of the maximum CPU consumption value 1001 and the maximum memory consumption value 1002 is reached. As a result, the server 101f is in a state where the CPU and the memory are ideally consumed.

  Next, an example of server resource consumption according to this embodiment of the present invention will be described. FIG. 11 is an explanatory diagram showing an example of resource consumption of the server according to this embodiment of the present invention. In FIG. 11, Xmax on the X coordinate indicates 100%, which is the maximum value of CPU consumption, and Ymax on the Y coordinate indicates 100%, which is the maximum value of memory consumption.

  A broken line 1101 allocates CPU and memory resources for each process, and finally reaches the coordinate 1102 of the intersection of the maximum value of CPU consumption Xmax and the maximum value of memory consumption Ymax. An example is shown. As a result, the broken line 1101 indicates a graph in the case where the process is allocated with no waste in CPU and memory consumption, and represents the result of ideal consumption of resources.

  A broken line 1103 shows an example in which the CPU consumption reaches the maximum value Ymax before the CPU consumption reaches the maximum value Xmax as a result of the allocation of CPU and memory resources for each process occurrence. Yes. This indicates that the broken line 1103 is a graph when the CPU consumption is wasted relative to the memory consumption, and that there is a usable CPU remaining in the amount indicated by the section 1105. Yes.

  A polygonal line 1106 shows an example in which the CPU consumption reaches the maximum value Xmax before the memory consumption reaches the maximum value Ymax as a result of allocating CPU and memory resources for each process occurrence. Yes. This indicates that the broken line 1106 is a graph when the memory consumption is wasted relative to the CPU consumption, and that the usable memory remains in the amount indicated by the section 1108. Yes.

  In the conventional process of assigning processes by the load balancer, load distribution is performed without determining the resource consumption of the server, so that the resources are not consumed evenly as indicated by the above-mentioned broken lines 1103 and 1106. It is normal. On the other hand, if a process is allocated such that the CPU and memory as resources are evenly consumed, it is possible to prevent useless resources that are not consumed, such as the broken line 1101, from remaining.

(Server evaluation method)
Next, a server evaluation method by the load distribution apparatus according to this embodiment of the present invention will be described. FIG. 12 is an explanatory diagram showing an outline of a server evaluation method by the load distribution apparatus according to the embodiment of the present invention. In FIG. 12, graphs 1202 and 1203 indicate the CPU consumption as an X coordinate and the memory consumption as a Y coordinate. Hereinafter, the graph 1202 shows the content of resource consumption of the server 101b, and the graph 1203 shows the content of resource consumption of the server 101c.

  First, a graph 1202 shows the characteristics of a server having resources whose maximum CPU consumption is twice and whose maximum memory consumption is 1/2 times that of the graph 1203. Processes 1204 and 1205 are already assigned to the graph 1202 and executed. On the other hand, a graph 1203 shows the characteristics of a server provided with a resource whose maximum CPU consumption is ½ times and whose maximum memory consumption is twice that of the graph 1202. Processes 1210 and 1211 are already assigned to the graph 1203 and executed. The following describes a method for determining whether resources are optimally consumed when a load is distributed to a server having any of the characteristics of the graph 1202 and the graph 1203 when the execution request of the process 1201 occurs.

  When the process 1201 is executed in the graph 1203, the coordinates (consumed resource coordinates) representing the entire consumed resource become coordinates 1215 by adding the resource consumption for the process 1201 from the vertex 1211 a of the process 1211. Then, a straight line connecting the origin and the coordinates 1213 of the intersection with the maximum value of CPU and memory consumption is the resource consumption optimum line 1212 (hereinafter, the coordinates of the intersection of the maximum value of CPU and memory consumption and the origin A perpendicular line 1214 is drawn from the coordinate 1215 to the resource consumption optimal line 1212.

  On the other hand, when the process 1201 is executed in the graph 1202, the resource consumption amount for the process 1201 is added from the vertex 1205 a of the process 1205, so that coordinates (consumed resource coordinates) representing the entire consumed resource are coordinates 1209. . A straight line connecting the coordinates 1207 and the origin of the intersection with the maximum values of CPU and memory consumption is defined as a resource consumption optimum line 1206, and a perpendicular line 1208 is drawn from the coordinates 1209 to the resource consumption optimum line 1206.

  Then, comparing the lengths of the vertical line 1208 and the vertical line 1214, it is shown that the length of the vertical line 1208 is shorter. This means that when the process 1201 is executed by the server 101b rather than being executed by the server 101c, the balance between the memory consumption and the CPU consumption, which are resources of the server 101b, is evenly close. Show. Therefore, the server 101c having the characteristics of the graph 1203 is not selected as the server for executing the process 1201, and the server 101b having the characteristics of the graph 1202 is selected. When the process 1201 is executed in the server 101b, resources are optimally consumed.

  At this time, there may be a case where the lengths of the vertical lines 1208 and 1214 are equal and it can be determined that the resource is consumed more appropriately in the same way when either server is selected. In such a case, the distances 1216 and 1217 between the consumption resource coordinates 1209 and 1215 and the origin are compared, and a server having a shorter length is assigned. By doing so, it is possible to assign a server with a small resource consumption amount when there is a server having a similar distance to the resource consumption optimum line. In addition, when both the resource consumption amount and the distance from the resource consumption optimum line are to be considered at the same time, the area (1208 × 1216, 1214 × 1217) multiplied by the distance (1216, 1217) between the vertical line (1208, 1214) and the origin is used. ) Can be realized by selecting a server having a smaller value.

  Next, functions of the allocation server evaluation method according to the embodiment of the present invention will be described. FIG. 13 is an explanatory diagram showing an example of a function of the allocation server evaluation method according to the embodiment of the present invention. In FIG. 13, Xmax on the X coordinate represents 100%, which is the maximum value of CPU consumption, and Ymax on the Y coordinate represents 100%, which is the maximum value of memory consumption. Then, in the space with the resource parameter as an axis, the maximum capacity that can be used for the origin and the parameter, that is, the coordinate of the intersection of Xmax and Ymax is (xmax, ymax), and (xmax, ymax) and the origin (0, 0). ) Is a resource consumption optimum line 1301. The characteristics of the graph shown in FIG. 13 will be described below assuming that the server 101b is provided.

  First, the process 1302 being executed first is shown as a rectangle from the origin where the memory consumption and the CPU consumption consumed by the process 1302 are the lengths of the sides. This is a rectangle representing the resources consumed by each process. The diagonal vertices from the origin of the rectangle represented by this process 1302 are shown as coordinates 1302a.

  The second process 1303 being executed can be shown as a rectangle with the coordinates 1302a as the starting point and the memory consumption and CPU consumption consumed by the process 1303 as the length of each side. The diagonal vertex from the coordinate 1302a which is the starting point of this rectangle is shown as a coordinate 1303a.

If the third process to be newly assigned is the process 1304, the process is shown as a rectangle having a coordinate 1303a as a starting point, a CPU usage amount cx, and a memory usage amount cy as shown in FIG. The vertex indicating the diagonal of the coordinate 1303a which is the start point of the rectangle is indicated as a coordinate 1304a.
Next, the distance of the perpendicular 1309 to be drawn from the coordinate 1304a to the resource consumption optimum line 1301 is obtained. First, the values of the x and y positions shown on the graph of FIG. 13 are expressed by the following formulas (1) and (2).

y = fxy (x) = (ymax / xmax) × x (1)
x = fyx (y) = (xmax / ymax) × y (2)

  As described above, the consumed resource coordinates consumed by the first process 1302 and the second process 1303 are indicated by the coordinates 1303a. The consumption resource coordinates consumed by the third process are indicated as coordinates 1304a. The coordinates of this coordinate 1304a are expressed as (x1, y1) = (x0 + cx, y0 + cy).

  On the other hand, an intersection of a straight line passing through the coordinate 1304a parallel to the vertical axis and the resource consumption optimal line 1301 is defined as a coordinate 1305. Further, an intersection of a straight line passing through the coordinate 1304 a parallel to the horizontal axis and the resource consumption optimal line 1301 is defined as a coordinate 1306. A straight line connecting the coordinate 1304a and the coordinate 1305 is a straight line 1307, and the length of the straight line 1307 is given by the following equation (3).

  Δy = | fxy (x1) −y1 | (3)

  A straight line connecting the coordinates 1304a and 1306 is a straight line 1308, and the length of the straight line 1308 is given by the following equation (4).

  Δx = | fyx (y1) −x1 | (4)

  A perpendicular drawn from the coordinate 1304a to the resource consumption optimum line 1301 is a perpendicular 1309, and the length of the perpendicular 1309 is given by the following equation (5).

  disxy = ΔxΔy / √ (Δx2 + Δy2) (5)

  This disxy is the distance from the resource consumption optimum line 1301 at the coordinate 1304a, and is a value to be evaluated when the load distribution apparatus 101a determines a server for executing a process from among a plurality of servers.

  Next, an expression for obtaining the distance from the resource consumption coordinate to the resource consumption optimum line in the case of an n-dimensional function including other parameters besides the CPU and memory described above will be described. First, a resource consumption optimum line that can be obtained on an n-order Euclidean space with n (xi) (1 ≦ i ≦ n) parameters used as resources is expressed by the following equation (6) using a two-dimensional case. (1 ≦ i ≦ n).

xi + 1 = fxi x (i + 1) mod n (xi) = (x (i + 1) mod n max / xi max) × xi
... (6)

  When the current server resource consumption is (xi0) (1 ≦ i ≦ n) and the process resource usage is (ci) (1 ≦ i ≦ n),

  xi1 = xi0 + ci (7)

  At this time, Equation (8) is

Δx (i + 1) mod n = | fxi x (i + 1) mod n (xi 1) −x (i + 1) mod n 1 |
... (8)

(1 ≦ i ≦ n), and the distance between the point (xi1) (1 ≦ i ≦ n) on the n-dimensional Euclidean space and the resource consumption optimal line 1301 can be expressed as the following equation (9).

  Therefore, the load balancer 101a calculates the distances (dis) indicated by the respective equations (5) or (9) of the servers 101b to 101n to be load-balanced, and the server with the shortest distance is calculated. Assign process execution.

  In addition, the distance 1310 between the point (xi1) (1 ≦ i ≦ n) and the origin on the n-dimensional Euclidean space can be expressed as in Expression (10).

  Therefore, when one server is selected from the servers having the same vertical line length dis, the server having the smaller diag value is selected and the process execution process is assigned. Furthermore, when it is desired to simultaneously evaluate the length dis of the perpendicular line and the distance diag between the origin, processing is assigned to the server having the smallest value of dis × diag.

  The above-mentioned resources other than the CPU and memory are, for example, a GPU, a video memory, a network interface card, and the like. The GPU is a graphic chip for performing rendering and geometry calculation by hardware in a game using 3D graphics. The video memory is a memory provided in the GPU used when the GPU performs image processing calculations. The network interface card is an Ethernet (registered trademark) adapter, for example, and terminates 100Base, 1000Base-TX, and the like.

  Next, the contents of server evaluation and determination processing by the load balancer according to the embodiment of the present invention will be described. FIG. 14 is a flowchart showing an example of a server evaluation and determination processing procedure by the load distribution apparatus according to the embodiment of the present invention. In the flowchart of FIG. 14, first, parameters used when the load distribution apparatus 101a evaluates the servers 101b to 101n are set (step S1401). The parameters are, for example, CPU consumption, memory consumption, GPU consumption, video memory consumption of the GPU, bandwidth consumption of the network interface card, and the like.

  Next, the usage amount of the resource used by each of the plurality of processes is defined and set for each parameter (step S1402). Then, for each of the servers 101b to 101n, the maximum resource value that is the parameter defined in step S1402 is set (step S1403).

  Next, the resource consumption of each parameter set in step S1401 is acquired and monitored for each of the servers 101b to 101n (step S1404). It is determined whether or not there is a process execution request from the terminal device 102 or the mobile phone 103 which is a client (step S1405). If there is no request for process execution (step S1405: No), the process returns to step S1404, and the consumption of resources consumed for each of the servers 101b to 101n is monitored (step S1404), and the process is continued.

  On the other hand, if there is a process execution request from the terminal device 102 that is a client (step S1405: Yes), server candidates for executing this process are selected from the servers 101b to 101n (step S1406). ). In this case, a server whose usage amount does not exceed the maximum value for any of the parameters set in step S1401 is selected from the servers 101b to 101n. Then, evaluation is performed by obtaining the distance of the perpendicular line 1309 drawn down from the resource consumption coordinate 1304a of FIG. 13 to the resource consumption optimum line 1301 from the candidates selected in step S1406, and the distance is the shortest. For example, the server 101b is determined as a load distribution destination (step S1407).

  Further, in the selection of server candidates in S1406, as described above, in addition to the length of the perpendicular 1309 drawn from the resource consumption coordinate 1304a to the resource consumption optimum line 1301, the distance 1310 between the resource consumption coordinate 1304a and the origin is evaluated together. Can do. When there are servers having the same length of the perpendicular line 1309, a server with a smaller resource consumption can be selected by selecting a server having a small distance 1310 from the origin. Alternatively, by selecting the server having the smallest value obtained by multiplying the length of the perpendicular line 1309 and the distance 1310 between the origin, it is possible to realize optimal resource consumption and to assign a process to a server having a small resource consumption.

  The load distribution apparatus 101a issues a process execution request to the load distribution destination server 101b determined in step S1407 (step S1408). Then, in order to indicate that the process has been executed to the terminal device 102 which is a client, notification of an IP address and a service port number which are information for accessing the load distribution destination server 101b is performed (step S1409). . The terminal device 102 notified of the information for accessing the server 101b can directly access the server 101b, and ends the process.

  Next, the contents of the server evaluation and decision processing procedures by the load balancer according to the embodiment of the present invention will be described. FIG. 15 is a schematic diagram showing an example of a server evaluation and determination processing procedure by the load distribution apparatus according to the embodiment of the present invention. 15 shows that the load distribution apparatus 101a distributes the load of the processes requested from the terminal apparatuses 102a to 102n to the servers 101b to 101n. In FIG. 15, the load distribution apparatus 101a performs processing according to the following procedure.

(1) Set parameters used for evaluation when performing load distribution.
(2) Define and set the usage for each parameter consumed by the processes executed on each server. This is done every time a new process is added. For example, when four parameters of CPU, memory, GPU, and video memory are used, if the process to be executed is a 3D game, resources are consumed for the GPU and video memory in addition to the CPU and memory. If the process to be executed is a 2D game, GPU and video memory resources are not consumed. In this way, the resources consumed vary depending on the contents of the process.

  (3) The maximum value of each parameter provided for each server is set. This setting may be set manually by the administrator or the like with respect to the load distribution apparatus 101a, or each server may be sequentially monitored via the network 100 to obtain a setting value. In addition, since various servers 101b to 101n such as games and other application servers may be installed depending on the purpose of use, the maximum capacity can be set for each parameter. This process is indicated by an arrow 1501a.

  (4) With respect to the resources set as parameters, the amount consumed for each of the servers 101b to 101n is monitored. This process is indicated by an arrow 1501a, and monitoring is sequentially performed from the load balancer 101a to the servers 101b to 101n to acquire the consumption.

  (5) A process execution request is received from the terminal device 102a, which is a client. At the same time, the amount of each resource used by the accepted process is acquired. This process is indicated by an arrow 1502.

  (6) The load distribution apparatus 101a selects server candidates for executing a process from the servers 101b to 101n. It is checked whether the consumption of resources used by the process that has accepted the execution request exceeds the maximum value of the resources of each server, and those that do not exceed are selected as candidates.

  (7) The load balancer 101a further selects the distance of the perpendicular line 1309 or the consumption resource coordinates 1304a from the candidates selected in (6) to the resource consumption optimum line 1301 from the consumption resource coordinates 1304a of FIG. Evaluation is performed by obtaining the distance 1310 from the origin. If the length of the perpendicular is the shortest or the length of the perpendicular is the same, the distance 1310 from the origin is short, or the length of the perpendicular and the distance from the origin are multiplied. For example, the server 101b is determined as the load distribution destination.

  (8) The load distribution apparatus 101a issues a process execution request to the load distribution destination server 101b determined in step S1407. At this time, if the server 101b accepts only an execution request from an authenticated client, the terminal device 102a also transmits an ID, an IP address, a service port number, an encryption key, a token, and the like to the server 101b. This process is indicated by an arrow 1501a.

  (9) In order to indicate that the process has been executed to the terminal device 102a, which is a client, the information of the load distribution destination server 101b is notified. This information is necessary when the terminal device 102a accesses the server 101b. For example, when the terminal device 102a directly accesses the server 101b, it indicates the IP address, port number, authentication / encryption key or token of the server 101b. This process is indicated by an arrow 1502.

  (10) The terminal device 102a notified of the information for accessing the server 101b can access the server 101b. Based on the information notified in (9), the terminal device 102a can access the server 101b. On the server 101b side, whether the access is from an authenticated client or not is confirmed based on whether the IP address, ID, token, etc. match. If the encryption key is delivered, the communication is encrypted. This process is indicated by an arrow 1503.

(Content of process information)
Next, process information held by the load balancer according to the embodiment of the present invention will be described. FIG. 16 is an explanatory diagram showing an example of a table holding process information according to the embodiment of the present invention. In FIG. 16, a process information table 1600 stores data contents for holding process information, an ID for uniquely identifying a process, a process name that is the name of the process, and a process executed by the process. Each item includes a CPU usage amount used at the time of execution and a memory usage amount used when the process is executed.

  The data attribute of the ID is a text type or an integer type if it can be expressed only by a numerical value. The data attribute of the process name is a text type, and the data attribute of the CPU usage and the memory usage is defined as an attribute such as a floating point type.

  First, the data content of the process indicated by the record whose ID is 100 is that the process name is process 1, the CPU usage is 0.15, and the memory usage is 0.2. Next, the data content of the process indicated by the record whose ID is 110 is that the process name is process 2, the CPU usage is 0.2, and the memory usage is 0.4. The same applies when the ID is 100. This is because when the process with ID 100 or 110 is executed, the CPU and memory consumption values consumed by the executed server are the CPU and memory usage values registered in the table, respectively. Indicates an increase by minutes.

  When receiving a process execution request from the terminal apparatus 102, the load distribution apparatus 101a uses the process information in the process information table 1600 to select a server for distributing the load. When the number of servers 101b to 101n to be distributed is small, it is more time-consuming to describe in a reference file such as a text file without managing the table using the database such as the table described above. May be less. In such a case, it is assumed that the table contents are created as a reference file as necessary, and the format can be arbitrarily determined.

(Table information)
Next, server information held by the load balancer according to the embodiment of the present invention will be described. FIG. 17 is an explanatory diagram showing an example of a table that holds server information of the server / client system according to the embodiment of the present invention. In FIG. 17, a server information table 1700 stores data for holding server information, and includes a connection destination server name for uniquely identifying a server, a usage rate of a CPU being used in each server, and the like. Each item includes the usage rate of the memory being used by each server.

  The data attribute of the connection destination server name is a text type or an integer type if it is expressed only by a numerical value. The data attribute of the CPU usage rate and the memory usage rate is a floating point type. Here, the CPU or memory usage rate shown in the server information table 1700 is a numerical value representing the percentage of the CPU or memory already used in each of the servers 101a, 101b, and 101c. Note that the maximum values of the CPU and memory usage rates in the servers 101b, 101c, and 101d are 1.0 (= 100%). The load distribution apparatus 101a may periodically monitor these values from each server and set them in a table, or may acquire them as necessary when performing load distribution. In the following, an evaluation is performed on which of the servers 101b, 101c, and 101d the process 1 whose ID is 100 shown in FIG.

  First, the record of the connection destination server name 101b shown in the first row of the server information table 1700 has a CPU usage rate of 0.9 and a memory usage rate of 0.88. This server 101b has a high usage rate for both the CPU and memory, and adding the CPU usage and memory usage of process 1 shown in FIG. 16 exceeds the maximum value of 1.0 (= 100%). It can be seen that there is no available resource for executing the process 1.

  The record with the connection server name 101c shown in the second row has a CPU usage rate of 0.5 and a memory usage rate of 0.2. When the CPU usage and memory usage of process 1 shown in FIG. 16 are added to this server 101c, the CPU usage rate is 0.65 and the memory usage rate is 0.4. It is less than 0.0 (= 100%), and it can be seen that there is an available resource for executing the process 1.

  The connection server name 101d shown in the third row has a CPU usage rate of 0.3 and a memory usage rate of 0.2. When the CPU usage and memory usage of the process 1 shown in FIG. 16 are added to the server 101d, the CPU usage is 0.45 and the memory usage is 0.4. It is less than 0.0 (= 100%), and it can be seen that there is an available resource for executing the process 1.

  Further, using the equation (5), when the distances dis of the perpendiculars to the resource consumption optimal line from the consumption resource coordinates of the servers 101c and 101d are respectively determined, the distance at the server 101c is about 0.177, and the server 104c The distance at is about 0.035. Thereby, the load distribution apparatus 101a selects and assigns the server 101d having a short distance as the optimum server.

  Further, the distance diag between the consumed resource coordinates and the origin is about 0.763 for the server 101c and about 0.602 for the server 101d according to the equation (10), and the product (dis × diag) of both is the value of the server 101c. In this case, it is about 0.135 in the case of the server 101d, and about 0.021 in the case of the server 101d. Even when both the length of the resource consumption optimum line and the distance between the consumed resource coordinates and the origin are considered at the same time, assign.

  If there are a small number of servers 101b, 101c, and 101d for distributing the load, the table management using the database as described above is not performed, and the file is described in a reference file such as a text file. In some cases, it is less time-consuming to update. In such a case, the contents of the table shown in FIG. 17 may be written out as a text file as necessary, and the format may be arbitrarily determined and created. Although not shown, not only the CPU and memory usage rates described above, but also items such as an IP address and a port number for accessing each server may be set as necessary.

(Contents of regional distribution of servers)
Next, load distribution of servers managed for each area according to the embodiment of the present invention will be described. FIG. 18 is an explanatory diagram showing an example of regional distribution of servers in the server / client system according to the embodiment of the present invention. In FIG. 18, a terminal device 102a, a load distribution device 1801a, and servers 1803a to 1803n installed in a center 1803 are installed in a region A, and a terminal device 102b, a load distribution device 1801b, and a center are installed in a region B. Servers 1804a to 1804n installed in 1804 are installed, and terminal C 102c, load balancer 1801c, and servers 1805a to 1805n installed in the center 1805 are installed in the store C. Thus, an example of a complex server-client system using each device distributed in each region or store is shown.

  The line A1 indicating the flow of the process execution request from the terminal device 102a installed in the area A is transmitted to the load balancer 1801a through the line A2 by the user navigation 1802. Here, the user navigation 1802 refers to a system that performs routing by determining a region from the IP address of a client or by determining each region from the address of a local DNS server. As a result, each area of the client can be specified, and a server installed in an area closest to the client (data transmission delay is small) can be preferentially assigned. The process execution request transmitted to the load distribution apparatus 1801a through line A2 is distributed to the server 1803a installed in the same area A through line A3, and the process execution request is transmitted.

  Also, line B1 indicating a process execution request from the terminal apparatus 102b installed in the region B is transmitted to the load distribution apparatus 1801b through line B2 by the user navigation 1802. Here, a process execution request is normally transmitted to any of the servers 1804a to 1804n installed in the region B. However, when the load balancer 1801a and the load balancer 1801b are configured to exchange resource monitoring information of the subordinate server group, the load is distributed to the server 1803a installed in another area A through the line B3. Will be able to. Further, if the resources of all the servers in the areas A, B, and the store C are monitored by the load balancer 1801b, a server that distributes loads from all other servers in addition to the server 1803a installed in the area A can be obtained. You will be able to choose.

  Also, line C1 indicating a process execution request from the terminal device 102c installed in the store C is transmitted to the load balancer 1801c through the line C2 by the user navigation 1802. The load balancer 1801c evaluates the servers 1805a to 1805n in the store C that can load balance. However, if there is no server having resources that can be allocated in the same store C as the terminal device 102c, the load can be distributed and allocated to the server 1804n in the adjacent region B through the line C3.

  In addition, it is known that the access frequency of many Internet applications including games etc. varies depending on the time. For example, it is assumed that an application reaches an access peak at 10 o'clock and the access frequency at, for example, 9 o'clock and 11 o'clock before and after the peak is several percent smaller than the peak time. When considering the time difference between east and west, servers deployed in the 1 hour front and back area will be used as surplus resources from clients in the peak time zone if the resources of the server in the area that has reached the peak time zone are exhausted. The load distribution in the east and west, such as receiving and processing process processing requests, becomes possible. Even on the equator, the average hourly time difference is less than 1700 km, and the round trip time due to propagation delay is about 17 ms. This is a delay that can be fully utilized even in applications such as games where communication between the server and the client is frequently performed, and it is possible to effectively distribute the load of applications that reach their peak depending on the time.

  As long as the resource capacity consumed by the requested process remains, the load distribution destination server may be installed anywhere regardless of the region. As a result, resources such as a hosting server that is generally used as a Web server that consumes less resources and multiple PC servers installed in network cafes in various locations can be partially borrowed to distribute the load. By making a contract as a server for this purpose, it becomes possible to perform load balancing with redundancy.

  As described above, according to the present embodiment, the server 101b to the server that can execute the process is evaluated by using the resource consumption at the time of execution of the process requested to be executed from the terminal device 102. 101n can be selected. In addition, each time a process to be executed is newly added or modified, the latest resource consumption is registered, so that the execution of the process can always be assigned to the optimum server. Since the server is selected using the resource consumption of the process that has requested execution and the remaining resources of the servers 101b to 101n that are constantly monitored, the amount of resources consumed by the process and the server It is possible to prevent mismatch with the remaining amount of resources.

  In addition, by using the resource consumption of the process that requested execution and the remaining amount of resources of the servers 101b to 101n that are constantly monitored, by evaluating with a function that consumes resources of the server without waste, It is possible to select an optimal server that consumes resources without waste. And it becomes possible to flexibly select from a plurality of servers that execute a game, an application program, or the like in which the amount of consumed resources differs for each process.

  In addition, it is possible to flexibly select a server having a remaining amount of resources that matches a resource consumed by a process desired to be executed from among a plurality of servers having different amounts of resources. This makes it possible to consume ideal server resources, maximize the simultaneous processing processes (such as the number of games) across multiple servers, and use up the capacity of each server's resource parameters without waste. .

  In addition, when the load balancer 101a according to the present embodiment is used, the load is distributed to a server or the like where there is a possibility that the process cannot be executed because the remaining amount of resources is insufficient or the process may be queued. Therefore, the waiting time for the process execution request can be reduced.

  Note that the load distribution method described in the present embodiment can be realized by executing a program prepared in advance on a computer such as a personal computer or a workstation. This program is recorded on a computer-readable recording medium such as a hard disk, a flexible disk, a CD-ROM, an MO, and a DVD, and is executed by being read from the recording medium by the computer. The program may be a transmission medium that can be distributed via a network such as the Internet.

  As described above, the server / client system, the load distribution device, the load distribution method, and the load distribution program according to the present invention are useful for consuming the resources of the server to which the load is distributed in a balanced manner, It is suitable for assigning processes that have been requested by many users, such as games, to multiple servers.

1 is a schematic diagram showing a system configuration of a server / client system according to an embodiment of the present invention. It is a block diagram which shows the hardware constitutions of the load distribution apparatus of the server client system concerning this Embodiment of this invention, a terminal device, and a server. It is a block diagram which shows the hardware constitutions of the mobile telephone of the server client system concerning this Embodiment of this invention. It is a block diagram which shows the functional structure of the server client system concerning this Embodiment of this invention. It is explanatory drawing which shows the outline | summary of the process sequence of the load distribution method using the load distribution apparatus concerning this Embodiment of this invention. It is explanatory drawing which shows an example of the ideal process allocation concerning this Embodiment of this invention. It is explanatory drawing which shows another example of ideal process allocation concerning this Embodiment of this invention. It is explanatory drawing showing an example of the optimal resource consumption in the server of the server client system concerning this Embodiment of this invention. It is explanatory drawing showing another example of the optimal resource consumption in the server of the server client system concerning this Embodiment of this invention. It is explanatory drawing showing another example of the optimal resource consumption in the server of the server client system concerning this Embodiment of this invention. It is explanatory drawing which shows the example of consumption of the resource concerning this Embodiment of this invention. It is explanatory drawing which shows the outline | summary of the evaluation method of the server by the load distribution apparatus concerning this Embodiment of this invention. It is explanatory drawing which shows an example of the function of the evaluation method of the allocation server concerning this Embodiment of this invention. It is a flowchart which shows an example of the process sequence of evaluation and determination of the server by the load distribution apparatus concerning this Embodiment of this invention. It is a schematic diagram which shows an example of the process sequence of evaluation and determination of the server by the load distribution apparatus concerning this Embodiment of this invention. It is explanatory drawing which shows an example of the table holding the information of the process concerning this Embodiment of this invention. It is explanatory drawing which shows an example of the table which hold | maintains the information of the server of the server client system concerning this Embodiment of this invention. It is explanatory drawing which shows an example of the regional distribution of the server of the server client system concerning this Embodiment of this invention. It is explanatory drawing which shows an example of the relationship between the resource amount with which a server is provided, and the resource amount which a process consumes. It is explanatory drawing which shows another example of the relationship between the resource amount with which a server is provided, and the resource amount which a process consumes.

Claims (12)

In a server-client system in which a plurality of servers and a plurality of clients are connected via a network, the server performs processing based on a process request from the client, and transmits a processing result to the client.
At least one of the servers is
Process information receiving means for receiving information on the process from the client via the network;
Determining means for determining a server to process the process from the plurality of servers, based on information about the process received by the process information receiving means;
Server information transmitting means for transmitting information about the server determined by the determining means to the client;
With
The client
Server information receiving means for receiving information on the server transmitted by the server information transmitting means via the network;
A process request transmitting means for transmitting information relating to the process processing request to a server related to the information received by the server information receiving means;
A server-client system comprising: The determining means includes
On the space centered on the parameter of the resource in each server, the origin and parameter of the expected point of the resource consumption obtained by adding the resource consumption of the process to the point representing the current resource consumption of each server First distance calculating means for calculating a first distance with a straight line connecting the maximum usable capacity;
In the space centered on the resource parameters of each server, the resource consumption amount of the process is added to the point representing the current resource consumption amount of each server. Second distance calculating means for calculating a distance of 2;
With
2. The server / client system according to claim 1, wherein a server that processes the process is determined based on one or both of the first distance and the second distance.   The parameter includes at least one of a load amount of a CPU of the server, a load amount of a system memory, a load amount of a graphic processing unit, a load amount of a video memory, and a load amount of a network interface card. The server client system according to claim 2. A load of the server in a server-client system in which a plurality of servers and a plurality of clients are connected via a network, the server performs processing based on a process request from the client, and transmits a processing result to the client A load balancer that distributes
Process information receiving means for receiving information on the process from the client via the network;
Determining means for determining a server to process the process from the plurality of servers, based on information about the process received by the process information receiving means;
Server information transmitting means for transmitting information about the server determined by the determining means to the client;
A load balancer comprising: The determining means includes
On the space centered on the parameter of the resource in each server, the origin and parameter of the expected point of the resource consumption obtained by adding the resource consumption of the process to the point representing the current resource consumption of each server First distance calculating means for calculating a first distance with a straight line connecting the maximum usable capacity;
In the space centered on the resource parameters of each server, the resource consumption amount of the process is added to the point representing the current resource consumption amount of each server. Second distance calculating means for calculating a distance of 2;
With
The load distribution apparatus according to claim 4, wherein a server that processes the process is determined based on one or both of the first distance and the second distance.   The parameter includes at least one of a load amount of a CPU of the server, a load amount of a system memory, a load amount of a graphic processing unit, a load amount of a video memory, and a load amount of a network interface card. The load distribution apparatus according to claim 5. A load of the server in a server-client system in which a plurality of servers and a plurality of clients are connected via a network, the server performs processing based on a process request from the client, and transmits a processing result to the client Load balancing method for distributing
A process information receiving step for receiving information on the process from the client via the network;
A determination step of determining a server to process the process from the plurality of servers based on information on the process received by the process information reception step;
A process request transmitting step of transmitting information on the process processing request to the server determined by the determining step;
A load balancing method comprising: The determination step includes
On the space centered on the parameter of the resource in each server, the origin and parameter of the expected point of the resource consumption obtained by adding the resource consumption of the process to the point representing the current resource consumption of each server A first distance calculating step of calculating a first distance with a straight line connecting the maximum usable capacity;
In the space centered on the resource parameters of each server, the resource consumption amount of the process is added to the point representing the current resource consumption amount of each server. A second distance calculating step of calculating a distance of 2;
Including
8. The load distribution method according to claim 7, wherein a server for processing the process is determined based on one or both of the first distance and the second distance.   The parameter includes at least one of a load amount of a CPU of the server, a load amount of a system memory, a load amount of a graphic processing unit, a load amount of a video memory, and a load amount of a network interface card. The load distribution method according to claim 8. A load of the server in a server-client system in which a plurality of servers and a plurality of clients are connected via a network, the server performs processing based on a process request from the client, and transmits a processing result to the client A load balancing program that distributes
A process information receiving step for receiving information on the process from the client via the network;
A determination step of determining a server to process the process from the plurality of servers based on information on the process received by the process information reception step;
Is executed by the server. The determination step includes
On the space centered on the parameter of the resource in each server, the origin and parameter of the expected point of the resource consumption obtained by adding the resource consumption of the process to the point representing the current resource consumption of each server A first distance calculating step of calculating a first distance with a straight line connecting the maximum usable capacity;
In the space centered on the resource parameters of each server, the resource consumption amount of the process is added to the point representing the current resource consumption amount of each server. A second distance calculating step of calculating a distance of 2;
Including
The load distribution program according to claim 10, wherein a server that processes the process is determined based on one or both of the first distance and the second distance.   The parameter includes at least one of a load amount of a CPU of the server, a load amount of a system memory, a load amount of a graphic processing unit, a load amount of a video memory, and a load amount of a network interface card. The load distribution program according to claim 11.

Images