JP6938944B2 - Information processing device and load balancing control method - Google Patents

Description

The present invention relates to an information processing device and a load balancing control method.

SDN (Software Defined Networking) technology has been developed for the purpose of flexibly changing network settings. The SDN includes a communication device, a communication control device that manages the communication device, and a control network that connects the communication device and the communication control device. In SDN, communication devices specialize in transfer functions. The communication device is managed by the communication control device. Hereinafter, the communication device in SDN will be referred to as a switch. Further, the communication control device in the SDN is referred to as a controller.

In SDN, network control logic such as path calculation is implemented in communication control software on the controller. For example, communication control software is software that calculates a path, notifies the controller of the path, and the controller sets the path to the switch. Hereinafter, the communication control software will be referred to as an application.

One of the applications in SDN is, for example, a path setting application that generates a control message to a switch connecting the specified two points when a user inputs a path setting request between the two points.

FIG. 24 is a diagram showing an example of the distributed communication control system P100. In SDN, when there is only one controller, the performance of the controller may not be able to keep up with the expansion of the network scale. The distributed communication control system P100 has a plurality of controllers, and the processing performance is improved by sharing the management of switches (SW in the figure) among the plurality of controllers. The controller that controls the switch is referred to as the "master" of the controlled switch. There is one master controller for one switch. Also, one controller can be the master of a plurality of switches. In FIG. 24, the controller is mounted on the server P1 and the server P2, and the controllers provided in the server P1 and the server P2 each serve as a master of three switches.

For example, since the service policy and bandwidth control requirements differ for each user's organization or area, a path setting application is prepared for each user's organization or area. That is, in the distributed communication control system P100, there are a plurality of types of path setting applications, and instances of the same type of applications operate on one or a plurality of servers. An instance is an application on a server. In the example shown in FIG. 24, an instance of one type of path setting application is running on the server P1 and the server P2.

FIG. 25 is a diagram showing an example of the flow of control messages when a path setting request is input to the path setting application in the distributed communication control system P100. In FIG. 25, the same distributed communication control system P100 as in FIG. 24 is shown. The example shown in FIG. 24 is an example in which a path setting request between the bases Ha and Hb is input to the path setting application on the server P1.

The path setting application on the server P1 calculates the path passing through the switch # 1 and the switch # 2 as the path between the bases Ha and Hb, and the path between the bases Ha and Hb addressed to the switches # 1 and the switch # 2. A control message is generated to set. The master of switch # 1 is the controller on server P1. The master of switch # 2 is the controller on server P2.

The master of switch # 1 is a controller on the same server P1 as the instance that generated the control message, and the control message to switch # 1 is transmitted from server P1 to the destination switch # 1. Hereinafter, the control message is simply referred to as a message.

Japanese Unexamined Patent Publication No. 2000-268004

However, in the distributed communication control system P100, the following problems may occur. In the distributed communication control system P100, if a message is transmitted from a plurality of controllers to one switch, the switch settings may become inconsistent. To reduce the possibility of inconsistent switch settings, the message to the switch is a specification sent from the controller of the master of the switch. The controller of the master of switch # 2 is a controller on server P2 different from server P1 of the instance that generated the message. Therefore, the message to the switch # 2 is transferred from the server P1 to the server P2 and transmitted from the server P2 to the switch # 2.

For example, the time it takes for the message to reach switch # 2 is longer than the time it takes for the message to reach switch # 1 due to the transfer of the message between the controllers. As a result, even though the path between the bases Ha and Hb has been set in the switch # 1, the switch # 2 does not have the path set between the bases Ha and Hb, and the data is transferred. May have an adverse effect on. Therefore, the transfer of messages between controllers causes deterioration in the performance of the distributed communication control system P100, so it is preferable that the number of messages is reduced.

FIG. 26 is a diagram showing an example of a distributed communication control system including a load balancer. The load balancer P3 receives requests for each application and evenly distributes the requests to each server. As a result, the processing load of the application of each server can be leveled. However, the transfer of control messages between controllers is not considered, and it is not always possible to reduce the transfer of messages between controllers by evenly distributing the requests to the application to each server.

One aspect of the present invention is to provide an information processing device capable of reducing message transfer between control devices and a load balancing control method for a plurality of control devices in which application requirements are distributed.

One aspect of the present invention is an information processing device. In the information processing device, the weight used for distributing the request for a predetermined process for each of the plurality of control devices is assigned to the control device as the number of destination devices to be controlled increases. A processing unit for determining the number of processing units is provided. Each of the plurality of control devices generates a message by executing a predetermined process according to the first receiving unit that receives the request distributed to the own device among the requests distributed among the plurality of control devices. If the control device that controls the destination device of the message is its own device, the message is sent to the destination device, and if the control device that controls the destination device of the message is another control device. Includes a control unit that transfers the message to the other control device.

According to the disclosed information processing device and the load balancing control method, it is possible to reduce the transfer of messages between control devices for a plurality of control devices in which application requests are distributed.

FIG. 1 is a diagram showing an example of a system configuration of the distributed communication control system according to the first embodiment. FIG. 2 is an example of the hardware configuration of the server. FIG. 3 is a diagram showing an example of the functional configuration of the server. FIG. 4 is a diagram showing an example of a distributed communication control system assumed in a specific example of the weight determination process used by the load balancer. FIG. 5 is an example of the measurement results of the request amount and the message amount of the server 1 in the example shown in FIG. FIG. 6 is an example of the measurement results of the request amount and the message amount of the server 2 in the example shown in FIG. FIG. 7 is a diagram showing a direct ratio for each application in each server in a specific example. FIG. 8 is an example of the generation request amount of each application in the specific example. FIG. 9 is an example of the capacity of each server according to the specific example. FIG. 10 is a diagram showing an example of a solution Xij of a linear programming problem according to a specific example. FIG. 11 is an example of the round robin weight Wij notified to the load balancer according to a specific example. FIG. 12 is an example of a flowchart of the round robin weight determination process of the load balancer of the control unit. FIG. 13 shows an example of the message amount of each server in the comparative example when the request amount state of each application is the same condition, and the message amount of each server when the weight Wij calculated in the specific example is used. It is a figure which shows. FIG. 14 is a diagram showing an example of the constant Aij of the linear programming problem according to the modified example of the first embodiment. FIG. 15 is a diagram showing an example of a round robin weight determination process of the load balancer according to the second embodiment. FIG. 16 is an example of a flowchart of the round robin weight determination process of the load balancer of the control unit according to the second embodiment. FIG. 17 is a diagram showing an example of the capacity Cj of the server according to the third embodiment. FIG. 18 is a diagram showing an example of the value of the requested amount Xij of the application i assigned to the server j calculated by the linear programming method for setting the value of the capacity Cj of the server of FIG. FIG. 19 is a diagram showing an example of the round robin weight of the load balancer corresponding to the value of the request amount Xij of the application i assigned to the server j of FIG. FIG. 20 shows the server 1 when the request amount of each application is distributed to the server 1 and the server 2 by using the weight Wij shown in FIG. 19 when the state of the request amount of each application is the same condition. It is a figure which shows an example of the message amount between and a server 2. FIG. 21 is an example of a flowchart of the round robin weight determination process of the load balancer in the process of the control unit according to the fourth embodiment. FIG. 22 is a diagram showing an example of a server having the highest direct ratio for each application according to the fifth embodiment. FIG. 23 shows the message amount of each server in the comparative example when the request amount status of each application is the same condition, and each server when the request is distributed to the server shown in FIG. 22 for each application. It is a figure which shows an example of the message amount of. FIG. 24 is a diagram showing an example of a distributed communication control system. FIG. 25 is a diagram showing an example of a flow of control messages when a path setting request is input to a path setting application in a distributed communication control system. FIG. 26 is a diagram showing an example of a distributed communication control system including a load balancer. FIG. 27 is a diagram showing an example of a distributed communication control system assumed in the specific example according to the sixth embodiment. FIG. 28 is an example of master information according to a specific example of the sixth embodiment. FIG. 29 is a diagram showing an example of master change information according to a specific example of the sixth embodiment. FIG. 30 is a diagram showing an example of a switch list according to a specific example of the sixth embodiment. FIG. 31 is a diagram showing an example of the direct ratio Aij in the controller j of the application i according to the specific example of the sixth embodiment. FIG. 32 is a diagram showing an example of an application generation request amount according to a specific example of the sixth embodiment. FIG. 33 is an example of the round robin weight Wij of the load balancer according to the specific example of the sixth embodiment. FIG. 34 is a diagram showing an example of simulation results according to a specific example of the sixth embodiment. FIG. 35 is a diagram showing an example of master information after application of master change information # 2 according to a specific example of the sixth embodiment. FIG. 36 is an example of a flowchart of the master change determination process according to the sixth embodiment. FIG. 37A is an example of a flowchart of the simulation process. FIG. 37B is an example of a flowchart of the simulation process. FIG. 38 is a diagram showing an example of the direct ratio Aij in the controller j of the application i according to the specific example of the seventh embodiment. FIG. 39 is a diagram showing an example of the variance value of the direct ratio of each application according to the specific example of the seventh embodiment. FIG. 40 is an example of the amount of messages transmitted to each switch of application # 1 according to the specific example of the seventh embodiment. FIG. 41 is an example of a flowchart of the master change determination process according to the seventh embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. The configurations of the following embodiments are examples, and the present invention is not limited to the configurations of the embodiments.

<First Embodiment>
FIG. 1 is a diagram showing an example of a system configuration of the distributed communication control system 100 according to the first embodiment. The distributed communication control system 100 includes a server 1, a server 2, a load balancer 3, a plurality of switches 9, and an external client 8. An application and a controller are mounted on the server 1 and the server 2, respectively. The load balancer 3 distributes and transmits the request to the application from the external client 8 to the server 1 and the server 2 by the weighted round robin. Each of the server 1 and the server 2 operates as a master of a plurality of switches 9.

Message transfer between controllers occurs when the instance that generates the message and the controller that is the master of the message destination switch 9 are on different servers. Therefore, the more likely it is that the server on which the instance that generates the message exists and the server on which the controller that is the master of the message destination switch 9 exists will be the same server, the more likely it is that message transfer will occur between the controllers. Will be low.

One of the factors that influences the possibility that the instance that generates the message and the controller that is the master of the message destination switch 9 are the same is that each controller of each controller is controlled as the master. The number of switches to do. Therefore, by allocating more requests to the controller with a large number of switches to be controlled, the instance that generates the message and the controller that is the master of the message destination switch 9 are located on the same server. It is possible to increase the possibility of becoming.

In addition, one of the other factors that influences the possibility that the instance that generates the message and the controller that is the master of the message destination switch 9 are the same is that the switch 9 that is the control target of the application is the same. There is a bias in the controller used as the master. There are a plurality of types of applications, and each type has a bias in the switch 9 to be controlled. For example, when a predetermined application targets a switch existing in a specific area, the controller in which the switch 9 in the area is the master is biased.

The bias of the controller controlled by the switch to be controlled by the application and the number of switches controlled by the controller as the master are the ratio of the message transferred directly to the switch or the message transferred to another controller. Projected on. In the first embodiment, the load balancer 3 uses linear programming to maximize the amount of messages directly transferred to the switch based on the ratio of the amount of messages directly transferred to the switch in each server. Determine the weights used in the round robin of. Round-robin weights are determined for each application.

In the first embodiment, the server 1 determines the weight of the round robin. However, the weight is not limited to this, and the round robin weight may be determined by either the server 2 or the load balancer 3. A device that determines the weight of round robin is an example of an "information processing device". The server 1 of the first embodiment is an example of an "information processing device". Further, the server 1 and the server 2 are examples of "control devices". The switch 9 is an example of a “destination device”.

<Device configuration>
FIG. 2 is an example of the hardware configuration of the server 1. The server 1 is, for example, a dedicated computer. The server 1 includes a CPU (Central Processing Unit) 101, a main storage device 102, an auxiliary storage device 103, and a network interface 104. Also, these are connected to each other by a bus 105.

The auxiliary storage device 103 stores an OS (Operating System), various programs, and data used by the CPU 101 when executing each program. The auxiliary storage device 103 is, for example, a non-volatile memory such as an EPROM (Erasable Programmable ROM), a flash memory, or a hard disk drive. Auxiliary storage device 103
Stores, for example, the program 103P for the controller and the path setting application.

The main storage device 102 is a storage device that provides the CPU 101 with a storage area and a work area for loading a program stored in the auxiliary storage device 103, or is used as a buffer. The main storage device 102 includes, for example, a semiconductor memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The main storage device 102 is an example of a “storage unit”.

The CPU 101 executes various processes by loading and executing the OS and various application programs held in the auxiliary storage device 103 in the main storage device 102. The number of CPUs 101 is not limited to one, and a plurality of CPUs 101 may be provided. The CPU 101 is an example of a " control unit" of a "control device" and a "processing unit" of an "information processing device".

The network interface 104 is an interface for inputting / outputting information to / from the control network. The network interface 104 may be an interface that connects to a wired network or an interface that connects to a wireless network. The network interface 104 is, for example, a NIC (Network Interface Card) or the like. The network interface 104 is an example of the "first receiving unit" of the "control device" and the "receiving unit" of the "information processing device".

The hardware configuration of the server 1 shown in FIG. 2 is an example, and is not limited to the above, and components can be omitted, replaced, or added as appropriate according to the embodiment. For example, the server 1 may include a portable recording medium driving device and execute a program recorded on the portable recording medium. The portable recording medium is, for example, an SD card, a miniSD card, a microSD card, a USB (Universal Serial Bus) flash memory, a CD (Compact Disc), a DVD (Digital Versatile Disc), a Blu-ray (registered trademark) Disc, or a flash. It is a recording medium such as a memory card. The server 2 and the load balancer 3 also include a CPU, a main storage device, an auxiliary storage device, and a network interface, similar to the hardware configuration shown in FIG.

FIG. 3 is a diagram showing an example of the functional configuration of the server 1. As functional components, the server 1 includes a request receiving unit 11, a message forwarding unit 12, a switch setting unit 13, an information collecting unit 14, a control unit 15, a measurement result receiving unit 16, a weight notification unit 17, and an application request amount storage unit 18A. , Message amount storage unit 18B, master information storage unit 18C, and application 19. Request receiving unit 11, message forwarding unit 12, switch setting unit 13, information collecting unit 14, control unit 15, measurement result receiving unit 16, weight notification unit 17, application request amount storage unit 18A, message amount storage unit 18B, master information The storage unit 18C is, for example, a functional component achieved by the CPU 101 executing the controller program 103P.

The request receiving unit 11 receives a request for any of the applications 19 distributed to the server 1 by the load balancer 3. The request receiving unit 11 outputs the request to the corresponding application 19. Further, the request receiving unit 11 records the received request amount in the application request amount storage unit 18A for each application 19.

The application 19 is, for example, a path setting application. For example, the application 19 is provided for each user organization and each area in response to the request for the policy and bandwidth control of each user organization and each area. The application 19 receives a request input from the request receiving unit 11. The application 19 performs a predetermined process in response to the request, generates a message including the identification information of the switch 9 to be set according to the process result, and outputs the message to the message forwarding unit 12.

In the first embodiment, the application 19 is assumed to be a path setting application. Therefore, the request to the application 19 is, for example, a request for setting a path between two bases. In addition, the application 19 performs a process of calculating the route of the path between the two bases specified by the request. The processing result of the application 19 is the calculated route information. The route information includes, for example, identification information of the switches 9 on the route, an output interface of each switch 9 on the route, and the like. Therefore, in the first embodiment, the message generated by the application 19 includes the identification information of the switch 9 on the route obtained as a result of the path calculation. Since there are one or a plurality of switches on the route, the number of messages generated for one request is the number of switches on the route.

Further, in the first embodiment, it is assumed that OpenFlow is used as a protocol for controlling the switch 9. Therefore, the message generated by the application 19 is an OpenFlow message.

The message forwarding unit 12 receives an input of a message from the application 19. The message forwarding unit 12 refers to the master information and transmits a message to the switch setting unit 13 or the server 2 according to the message destination switch 9. The master information is correspondence information between the identification information of the switch and the identification information of the controller that is the master of the switch. The switch identification information and the identification information of the master controller are, for example, a switch name / controller name and an IP (Internet Protocol) address.

When the controller that is the master of the message destination switch 9 is mounted on the local server 1, the message forwarding unit 12 outputs the message to the switch setting unit 13. When the controller that is the master of the message destination switch 9 is mounted on another server 2, the message forwarding unit 12 transfers the message to the server 2.

The message forwarding unit 12 records the message amount transferred to the switch setting unit 13 and the message amount transferred to the server 2 in the message amount storage unit 18B for each application 19. Hereinafter, the message transferred to the switch setting unit 13 will be referred to as a direct message. Further, a message transferred to another server 2 is referred to as an indirect message.

The switch setting unit 13 receives input of a message from the message forwarding unit 12 or another server 2. The switch setting unit 13 transmits a message to the control target switch 9 whose master is the own device.

The information collecting unit 14 acquires the request amount, the direct message amount, and the indirect message amount for each application from the application request amount storage unit 18A and the message amount storage unit 18B at a predetermined cycle, and outputs the request amount to the control unit 15. .. The acquisition cycle of the request amount, the direct message amount, and the indirect message amount for each application is set in seconds by, for example, the administrator of the distributed communication control system 100.

The measurement result receiving unit 16 receives the request amount, the direct message amount, and the indirect message amount for each application of the server 2 transmitted from the other server 2 at a predetermined cycle, and outputs the request amount to the control unit 15.

The control unit 15 receives input from the information collection unit 14 for the request amount, the direct message amount, and the indirect message amount for each application in the server 1. Further, the control unit 15 receives input from the measurement result receiving unit 16 of the request amount, the direct message amount, and the indirect message amount for each application in the server 2.

The control unit 15 determines the round robin weight of the load balancer 3 based on the request amount, the direct message amount, and the indirect message amount for each application in each of the server 1 and the server 2. Details of the weight determination process will be described later. The control unit 15 outputs the determined weight to the weight notification unit 17.

The weight notification unit 17 receives an input of the round robin weight of the load balancer 3 from the control unit 15 and notifies the load balancer 3 of the weight.

The application request amount storage unit 18A, the message amount storage unit 18B, and the master information storage unit 18C each correspond to a predetermined area in the storage area of the main storage device 102 of the server 1. The application request amount storage unit 18A stores the request amount input from the load balancer 3 to the server 1 for each application. The message amount storage unit 18B stores the direct message amount and the indirect message amount for each application. The values held by the application request amount storage unit 18A and the message amount storage unit 18B may be reset by, for example, acquiring information by the information collecting unit 14, or may be held without being reset. When the values held by the application request amount storage unit 18A and the message amount storage unit 18B are not reset by the acquisition of information by the information collection unit 14, the information collection unit 14 determines the difference from the previously read value. Acquire as request amount, direct message amount, and indirect message amount within the cycle.

The master information storage unit 18C stores the master information. The master information is set in advance by, for example, the administrator of the distributed communication control system 100. Alternatively, the master information may be determined by negotiation between controllers. The method of determining the master information is not limited to these.

The functional configuration of the server 2 includes components other than the control unit 15, the measurement result receiving unit 16, and the weight notification unit 17 among the functional configurations of the server 1. In the server 2, the information collecting unit 14 transmits the information acquired from the application request amount storage unit 18A and the message amount storage unit 18B to the server 1 at a predetermined cycle.

<Details of round-robin weight determination process>
FIG. 4 is a diagram showing an example of a distributed communication control system 100 assumed in a specific example of the weight determination process used by the load balancer 3. The network formed by the switch 9 is divided into three areas: data center (DC) # 1, data center # 2, and WAN. Applications are prepared for each area of DC # 1, DC # 2, and WAN. The load balancer 3, the server 1, and the server 2 have instances of a DC # 1 control application, a DC # 2 control application, and a WAN control application for each of DC # 1, DC # 2, and WAN.

(Step 1)
The control unit 15 of the server 1 acquires the measurement results of the request amount and the message amount of the server 1 and the server 2. In a specific example, it is assumed that the results of FIGS. 5 and 6 have been obtained.

FIG. 5 is an example of the measurement results of the request amount and the message amount of the server 1 in the example shown in FIG. From FIG. 5, it is shown that the DC # 1 control application of the server 1 has received 20 requests and generated a total of 60 messages for the 20 requests. Of the messages generated by the DC # 1 control application of server 1, 40 are direct messages and 20 are indirect messages.

It is shown that the DC # 2 control application of server 1 has received 10 requests and generated a total of 40 messages for 10 requests. Of the messages generated by the DC # 2 control application of server 1, 10 are direct messages and 30 are indirect messages.

It is shown that the WAN control application of server 1 has received 5 requests and generated a total of 10 messages for 5 requests. Of the messages generated by the WAN control application of server 1, 5 is a direct message and 5 is an indirect message.

FIG. 6 is an example of the measurement results of the request amount and the message amount of the server 2 in the example shown in FIG. From FIG. 6, it is shown that the DC # 1 control application of the server 2 has received 20 requests and generated a total of 60 messages for the 20 requests. Of the messages generated by the DC # 1 control application of the server 2, 20 are direct messages and 40 are indirect messages.

It is shown that the DC # 2 control application of server 2 has received 10 requests and generated a total of 40 messages for 10 requests. Of the messages generated by the DC # 2 control application of the server 2, 30 are direct messages and 10 are indirect messages.

It is shown that the WAN control application of server 2 has received 5 requests and generated a total of 10 messages for 5 requests. Of the messages generated by the WAN control application of the server 2 , 5 is a direct message and 5 is an indirect message.

The unit of the requested amount and the message amount may be, for example, the number, the inflow / outflow amount (kbps) in a unit time, or the like.

(Step 2)
The control unit 15 of the server 1 calculates the ratio of the direct message amount to the total message amount for each combination of the application and the server based on the request amount and the message amount collected from the server 1 and the server 2. The ratio of the amount of direct messages to the total amount of messages of a predetermined application on a predetermined server is referred to as a direct ratio.

FIG. 7 is a diagram showing a direct ratio for each application on each server in a specific example. In the example shown in FIG. 5, the total number of messages of the DC # 1 control application on the server 1 is 60, and the amount of direct messages is 40. Therefore, the direct ratio for the DC # 1 control application on the server 1 is 40 /. 60 = 2/3. In the example shown in FIG. 6 , the total number of messages of the DC # 1 control application on the server 2 is 60, and the amount of direct messages is 20, so that the direct ratio for the DC # 1 control application on the server 2 is 20 /. 60 = 1/3. Similarly, for the DC # 2 control application and the WAN application, the direct ratio is calculated for each server.

(Step 3)
The control unit 15 of the server 1 obtains the request amount generated from each application based on the request amount for each application collected from the server 1 and the server 2. The request amount generated from the application is hereinafter referred to as the generated request amount.

FIG. 8 is an example of the generation request amount of each application in the specific example. The generated request amount of each application is calculated as the sum of the request amounts of the application collected from each server. For example, with respect to the DC # 1 control application, the request amount distributed to the server 1 is 20 from FIG. 5, and the request amount distributed to the server 2 is 20 from FIG. The amount is calculated as 20 + 20 = 40. Similarly for the DC # 2 control application and the WAN control application, the required amount of generation is required.

(Step 4)
The control unit 15 of the server 1 obtains the request amount that can be accommodated by each server based on the request amount for each application collected from the server 1 and the server 2. The required amount that can be accommodated by the server is hereinafter referred to as an accommodating amount.

FIG. 9 is an example of the capacity of each server according to the specific example. In the first embodiment, the capacity of each server is obtained as a value obtained by dividing the total amount of generation requests of all applications in the distributed communication control system 100 by the number of servers.

In a specific example, from FIG. 8, the total amount of generated requests for all applications is calculated to be 40 + 20 + 10 = 70 by adding the generated requests for each of the DC # 1 control application, the DC # 2 control application, and the WAN application. In a specific example, the number of servers is two, server 1 and server 2. Therefore, the capacity of each server in the specific example is calculated to be 70/2 = 35.

(Step 5)
The control unit 15 of the server 1 solves the linear programming problem by using the direct ratio, the generated request amount of each application, and the capacity of each server. In the first embodiment, the objective function and the constraint condition are set as follows.

The objective function is a function that estimates the total amount of direct messages for each server and each application. Constraint 1 imposes an upper limit on the amount of request allocated to each server. The constraint condition 2 limits the generated request amount of each application to the same state as at the time of measurement.

The control unit 15 of the server 1 solves a linear programming problem in which the objective function is maximized under the above constraints. As a solution to the linear programming problem, the value of Xij when the objective function is maximized is obtained. That is, the control unit 15 of the server 1 sends each application to each server so that the total amount of direct messages is maximized in the state where the same request is generated as at the time of measurement and the capacity of the server is limited. Find the required amount to be allocated. When the total amount of direct messages is maximum, the amount of indirect messages, that is, the amount of messages transferred between controllers is minimum.

In a specific example, the value of Aij is the value of the direct ratio for each application on each server shown in FIG. The value of Ri is the value of the generated request amount of each application shown in FIG. The value of C is the capacity value of each server shown in FIG.

FIG. 10 is a diagram showing an example of a solution Xij of a linear programming problem according to a specific example. The server 1 is assigned 35 requests for the DC # 1 control application. The server 2 is assigned 5 requests for the DC # 1 control application, 20 requests for the DC # 2 control application, and 10 requests for the WAN control application.

(Step 6)
The control unit 15 of the server 1 obtains the ratio of requests assigned to each server for each application from the solution Xij of the linear programming problem. The control unit 15 of the server 1 notifies the load balancer 3 through the weight notification unit 17 of the ratio of requests assigned to each server for each application as the weight used by the load balancer 3 in round robin.

The weight Wij of distribution of requests to the server j for the application i is obtained by Wij = Xij / Ri. Ri is the requested amount of application i.

FIG. 11 is an example of the round robin weight Wij notified to the load balancer 3 according to a specific example. After that, the load balancer 3 distributes the requests of the DC # 1 control application to the server 1 with a weight of 7/8 and the server 2 with a weight of 1/8, and requests the DC # 2 control application and the WAN control application. All will be distributed to server 2.

The control unit 15 of the server 1 repeatedly executes the above steps 1 to 6 at a predetermined cycle. Since the weight of the round robin of the load balancer 3 also changes according to the change in the demand amount that occurs, the message transferred between the controllers is adjusted to be minimized according to the change in the demand amount. The direct message is an example of the "first message". The indirect message is an example of a "second message". The "direct ratio" is an example of the "first ratio".

<Processing flow>
FIG. 12 is an example of a flowchart of the round robin weight determination process of the load balancer 3 of the control unit 15 of the server 1. The process shown in FIG. 12 is repeatedly executed in the same cycle as the cycle in which the information on the request amount and the message amount of the server 1 and the server 2 is input to the control unit 15 of the server 1. The execution subject of the process shown in FIG. 12 is the CPU 101, but for convenience, the control unit 15 which is a functional component will be described as the main body.

In OP1, the control unit 15 acquires the request amount, the direct message amount, and the indirect message amount for each combination of the server and the application from the information collection unit 14.

In OP2, the control unit 15 obtains the direct ratio Aij for each combination of the server and the application. In OP3, the control unit 15 obtains the generated request amount Ri for each application. In OP4, the control unit 15 obtains the capacity C of the server. In OP5, the control unit 15 solves a linear programming problem.

In OP6, the control unit 15 calculates the weight Wij for distributing the request to the server j for the application i, and notifies the load balancer 3 of the weight Wij. After that, the process shown in FIG. 12 ends.

<Action and effect of the first embodiment>
FIG. 13 shows the message amounts of the server 1 and the server 2 in the comparative example when the state of the requested amount of each application is the same, and the server 1 and the server 2 when the weight Wij calculated in the specific example is used. It is a figure which shows an example with the message amount of. The comparative example in FIG. 13 is an example in which the requests are evenly distributed between the server 1 and the server 2. The measurement results of the server 1 and the server 2 of the comparative example in FIG. 13 are the same as those shown in FIGS. 5 and 6, respectively.

In the comparative example, the total amount of indirect messages of all applications on the server 1 is 55, and the total amount of indirect messages of all applications on the server 2 is 55. The total amount of indirect messages between the server 1 and the server 2 in the comparative example is 55 + 55 = 110.

When the weight Wij calculated in the specific example is used, the total amount of indirect messages of all applications on the server 1 is 35, and the total amount of indirect messages of all applications on the server 2 is 40. In a specific example, the total amount of indirect messages between the server 1 and the server 2 is 35 + 40 = 75.

Therefore, according to the first embodiment, the amount of messages (indirect messages) transferred between the controllers can be reduced by changing the weight of the round robin of the load balancer 3. Further, in FIG. 13, the total request amount of the server 1 and the total request amount of the server 2 of the specific example are the same. This is due to the fact that the upper limit of the server capacity is set as a constraint in the linear programming problem. Therefore, according to the first embodiment, it is possible to distribute the processing load between the servers while reducing the amount of messages transferred between the controllers. In addition, linear programming is relatively easy to implement because existing libraries can be used.

In the first embodiment, the direct ratio is set to the constant Aij of the linear programming problem, the solution Xij that maximizes the objective function is obtained, and the round robin weight of the load balancer 3 is determined based on the solution Xij. NS. Instead of this, the indirect ratio, which is the ratio of the indirect messages to the total amount of messages of the application i on the server j, may be set as the constant Aij of the linear programming problem. In this case, the objective function is the amount of messages transferred between the controllers (indirect message amount), so the solution Xij is calculated assuming that the objective function is minimized. The round robin weight of the load balancer 3 is the ratio of the requests assigned to each server for each application obtained from the solution Xij.

<Modified example of the first embodiment>
In the first embodiment, the direct ratio of each server in each application is used for the constant Aij of the linear programming problem. Instead of this, if the switch 9 to be set for each application is known in advance, the ratio of the switches in which each controller is the master may be set as the constant Aij of the linear programming problem. For example, in the example shown in FIG. 4, in the DC # 1 control application, of the three switches to be set, the server 1 is the master switch is two, and the server 2 is the master switch. In the DC # 2 control application, out of the four switches to be set, one switch has the server 1 as the master, and three switches have the server 2 as the master. In the WAN control application, out of the four switches to be set, the server 1 is the master switch is two, and the server 2 is the master switch is two.

FIG. 14 is a diagram showing an example of the constant Aij of the linear programming problem according to the modified example of the first embodiment. The example shown in FIG. 14 is premised on the system configuration of FIG. The constant Aij of the example linear programming problem shown in FIG. 14 is the percentage of switches in each application for which each controller is the master. Since the value of the constant Aij of the linear programming problem of the example shown in FIG. 14 matches the value of the constant Aij of the linear programming problem according to the first embodiment shown in FIG. 7, as a solution of the linear programming problem, The same values as in the example shown in FIG. 10 are obtained. Therefore, the amount of messages transferred between the controllers can be reduced even by the modification of the first embodiment.

<Second Embodiment>
In the second embodiment, heuristic algorithms are used instead of linear programming to determine the round robin weights that the load balancer 3 uses to assign requests to each server for each application. A heuristic algorithm is a method that does not always lead to the correct answer, but can obtain a solution that is close to the correct answer at a certain level.

In the second embodiment, the description common to the first embodiment is omitted. In the second embodiment, the system configuration, the hardware configuration of each device, and the functional configuration are the same as those in the first embodiment. In the second embodiment, the processing of the control unit 15 in the functional configuration is different from that in the first embodiment.

FIG. 15 is a diagram showing an example of a round robin weight determination process of the load balancer 3 according to the second embodiment. In FIG. 15, a specific example shown in FIG. 4 is assumed. In the second embodiment, the control unit 15 performs step 1 (information collection), step 2 (direct ratio), step 3 (calculation of the generation request amount of each application), and step of the round robin weight determination process of the first embodiment. 4 (Calculation of capacity of each server) is executed. Subsequent processing is as follows.

Table shown in FIG. 15, for each application i, the required amount r (i), direct ratio a (i, j) and the including of each server j. The direct ratios a (i, j) of each server j are sorted in descending order of value. The initial value of the request amount r (i) of the application i is the request generation amount Ri of the application i. Further, in FIG. 15, the capacity c (j) of the server j is shown. The initial value of the capacity c (j) of the server j is the capacity C obtained by dividing the total amount required by all applications by the number of controllers. The direct ratio a (i, j), the initial value Ri of the requested quantity r (i), and the initial value C of the server capacity c (j) are the same as the examples shown in FIGS. 7, 8 and 9, respectively. The value.

First, the control unit 15 selects the combination of the DC # 2 control application and the server 2 having the highest direct ratio a (2,2) (= 3/4), and the server 2 can accommodate the server 2. Allocate the requirements of the DC # 2 control application. In FIG. 15, since the capacity c (2) of the server 2 is 35 and the required amount r (2) of the DC # 2 control application is 20, the required amount r (2) of the DC # 2 control application is sent to the server 2. = 20 All are assigned. The required amount r (2) of the DC # 2 control application can be rewritten from 20 to 0. Since the requested amount r (2) of the DC # 2 control application has become 0, the processing for the DC # 2 control application ends.

Next, the control unit 15 selects the combination of the DC # 1 control application and the server 1 having the second highest direct ratio a (1,1) (= 2/3), and the server 1 is set to the server 1. Allocate the required amount of DC # 1 control application that can be accommodated. In FIG. 15, since the capacity c (1) of the server 1 is 35 and the required amount r (1) of the DC # 1 control application is 40, the server 1 has the capacity c (1) of the server 1. Up to 35 DC # 1 control application requirements are allocated. The required amount c (1) of the DC # 1 control application can be rewritten from 40 to 5. Since the request of the DC # 1 control application is assigned to the server 1 up to the upper limit of the capacity, the server 1 is excluded from the processing target thereafter.

Next, the control unit 15 selects a combination of the WAN control application and the server 2 having a direct ratio a (3,2) (= 1/2), and the WAN control application that the server 2 can accommodate in the server 2. Allocate the required amount of. Since the server 2 has already been assigned 20 requests for the DC # 2 control application, the capacity c (2) of the server 2 is 15. Since the request amount r (3) of the WAN control application is 10, the server 2 is assigned all the request amounts r (3) = 10 of the WAN control application. The required quantity r (3) of the WAN control application can be rewritten from 10 to 0. Since the request amount r (3) of the WAN control application has become 0, the processing for the WAN control application ends.

Next, the control unit 15 selects a combination of the DC # 1 control application and the server 2 having a direct ratio a (1,2) (= 1/3), and the server 2 is DC enough to accommodate the server 2. # 1 Allocate the required amount of the control application. Since the request of the DC # 2 control application is already 20 and the request of the WAN control application is 10 assigned to the server 2, the capacity c (2) of the server 2 is 5. In FIG. 15, since the capacity c (2) of the server 2 is 5 and the request amount r (1) of the DC # 1 control application is 5, the remaining request amount r (1) of the DC # 1 control application is sent to the server 2. 1) = 5 All are assigned. The required amount r (1) of the DC # 1 control application can be rewritten from 5 to 0. Since the request amount r (1) of the DC # 1 control application becomes 0 and the request amount of all the applications becomes 0, the process ends.

The ratio of the request amount allocated to each server for each application in FIG. 15 is the same as the result when the linear programming method of the first embodiment is used (see FIG. 10).

FIG. 16 is an example of a flowchart of the round robin weight determination process of the load balancer 3 of the control unit 15 of the server 1 according to the second embodiment. The process shown in FIG. 16 is repeatedly executed in the same cycle as the cycle in which the information on the request amount and the message amount of the server 1 and the server 2 is input to the control unit 15 of the server 1. The execution subject of the process shown in FIG. 16 is the CPU 101, but for convenience, the control unit 15 which is a functional component will be described as the main body.

In OP11, the control unit 15 acquires the request amount, the direct message amount, and the indirect message amount for each combination of the server and the application from the information collection unit 14.

In OP12, the control unit 15 uses the direct ratio a (i, j) in the application i on the server j, the initial value Ri of the requested quantity r (i) of the application i, and the initial value of the accommodation capacity C (j) of the server j. Find C. Each method is the same as in the first embodiment.

In OP13, the control unit 15 initializes the request amount x (i, j) assigned to the application i on the server j and the flag f (i, j) corresponding to the application i on the server j with 0, respectively. do. If the flag f (i, j) corresponding to the application i on the server j is 0, it is unprocessed, and if it is 1, it is indicated that it has been processed.

In OP14, the control unit 15 determines whether or not the requested amounts r (i) of the application i are all 0. When the requested amounts r (i) of the application i are all 0 (OP14: YES), the process shown in FIG. 16 ends. If any of the request amounts r (i) of the application i is not 0 (OP14: NO), the process proceeds to OP15.

In OP15, the control unit 15 obtains the application i and the server j having the maximum direct ratio a (i, j) among the combinations of the application i and the server j in which the flag f (i, j) is 0. .. The application i and the server j obtained in OP15 are hereinafter referred to as the application ii and the server jj, respectively.

In OP16, the control unit 15 uses the requested amount r (ii) of the application ii and the server JJ.
The capacity c (jj) of the above and the required amount x (ii, JJ) allocated to the application ii on the server jj are updated. m is the smaller value of r (ii) and c (jj).

The requested amount r (ii) of the application ii is updated to a value obtained by subtracting m from r (ii). The requested amount x (ii, jj) allocated to the application ii on the server jj is updated to a value obtained by subtracting m from x (ii, jj). The requested amount x (ii, jj) allocated to the application ii on the server jj is updated to a value obtained by adding m to x (ii, JJ).

In OP17, the control unit 15 sets the flags f (ii, j) (j is all servers) of all servers jj to 1 for the application ii in which the requested amount r (ii) of the application ii is 0. If the requested amount r (ii) of the application ii does not become 0, the processing of OP17 is omitted.

In OP18, the control unit 15 sets the flags f (i, JJ) of all applications ii (i is all applications ) to 1 for the server jj in which the capacity c (jj) of the server jj is 0. If the capacity c (jj) of the server jj does not become 0, the processing of OP18 is omitted. After that, the process proceeds to OP14, and the process from OP14 is repeatedly executed.

In the second embodiment, the heuristic algorithm determines the amount of demand allocated to each server for each application. The result is the same as when linear programming is used to determine the amount of demand allocated to each server for each application. Therefore, according to the second embodiment as well, it is possible to reduce the transfer of messages between the controllers while distributing the processing load between the servers.

In addition, by using a heuristic algorithm, it is possible to obtain the required amount assigned to each server for each application at a higher speed.

<Third Embodiment>
In the first embodiment, the server capacity C, which is one of the constraints of the linear programming problem, is a constant having a value obtained by dividing the sum of the demands of all applications by the number of controllers and equalizing among the servers. be. Further, also in the second embodiment, the initial value of the accommodation capacity c (j) of the server j is a value equal among the servers obtained by dividing the required amount of all applications by the number of controllers. In the third embodiment, an allowable range is set for the capacity of the server. In the third embodiment, the description common to the first embodiment and the second embodiment will be omitted. The third embodiment assumes that the system configuration, the hardware configuration, and the functional configuration are the same as those of the first embodiment.

The constraints on the capacity of the server of the third embodiment are as follows.

α is a coefficient representing the allowable range of error. The expression of the error is not limited to the method of specifying by the magnification with respect to C, and may be expressed by adding or subtracting the permissible range α of the error to C, for example.

FIG. 17 is a diagram showing an example of the capacity Cj of the server j according to the third embodiment. The premise of FIG. 17 is the same as the specific example of the first embodiment. For example, the capacity C1 of the server 1 is required to be 40, and the capacity C2 of the server 2 is required to be 35.

FIG. 18 is a diagram showing an example of the value of the required amount Xij assigned to the application i on the server j calculated by the linear programming method for setting the value of the capacity Cj of the server j in FIG.

FIG. 19 is a diagram showing an example of the round robin weight of the load balancer 3 corresponding to the value of the request amount Xij assigned to the application i on the server j of FIG.

FIG. 20 shows the server 1 when the request amount of each application is distributed to the server 1 and the server 2 by using the weight Wij shown in FIG. 19 when the state of the request amount of each application is the same condition. It is a figure which shows an example of the message amount between and a server 2.

When the weight Wij shown in FIG. 19 is used, the total amount of indirect messages of all applications on the server 1 is 40, and the total amount of indirect messages of all applications on the server 2 is 30. In a specific example, the total amount of indirect messages between the server 1 and the server 2 is 30 + 30 = 70. For example, when compared with the measurement result of the comparative example shown in FIG. 13, the total amount of indirect messages in the measurement result of the comparative example is 110, so that the amount of messages transferred between the controllers is reduced. I understand.

Therefore, according to the third embodiment, the amount of messages (indirect messages) transferred between the controllers can be reduced more efficiently by providing an error tolerance in the capacity of the server.

Further, in the third embodiment, an example by the linear programming method of the first embodiment has been described, but the technique described in the third embodiment can also be applied when the heuristic algorithm of the second embodiment is used. ..

<Modified example of the third embodiment>
If the performance difference between the servers is known in advance, the accommodation capacity Cj may be changed according to the performance difference between the servers. When the capacity Cj is changed according to the performance difference between the servers, the constraint conditions for the capacity of the server are as follows.

αj is a coefficient representing the performance of the server j. The value of the coefficient αj of the server j having the lowest performance is set to 1. It should be noted that the constraint condition of the capacity of the server when the capacity Cj is changed according to the performance difference between the servers is not limited to the method of specifying by the magnification with respect to the capacity C of the server. For example, the constraint condition of the capacity of the server when the capacity Cj is changed according to the performance difference between the servers may be expressed by addition or subtraction with respect to the capacity C of the server.

<Fourth Embodiment>
In the first embodiment, the second embodiment, and the third embodiment, the instance of the application i for which the round robin weight Wij of the load balancer 3 is calculated to be 0 is on the server j even though the request is not input. Continues to work with. The resource on the server j is consumed by the instance on the server j of the application i for which no request is input.

In the fourth embodiment, the operation of the instance of the application i on the server j whose calculated round robin weight Wij of the load balancer 3 is less than a predetermined threshold value δ is stopped. In the fourth embodiment, the description common to the first embodiment, the second embodiment, and the third embodiment is omitted. In the fourth embodiment, the system configuration, the hardware configuration, and the functional configuration are assumed to be the same as those in the first embodiment.

For example, when the round robin weight Wij of the load balancer 3 shown in FIG. 11 is acquired and the threshold value δ = 0.01, the control unit 15 and the DC # 2 control application on the server 1 Determines whether to stop the operation of the instance with the WAN control application. The control unit 15 notifies the load balancer 3 of the stop of the operation of the instances of the DC # 2 control application and the WAN control application on the server 1 together with the weight Wij through the weight notification unit 17.

FIG. 21 is an example of a flowchart of the round robin weight determination process of the load balancer 3 in the process of the control unit 15 according to the fourth embodiment. The process shown in FIG. 21 is repeatedly executed in the same cycle as the cycle in which the information on the request amount and the message amount of the server 1 and the server 2 is input to the control unit 15 of the server 1. The execution subject of the process shown in FIG. 21 is the CPU 101, but for convenience, the control unit 15 which is a functional component will be described as the main body.

In OP21, the control unit 15 acquires the request amount, the direct message amount, and the indirect message amount for each combination of the server and the application from the information collection unit 14.

In OP22, the control unit 15 obtains the direct ratio Aij for each combination of the server and the application. In OP23, the control unit 15 obtains the generated request amount Ri for each application. In OP24, the control unit 15 obtains the capacity C of the server. In OP25, the control unit 15 solves a linear programming problem. In OP26, the control unit 15 determines the ratio of the requests assigned to each server for each application from the request amount assigned to each server of each application obtained by solving the linear programming problem in the round robin of the load balancer 3. Calculated as the weight Wij.

In OP27, the control unit 15 determines whether or not there is a weight Wij less than the threshold value δ. If there is a weight Wij less than the threshold value δ (OP27: YES), the process proceeds to OP28. If there is no weight Wij less than the threshold value δ (OP27: NO), the process proceeds to OP29.

In OP28, the control unit 15 notifies the load balancer 3 that the instance of the application i on the server j corresponding to the weight Wij less than the threshold value δ is stopped.

In OP29, the control unit 15 notifies the load balancer 3 of the round robin weight Wij. After that, the process shown in FIG. 21 ends.

In the fourth embodiment, when the calculated round robin weight Wij of the load balancer 3 is less than the threshold value δ, the operation of the instance of the application i on the corresponding server j is stopped. As a result, the resource consumption of the server j can be reduced.

In the fourth embodiment, the case of applying to the first embodiment has been described, but the technique described in the fourth embodiment is not limited to this, and the techniques described in the fourth embodiment are the second embodiment and the third embodiment. It is also applicable to.

<Fifth Embodiment>
In the fifth embodiment, the round robin weight of the load balancer 3 is determined so that the request of each application is directly assigned to the server having the highest ratio for each application.

FIG. 22 is a diagram showing an example of a server having the highest direct ratio for each application according to the fifth embodiment. The figure shown in FIG. 22 is an example in the case where the direct ratio of each server is the value shown in FIG. 7 for each application. When the control unit 15 calculates the direct ratio of each server for each application, the control unit 15 selects the server having the highest direct ratio for each application. When there are a plurality of servers having the highest direct ratio, for example, the control unit 15 randomly selects one of the servers.

In FIG. 7, the server 1 has the highest direct ratio in the DC # 1 control application, so the server 1 is selected in FIG. 22. In FIG. 7, the server 2 has the highest direct ratio in the DC # 2 control application, so the server 2 is selected in FIG. 22. In FIG. 7, in the WAN control application, both the server 1 and the server 2 have the same value as the direct ratio 1/2. Therefore, in FIG. 22, the server 1 is selected as a result of being randomly selected.

Therefore, in FIG. 22, it is shown that all the requests of the DC # 1 control application are allocated to the server 1. It is shown that all requests of the DC # 2 control application are allocated to server 2. It is shown that all WAN control application requests are allocated to server 1.

FIG. 23 shows the amount of messages between the server 1 and the server 2 in the comparative example when the state of the request amount of each application is the same, and the case where the request is distributed to the server shown in FIG. 22 for each application. It is a figure which shows an example of the message amount between the server 1 and the server 2. The comparative example shown in FIG. 23 is the same as the comparative example shown in FIG. 13 when the request is evenly distributed between the server 1 and the server 2.

For each application, when requests are allocated to the server shown in FIG. 22, the total amount of indirect messages of all applications on server 1 is 50, and the total amount of indirect messages of all applications on server 2 is 20. .. In a specific example, the total amount of indirect messages between the server 1 and the server 2 is 50 + 20 = 70. Since the total amount of indirect messages between the server 1 and the server 2 in the comparative example is 110, it is shown that the transfer of messages between the controllers is reduced in the fifth embodiment.

Therefore, according to the fifth embodiment, it is possible to reduce the amount of messages (indirect messages) transferred between the controllers, which allocate all requests to the server having the highest direct ratio for each application.

<Others>
The round robin weight of the load balancer 3 may be the ratio of the number of switches whose masters are each controller in the distributed communication control system 100 as a whole. In this case, for example, the control unit 15 calculates the ratio of the switches that are the masters for each server from the master information, and notifies the load balancer 3 of the calculated ratio. As a result, the larger the number of switches that are the master, the more requests can be distributed. Further, in this case, the bias of the controller, which is the master of the switch to be set by the application, is not taken into consideration.

<Sixth Embodiment>
In the sixth embodiment, by changing the controller that becomes the master of the switch 9 to another controller, the number of switches 9 for which the controller is in charge of the master is biased in the switch 9 in the setting area of the application. In the sixth embodiment, the description overlapping with the first to fifth embodiments is omitted.

For example, in the distributed communication control system 100 shown in FIG. 4, there are two switches whose master is the server 1 and two switches whose master is the server 2 in the setting area of the WAN control application. .. In this case, within the setting area of the WAN control application, there is no bias in the number of switches in which the controller is in charge of the master between the server 1 and the server 2.

If the number of switches 9 in which the controller is in charge of the master is not biased, the effect of reducing the amount of messages between the controllers can be obtained regardless of how the round robin weight is set for the WAN control application of the load balancer 3. small. This is because, for example, assuming that messages are evenly transmitted to each switch 9, the amount of messages corresponding to the distributed request amount in each controller is set regardless of the value of the round robin weight. This is because it is evenly distributed among the controllers.

In the sixth embodiment, the master of the switch group is intentionally biased to a specific controller to increase the effect of reducing messages between controllers. However, multiple applications are running in the system. Therefore, in the sixth embodiment, the switch to be changed to the master is determined so that the total amount of inter-controller messages of all applications is reduced.

Further, even if the master of the switch 9 is changed to reduce the amount of messages between controllers, if the message processing is concentrated on a specific controller, the load on the controller becomes large. In this case, the performance of the entire system may deteriorate. Therefore, in the sixth embodiment, the switch to be changed and the changed master of the switch are determined so that the load between the controllers is equalized.

Specifically, as the number of messages sent from the Con preparative roller to the switch 9 is equalized and further, as the amount required to be allocated to each controller is equalized, the switch 9 for which you want to change the master And the changed master of the switch 9 are determined.

In the sixth embodiment, the server 1 determines the switch to be changed and the master after the change of the switch. However, the determination of the switch for which the master is to be changed and the master after the change of the switch is not limited to the server 1, and either the server 2 or the load balancer 3 may make the determination.

In the sixth embodiment, the system configuration of the distributed communication control system 100, the hardware configuration and the functional configuration of the server 1 are the same as those of the first embodiment. However, the technique described in the sixth embodiment can also be applied to the server 1 of the second to fifth embodiments.

In the sixth embodiment, the control unit 15 of the server 1 determines a switch 9 for which the master is to be changed and a master after the change of the switch 9. Specifically, the control unit 15 determines the round robin weight of the load balancer 3 for all cases of the combination of one or a plurality of switches 9 to be changed from all the switches in the system and the changed master. Wij is obtained, and the distribution of requests of each application according to the obtained weight Wij is simulated. As a result of the simulation, the bias of the message amount between the controllers and the message amount transmitted from the controller to the switch 9 between the controllers is acquired.

The simulation is based on the required amount of each application that occurred within a predetermined period.
This is done based on the information in the message sent to the switch 9. The above-mentioned information used in the simulation is collected by another controller and the information collecting unit 14.

From the simulation result, the control unit 15 selects a case in which the amount of messages transmitted from the controller to the switch 9 is smaller than a predetermined threshold value among the cases in which the amount of messages transmitted between the controllers is small. The control unit 15 determines the switch 9 to be changed and the changed master of the switch 9 based on the selected case. The process of determining the switch 9 for which the master is to be changed and the changed master of the switch 9 is hereinafter referred to as a master change determination process.

<Details of master change decision processing>
FIG. 27 is a diagram showing an example of the distributed communication control system 100A assumed in the specific example according to the sixth embodiment. The distributed communication control system 100A according to the specific example includes two controllers # 1 and a controller # 2. Further, the distributed communication control system 100A includes seven switches # 1 to # 7.

It is assumed that the controller # 1 and the controller # 2 are instances existing in the server 1 and the server 2, respectively. Further, it is assumed that the instances of application # 1 and application # 2 exist in both the server 1 and the server 2 on which the controller # 1 and the controller # 2 are mounted. Hereinafter, the wording of the server and the controller will be used without distinction. For example, when it is referred to as controller # 1, it means server 1. When referred to as server 1, controller # 1 is indicated.

Further, in the specific example of the sixth embodiment, it is assumed that the server 1, that is, the server 1 on which the controller # 1 is mounted performs the master change determination process. The master change determination process is a process of determining a switch 9 for which a controller that is a master is to be changed and a master after the change of the switch 9. However, the device that executes the master change determination process is not limited to the server 1, and may be the server 2 and the load balancer 3.

FIG. 28 is an example of master information according to a specific example of the sixth embodiment. The master information stores information on the controller that is the master of each switch in the distributed communication control system 100A. The master information is stored in the master information storage unit 18C of the server 1. The master information may be preset by the administrator of the distributed communication control system 100A, or may be acquired by negotiation between the controllers.

In FIG. 28, the controller # 1 and the controller # 2 are described as C # 1 and C # 2, respectively. Switches # 1 to switch # 7 are described as SW # 1 to SW # 7, respectively. The same applies to the following figures.

In the distributed communication control system 100A according to the specific example, the switches whose master is the controller # 1 are switches # 1, # 2, # 5, and # 6. The switches 9 whose master is the controller # 2 are switches # 3, # 4, and # 7.

In the master change determination process according to the specific example, it is premised that the number of switches whose master is changed in the distributed communication control system 100A is limited to one. However, the number of switches whose masters are changed is not limited to one, and the masters may be changed for a plurality of switches.

(Step 1)
Based on the current master information, the control unit 15 of the server 1 shows each case of the selection of the switch to be changed and the combination of the selected switch with the changed master. Generate information. In the specific example, since it is assumed that there are two controllers, seven switches, and one switch whose master is changed, the master change information is (number of switches) × (number of controllers-1) = 7. × 1 = 7 are created. The master change information is an example of "case information". The process of step 1 is an example of "generating a plurality of case information indicating a case of selecting one or a plurality of destination devices from the plurality of destination devices".

FIG. 29 is a diagram showing an example of master change information according to a specific example of the sixth embodiment. The master change information is information regarding an example of a combination of the selection of the switch 9 for which the controller to be changed is to be changed and the changed master of the selected switch 9. In the example shown in FIG. 29, the master change information includes the identification information of the master change information, the identification information of the switch 9 to be changed to the master, and the identification information of the controller to be the master after the change of the switch 9. It is included.

In the example shown in FIG. 29, in the master change information # 1, the switch 9 to be changed is the switch # 1, and the changed master of the switch # 1 is the controller # 2. In the master change information # 2, the switch 9 to be changed is the switch # 2, and the master after the change of the switch # 2 is the controller # 2. In the master change information # 3, the switch 9 to be changed is the switch # 3, and the master after the change of the switch # 3 is the controller # 1. In the master change information # 4, the switch 9 to be changed is the switch # 4, and the master after the change of the switch # 4 is the controller # 1. In the master change information # 5, the switch 9 to be changed is the switch # 5, and the master after the change of the switch # 5 is the controller # 2. In the master change information # 6, the switch 9 to be changed is the switch # 6, and the master after the change of the switch # 6 is the controller # 2. In the master change information # 7, the switch 9 to be changed is the switch # 7, and the master after the change of the switch # 7 is the controller # 1.

For example, when the number of controllers is set to 2 and the number of switches to be changed to the master is set to 2 at the maximum, 7 to 2 switches 9 are added to the master change information shown in FIG. 29.
Master change information is created for the number of combinations (21 combinations) to be selected. The master change information is temporarily stored in the storage area in the main storage device 102, and is deleted when the master change determination process is completed.

(Step 2)
When the master change information is applied to all the master change information obtained in step 1, the control unit 15 of the server 1 changes the master of the switch 9 to be master-changed to the controller indicated by the master change information. Simulate the distribution of application requirements. As a result of the simulation, the control unit 15 acquires, for example, the amount of messages between controllers and the distributed value of the amount of messages transmitted from the controller to the switch 9. The simulation procedure is as follows. The index value indicating the variation such as the dispersion value of the amount of messages transmitted from the controller to the switch 9, the standard deviation, and the difference between the maximum value and the minimum value is "from each of the plurality of control devices to a plurality of destination devices to be controlled. This is an example of "an index value indicating a variation in the amount of transmitted messages among the plurality of control devices".

(Step 2-1)
The control unit 15 of the server 1 acquires the switch list. The switch list is a list of switches set by the application within a predetermined period. The switch set by the application is, for example, the switch to which the message created by the controller is sent in response to the request of the application. Alternatively, the switch set by the application is, for example, a switch that has received a message created by the controller in response to a request from the application.

FIG. 30 is a diagram showing an example of a switch list according to a specific example of the sixth embodiment. Since it is assumed that there are two applications in the specific example, FIG. 30 shows a switch list between application # 1 and application # 2.

The switch identification information included in the switch list is included once for each message created by the controller and sent to the switch 9 in response to the request of the corresponding application. The switch identification information included in the switch list is the identification information of the switch to which the message created by the controller is sent in response to the request of the corresponding application. For example, in the example shown in FIG. 30, the switch list of application # 1 includes SW # 1, SW # 1, SW # 5, and SW # 4. .. .. Is stored. This indicates that in application # 1, a message transmitted to SW # 1, a message transmitted to SW # 1, a message transmitted to SW # 5, and a message transmitted to SW # 4 have occurred.

Among the information of the switch of the destination of the message included in the switch list of each application, the information of the switch of the destination of the message regarding the own controller is stored in, for example, the message amount storage unit 18B. The information of the switch of the transmission destination of the message regarding the own controller is the information of the switch 9 of the transmission destination of the message transmitted by the local controller. The information collecting unit 14 reads the information of the switch to which the message regarding the own controller is transmitted from the message amount storage unit 18B, and outputs the information to the control unit 15.

Among the information of the switch of the destination of the message of each application, the information of the switch of the destination of the message regarding the other controller is acquired by being received from each controller through the measurement result receiving unit 16, for example. The information of the switch of the destination of the message regarding the other controller is the information of the switch 9 of the destination of the message transmitted by the other controller. The switch list is temporarily stored in the storage area in the main storage device 102, and is deleted when the master change determination process is completed.

The switch identification information included in the switch list is information indicating the destination of one message, and corresponds to one message. Therefore, it can be said that the switch list contains information about the message created in response to the request of the corresponding application. Therefore, thereafter, the element (switch identification information) included in the switch list may be expressed as a message.

(Step 2-2)
The control unit 15 of the server 1 directly obtains the ratio Aij for all combinations of the application i and the controller j. In the sixth embodiment, the direct ratio Aij is obtained in the same manner as in the first embodiment.

FIG. 31 is a diagram showing an example of the direct ratio Aij in the controller j of the application i according to the specific example of the sixth embodiment. In a specific example, since there are two controllers and two applications, the direct ratio Aij is a 2 × 2 matrix .

In the example shown in FIG. 31, the direct ratio A11 in controller # 1 of application # 1 is 1/4. The direct ratio A12 in controller # 2 of application # 1 is 3/4. The direct ratio A21 in controller # 1 of application # 2 is 3/5. The direct ratio A22 in controller # 2 of application # 2 is 2/5.

(Step 2-3)
The control unit 15 of the server 1 obtains the round robin weight Wij of the load balancer 3 that minimizes the message amount between controllers based on the generation request amount of each application within a predetermined period and the direct ratio Aij. The generated request amount of each application within a predetermined period is acquired from the information collecting unit 14. The round robin weight Wij of the load balancer 3 is obtained in the same manner as in the first embodiment or the second embodiment. The process of step 2-3 is to "determine the weight for each of the generated plurality of case information when the control device whose control target is the selected destination device is changed to another control device". This is an example.

FIG. 32 is a diagram showing an example of an application generation request amount according to a specific example of the sixth embodiment. In a specific example, the generation request amount of application # 1 is set to 20. Let the generation request amount of application # 2 be 10.

FIG. 33 is an example of the round robin weight Wij of the load balancer 3 according to the specific example of the sixth embodiment. In a specific example, for application # 1, the weight W11 to controller # 1 is 1/4 and the weight W12 to controller # 2 is 3/4. For application # 2, the weight W21 for controller # 1 is 1, and the weight W22 for controller # 2 is 0.

(Step 2-4)
The control unit 15 of the server 1 uses the switch list and the round robin weight Wij of the load balancer 3 to simulate the distribution of requests from each application when each master change information is applied. For example, the simulation is performed as follows.

For the application i, the control unit 15 uses the controller k (a positive integer not including k: 0), which is the master of the switch 9 to which the target message in the switch list is sent, as the target master change information and the master information. Get based on. That is, the controller k, which is the master of the switch 9 to which the target message is transmitted, is acquired based on the changed master information when the master change information is applied. The target message is transmitted from the master controller k to the transmission destination switch 9 regardless of which controller is created. Therefore, since the target message in the switch list is a message transmitted from the controller k to the switch 9, the control unit 15 increments 1 in the amount of messages transmitted from the controller to the switch 9 CMik.

The control unit 15 generates a random number using the round-robin weight Wiki of the controller k with respect to the application i, and selects the controller j to be the distribution destination of the request of the load balancer 3 based on the generated random number.

When the controller k and the controller j do not match, it is shown that the request that is the cause of creating the message is distributed by the load balancer 3 to the controllers other than the controller k. Since the master of the target message is the controller k, the target message is created by the distribution destination controller j (j ≠ k) and transmitted to the master controller k, and is transmitted from the master controller k to the destination switch 9. Will be. Therefore, when the controller k and the controller j do not match, the target message becomes an inter-controller message, and the control unit 15 increments 1 in the inter-controller message amount interMsgs.

When the controller k and the controller j match, it is shown that the request that causes the creation of the target message is distributed to the controller k by the load balancer 3. Since the master of the target message is the controller k, the target message is created by the distribution destination controller j (j = k) and transmitted to the destination switch 9. In this case, since the inter-controller message does not occur, the inter-controller message amount interMsgs is not updated.

The control unit 15 of the server 1 performs the above processing for all the messages included in the switch list of each application. When the above processing is performed for all the messages included in the switch list of each application, the control unit 15 determines the amount of messages transmitted from the controller to the switch 9 based on the amount of messages transmitted from the controller k to the switch 9. Find the variance value. The method for obtaining the variance value may be any of the well-known methods.

The value indicating the bias of the amount of transmitted messages from the controller to the switch 9 is not limited to the distributed value. As a value indicating the bias of the amount of messages directly transmitted between the controllers, for example, the standard deviation, the difference between the maximum value and the minimum value, or the like may be used.

The control unit 15 performs a simulation in all cases to which each master change information is applied, and obtains an inter-controller message amount interMsgs and a distributed value of the transmission message amount from the controller to the switch 9. The process of step 2-4 is "simulating the distribution of the request for the predetermined process based on the determined weight, the amount of messages transferred between the plurality of control devices, and from each of the plurality of control devices. This is an example of "acquiring an index value indicating variation among the plurality of control devices in the amount of messages transmitted to a plurality of destination devices to be controlled".

(Step 3)
The control unit 15 of the server 1 selects the master change information to be adopted based on the simulation result. Specifically, for example, the control unit 15 adopts the master change information having the smallest inter-controller message amount among the master change information in which the dispersion value of the transmission message amount from the controller to the switch 9 is equal to or less than the threshold value. When the master change information to be adopted is determined, the master change target switch 9 included in the master change information is selected as the master change target switch.

Of the master change information in which the distribution value of the amount of messages transmitted from the controller to the switch 9 is equal to or less than the threshold value, adopting the master change information having the smallest amount of messages between controllers is described as "transferred between the plurality of control devices. This is an example of "a condition including a condition in which the amount of messages is reduced and the amount of messages transmitted from each of the plurality of control devices to the plurality of destination devices to be controlled is equal among the plurality of control devices". .. The fact that the dispersion value of the amount of transmitted messages from the controller to the switch 9 is equal to or less than the threshold value means that "the amount of messages transmitted from each of the plurality of control devices to the plurality of destination devices to be controlled is between the plurality of control devices. It is an example of "being even."

The process of step 3 is "extracting the case information in which the amount of messages transferred between the control devices is the smallest among the case information in which the index value indicating the variation is equal to or less than a predetermined threshold value, and is selected in the extracted case information. It is an example of "determining that the control device whose control target is one or more destination devices is changed to another control device".

FIG. 34 is a diagram showing an example of simulation results according to a specific example of the sixth embodiment. For example, assume that the threshold value of the distributed value of the amount of messages transmitted from the controller to the switch 9 is 5.

In the case of the example shown in FIG. 34, the control unit 15 has master change information # 1, # 2, # 5, # 6, # 7 in which the distributed value of the amount of messages transmitted from the controller to the switch 9 is equal to or less than the threshold value (5). Is extracted. Next, the control unit 15 selects the master change information # 2 having the smallest inter-controller message amount of 10 among the master change information # 1, # 2, # 5, # 6, and # 7. That is, the control unit 15 determines the switch # 2 as the switch for changing the master.

FIG. 35 is a diagram showing an example of master information after applying the master change information # 2 according to the specific example of the sixth embodiment. The master change information # 2 indicates that the master of the switch # 2 is changed from the controller # 1 to the controller # 2. Therefore, the control unit 15 updates the master information in the master information storage unit 18C to the one in which the master of the switch # 2 is changed from the controller # 1 to the controller # 2.

Further, the control unit 15 notifies another controller, for example, the master change information # 2. The master change information is notified to other controllers, for example, through an inter-controller interface unit, which is a functional component not shown in FIG. The controller that receives the master change information updates the master information according to the master change information. This is an example of how to keep the master information consistent between the controllers in the system.

<Processing flow>
FIG. 36 is an example of a flowchart of the master change determination process according to the sixth embodiment. The process shown in FIG. 36 may be executed, for example, at a predetermined cycle, or may be executed when a predetermined event occurs. The event that triggers the execution of the process shown in FIG. 36 is, for example, that the bias of the direct ratio Aij between the controllers is less than the threshold value for the application i. The execution subject of the example shown in FIG. 36 is the CPU 101 of the server 1, but for convenience, the control unit 15 which is a functional component will be described as the main body.

In OP31, the control unit 15 generates master change information for all patterns of the combination of the switch 9 that changes the master and the changed master of the switch 9. For example, when there are two controllers and the master of one of the seven switches 9 is changed, seven master change information is generated. When there are two controllers and the master of two of the seven switches 9 is changed, 21 master change information is generated.

In OP32, the control unit 15 performs a simulation when all the master change information is applied. By the simulation, the inter-controller message amount interMsgs and the distributed value of the transmission message amount from the controller to the switch 9 are acquired when each master change information is applied. Details of the simulation process will be described later.

In OP33, the control unit 15 extracts master change information in which the dispersion value of the amount of messages transmitted from the controller to the switch 9 is equal to or less than the threshold value from the simulation result.

In the OP 34, the control unit 15 acquires the master change information with the fewest inter-controller messages among the master change information extracted in the OP 33 as the master change information to be adopted. The control unit 15 determines the switch 9 for which the master is to be changed in the acquired master change information as the switch for which the master is to be changed. After that, the process shown in FIG. 36 ends.

In OP33, if there is no master change information in which the distribution value of the amount of transmitted messages from the controller to the switch 9 is equal to or less than the threshold value, it may be determined that neither the master of any switch 9 is changed. Alternatively, the master change information having the smallest distributed value of the amount of messages transmitted from the controller to the switch 9 may be selected.

The control unit 15 may execute the following processing instead of the processing of OP33 and OP34. For example, the control unit 15 selects the master change information having the smallest amount of messages between controllers from the simulation result. Next, the control unit 15 determines whether or not the distribution value of the amount of messages transmitted from the controller to the switch 9 of the selected master change information is equal to or less than the threshold value.

When the distributed value of the amount of messages transmitted from the controller to the switch 9 of the selected master change information is equal to or less than the threshold value, the control unit 15 acquires the selected master change information as the master change information to be adopted. When the distribution value of the amount of messages transmitted from the controller to the switch 9 of the selected master change information is larger than the threshold value, the control unit 15 selects the master change information with the next smallest amount of messages between controllers, and performs the same processing. I do.

The process of OP31 in FIG. 36 corresponds to (step 1) of the master change determination process. The process of OP32 in FIG. 36 corresponds to (step 2) of the master change determination process. The processes of OP33 and OP34 in FIG. 36 correspond to (step 3) of the master change determination process.

37A and 37B are examples of flowcharts of simulation processing. The process shown in FIGS. 37A and 37B is the process executed in OP 32 of FIG. The execution subject of the example shown in FIGS. 37A and 37B is the CPU 101 of the server 1, but for convenience, the control unit 15 which is a functional component will be mainly described. The processing shown in FIGS. 37A and 37B is performed based on a message transmitted to the switch 9 in response to an application request generated during a predetermined period.

In OP41, the control unit 15 acquires a switch list for each application. The switch list is acquired from, for example, information received from another controller through the measurement result receiving unit 16 and information stored in the message amount storage unit 18B.

In OP42, the control unit 15 acquires the generation request amount of each application. The generation request amount of each application is acquired by the control unit 15 referring to the application request amount storage unit 18A.

The processes of OP43 to OP50 are repeatedly executed for each master change information. In OP43, the control unit 15 directly acquires the ratio Aij for all the applications i and all the controllers j. The direct ratio Aij is obtained, for example, in the same manner as in the first embodiment.

In OP44, the control unit 15, the weight Wij of Round Robin load balancer 3, all applications i, for all controllers j, obtains. Weight Wij of Round Robin load balancers 3 is determined in the same manner as any of the first to fifth embodiments.

The processes of OP45 to OP49 in FIG. 37B are repeated for all applications. Further, the processes of OP45 to OP49 in FIG. 37B are repeatedly executed for the number of switches to which the messages included in the switch list of one application are transmitted. That is, the processes of OP45 to OP49 in FIG. 37B are repeated a number of times corresponding to (the number of messages included in the switch list of one application) × (the number of applications) × (the number of master change information).

Hereinafter, the application to be processed is referred to as application i. The message to be processed in the switch list of application i is referred to as a target message.

In OP45, the control unit 15 acquires the controller k, which is the master of the switch 9, which is the transmission destination of the target message in the switch list of the application i, based on the target master change information and the master information. That is, the controller k, which is the master of the switch 9 to which the target message is transmitted in the switch list of the application i, is obtained based on the master information after the target master change information is applied.

In OP46, the control unit 15 adds 1 to the amount of messages sent from the controller k to the switch 9 and updates the CMik.

In OP47, the control unit 15 generates a random number based on the weight Wiki of the controller k with respect to the application i, and determines the controller j to be the distribution destination of the target message based on the generated random number.

In OP48, the control unit 15 determines whether or not the controller k, which is the master of the target message, and the controller j, which is the distribution destination of the target message, match. If the controller k and the controller j match (OP48: YES), the process proceeds to OP45 or OP50 for the next message or the next application in the switch list of the target application i. If the controller k and the controller j do not match (OP48: NO), the process proceeds to OP49.

Since it is shown in OP49 that the target message is an inter-controller message, the control unit 15 updates the inter-controller message amount interMsgs by adding 1. After that, the process proceeds to OP45 or OP50 for the next message or the next application in the switch list of the target application i.

When the processing of OP45 to OP49 is completed for all the messages in the switch list of all applications for one master change information, the processing proceeds to OP50.

In OP50, the control unit 15 obtains a distributed value of the amount of transmitted messages from each controller to the switch 9. After that, the process proceeds to OP45 for the next master change information. Alternatively, when the processing for all the master change information is completed, the processing proceeds to OP33 in FIG.

The process of OP41 in FIG. 37A corresponds to (step 2-1) of the master change determination process. The process of OP43 in FIG. 37A corresponds to (step 2-2) of the master change determination process. The processes of OP42 and OP44 in FIG. 37A correspond to (step 2-3) of the master change determination process. The process of FIG. 37B corresponds to (step 2-4) of the master change determination process.

<Action and effect of the sixth embodiment>
In the sixth embodiment, by changing the master of the switch 9, the controller in charge of the master of the switch 9 is biased. The round robin weight Wij is determined by the round robin weight Wij determination process of the load balancer 3 according to the first to fifth embodiments according to the bias of the controller in charge of the master of the switch 9. As a result, the effect of reducing the amount of messages between controllers can be improved.

The switch 9 to be changed in the master is determined by simulation so that the amount of messages transmitted from each controller to the switch 9 is not biased and the amount of messages between controllers is small. Thereby, the processing efficiency of the entire distributed communication control system 100 can be improved.

Further, in the simulation of distribution of application requests when the master change information is applied, the round robin weight Wij of the load balancer 3 is obtained based on the master information after the application of the master change information. The required round-robin weight Wij is required to minimize the total amount of inter-controller messages for each application. Therefore, according to the sixth embodiment, the switch that changes the master is determined so that the total amount of inter-controller messages of all applications is reduced.

<7th Embodiment>
In the sixth embodiment, a simulation is performed for all the master change information when the master change information is applied. In the seventh embodiment, the load related to the simulation processing is reduced and the processing speed is increased by narrowing down the master change information for executing the simulation. In the seventh embodiment, the description overlapping with the sixth embodiment is omitted. In the seventh embodiment, the system configuration of the distributed communication control system 100, the hardware configuration and the functional configuration of the server 1 are the same as those of the sixth embodiment.

<Processing to narrow down master change information>
The process of narrowing down the master change information is a process executed before the process of creating the master change information in (step 1) in the master change determination process. Therefore, the process of narrowing down the master change information will be described as a sub-step of the process of (step 1) in the master change determination process. Hereinafter, the same distributed communication control system 100A as in the sixth embodiment will be described together with a assumed specific example.

(Step 1-1)
The control unit 15 of the server 1 directly acquires the ratio Aij for all combinations of the application and the controller. The direct ratio Aij is acquired in the same manner as in the first embodiment, for example.

FIG. 38 is a diagram showing an example of the direct ratio Aij in the controller j of the application i according to the specific example of the seventh embodiment. In the specific example of the seventh embodiment, since there are two controllers and two applications, the direct ratio Aij is a 2 × 2 matrix .

In the example shown in FIG. 38, the direct ratio A11 in controller # 1 of application # 1 is 1/2. The direct ratio A12 in controller # 2 of application # 1 is 1/2. The direct ratio A21 in controller # 1 of application # 2 is 2/3. The direct ratio A22 in controller # 2 of application # 2 is 1/3.

(Step 1-2)
The control unit 15 of the server 1 obtains a distributed value for each application with respect to the direct ratio Aij obtained in step 1-1. However, the value is not limited to the variance value, and any index indicating the variation may be used. For example, the standard deviation, the difference between the maximum value and the minimum value, and the like may be used.

FIG. 39 is a diagram showing an example of the variance value of the direct ratio of each application according to the specific example of the seventh embodiment. The variance value of the direct ratio of each application shown in FIG. 39 is obtained based on the direct ratio Aij of the example shown in FIG. 38.

In the example shown in FIG. 39, the variance value of the direct ratio of application # 1 is 0. The variance value of the direct ratio of application # 2 is about 0.28.

(Step 1-3)
The control unit 15 of the server 1 extracts applications whose direct ratio variance value is equal to or less than the threshold value. The smaller the variance value of the direct ratio, the smaller the bias of the controller mastered by the switch 9 in the setting area of the application. Therefore, in an application in which the variance value of the direct ratio is equal to or less than the threshold value, the controller used by the switch 9 as the master is biased, so that a great effect can be obtained by reducing the messages between the controllers.

For example, when the variance value of the direct ratio of the application is the value of the example shown in FIG. 39 and the threshold value is 0.1, the control unit 15 extracts the application # 1. In the processes of steps 1-1 to 1-3, "the amount of messages transmitted from each of the plurality of control devices to the plurality of destination devices to be controlled from the plurality of types of predetermined processes is between the plurality of control devices. It is an example of "obtaining the first kind that is equal in". The fact that the dispersion value of the direct ratio of the application is equal to or less than the threshold value means that "the amount of messages transmitted from each of the plurality of control devices to the plurality of destination devices to be controlled is equal among the plurality of control devices". This is an example.

(Step 1-4)
The control unit 15 of the server 1 extracts the switch 9 having the largest amount of transmitted messages for the application extracted in steps 1-3. The amount of messages transmitted by a predetermined application to each switch 9 is acquired by aggregating the identification information of the switches included in the switch list of the predetermined application. Therefore, in the seventh embodiment, the switch list is acquired before the process of steps 1-4. The process of step 1-4 is an example of "acquiring a higher-order one or a plurality of destination devices having a large amount of received messages of the message of the predetermined process of the first type".

FIG. 40 is an example of the amount of messages transmitted to each switch 9 of the application # 1 according to the specific example of the seventh embodiment. In the example shown in FIG. 40, the switch # 4 is extracted because the amount of messages transmitted to SW # 4 is the largest.

(Step 1-5)
Based on the current master information, the control unit 15 of the server 1 generates master change information for which the switch extracted in step 1-4 is the master change target. For example, in the specific example of the seventh embodiment shown in FIG. 40, since the switch # 4 is extracted, the control unit 15 changes the master of the switch # 4 from the controller # 1 (see FIG. 28) to the controller # 2. Create master change information. The process of step 1-5 is an example of "generating case information for selecting the acquired one or a plurality of destination devices from a plurality of destination devices".

This can reduce the number of master change information for which the simulation is performed. In the specific example, since it is assumed that the master of one of the seven switches 9 is changed, the switch 9 having the largest amount of transmitted messages is extracted in steps 1-4. However, it is not limited to this. For example, when changing the master of a plurality of switches 9, the corresponding number of switches 9 having a large amount of transmitted messages may be extracted. The processing after step 2 is the same as that of the sixth embodiment.

FIG. 41 is an example of a flowchart of the master change determination process according to the seventh embodiment. Similar to the process shown in FIG. 36, the process shown in FIG. 41 may be executed, for example, at a predetermined cycle, or may be executed when a predetermined event occurs. The event that triggers the execution of the process is, for example, that the bias of the direct ratio Aij between the controllers is less than the threshold value for the application i. The execution subject of the example shown in FIG. 41 is the CPU 101 of the server 1, but for convenience, the control unit 15 which is a functional component will be described as the main body.

In OP60, the control unit 15 creates a switch list for each application.

In OP61, the control unit 15 directly obtains the ratio Aij for all combinations of the controller j and the application i. The method of obtaining the direct ratio Aij is, for example, the same as in the first embodiment.

In OP62, the control unit 15 obtains the variance value of the direct ratio of each application. In OP63, the control unit 15 extracts applications whose direct ratio variance value is equal to or less than the threshold value.

In OP64, the control unit 15 acquires the amount of messages to be transmitted to each switch 9 of the application extracted by OP63. In OP65, the control unit 15 sets the switch with the largest amount of transmitted messages as the master change target. However, when the number of switches 9 to be changed to the master is plural, the corresponding number of upper switches 9 having a large amount of transmitted messages is selected.

In OP66, the control unit 15 creates master change information for changing the master of the switch 9 set as the master change target in OP65. In OP67 to OP69, the same processing as OP32 to OP34 in FIG. 36 is performed on the master change information created in OP66.

When the master change information created in OP66 is one and the variance of the amount of messages transmitted from the controller to the switch 9 of the master change information is larger than the threshold value, for example, the following may be performed. Processing proceeds to OP65, and the control unit 15 sets the switch with the next largest amount of transmitted messages as the master change target, creates master change information (OP66), and simulates the master change information.

<Action and effect of the seventh embodiment>
In the seventh embodiment, master change information is created in which the master of the switch satisfying the condition is changed, not for all the patterns of the switch whose master is changed. As a result, the simulation processing can be reduced and the master change determination processing can be speeded up. In addition, the processing load of the server 1 related to the simulation can be reduced.

Further, for an application in which the bias of the controller as the master is small, the switch 9 having a large amount of transmitted messages is set as the master change target, so that it is possible to create master change information capable of further reducing the messages between controllers.

<Recording medium>
A program that enables a computer or other machine or device (hereinafter, computer or the like) to realize any of the above functions can be recorded on a recording medium that can be read by the computer or the like. The function can be provided by causing a computer or the like to read and execute the program of this recording medium.

Here, a recording medium that can be read by a computer or the like is a non-temporary recording medium that can store information such as data and programs by electrical, magnetic, optical, mechanical, or chemical action and can be read from the computer or the like. Recording medium. Among such recording media, those that can be removed from a computer or the like include, for example, a memory such as a flexible disk, a magneto-optical disk, a CD-ROM, a CD-R / W, a DVD, a Blu-ray disk, a DAT, an 8 mm tape, or a flash memory. There are cards etc. Further, as a recording medium fixed to a computer or the like, there are a hard disk, a ROM (read-only memory), and the like. Further, the SSD (Solid State Drive) can be used as a recording medium that can be removed from a computer or the like, or as a recording medium that is fixed to the computer or the like.

1 Server 2 Server 3 Load balancer 9 Switch 11 Request receiving unit 12 Message forwarding unit 13 Switch setting unit 14 Information collecting unit 15 Control unit 16 Measurement result receiving unit 17 Weight notification unit 18A Application request amount storage unit 18B Message amount storage unit 18C Master information storage unit 101 CPU
102 Main storage 103 Auxiliary storage 104 Network interface

Claims (17)

A first receiving unit that receives a request that has been distributed to the own device among the requests that are distributed among a plurality of control devices, said executes a predetermined process in response to requests that are distributed to the own device, the predetermined depending on the result of the processing, to generate a message, when equipment of the destination of the message is a device of the control target of its own device transmits the message to the device of the destination equipment of the destination of the message When is a device to be controlled by another control device, the plurality of control devices including a control unit that transfers the message to the other control device are used for sorting the request for the predetermined process. the weight to be, as the made many requirements of a given process is distributed as the number of that equipment to the control target is often the controller to the control unit, determining unit,
Information processing device equipped with. For each of the plurality of control devices, further comprising a storage unit for storing the equipment information shall be the control object,
Wherein the processing unit, for each of the plurality of control devices, the ratio of the number of said plurality of control apparatus the plurality of control units to that equipment respectively the controlled object to the total number of devices each of which the control target, the weight Decide as,
The information processing device according to claim 1. The information processing device
A receiving unit that receives a request amount of each of a plurality of types of predetermined processes received by each of the plurality of control devices is further provided from each of the plurality of control devices.
In the linear planning method, the processing unit is a control device for the total number of devices to which messages of a predetermined process of type i (1 ≤ i ≤ number of predetermined process types) among the plurality of types of predetermined processes are addressed. j the ratio of the number of equipment of the control object (the number of 1 ≦ j ≦ controller) and a constant Aij, the required amount of a given process of the type i to be allocated to the control unit j as variables Xij, sigma _j X
ij (addition _{related to Σ j} : j) is set as the upper limit of the required amount that can be distributed to the control device j, and Σ _i Xij (addition _{related to Σ i} : i) is set as the plurality of the predetermined processes of the type i. Predetermined processing of the type i assigned to the control device j that maximizes the objective function ΣAij * Xij (* is a multiplication symbol) under a constraint condition that is the sum of the requested amounts received by each of the control devices of The value of the requested amount Xij of the above is acquired, and the ratio of the acquired value of the Xij to the sum of the requested amounts received by each of the plurality of control devices for the predetermined process of the type i is determined by the predetermined type i. Is determined as a weight used for distributing the processing request to the control device j.
The information processing device according to claim 1. A first receiving unit that receives a request distributed to its own device among the requests distributed among a plurality of control devices, and a predetermined process is executed according to the request distributed to the own device to generate a message. and, if the equipment of the destination of the message is a device of the control target of its own device transmits the message to the device of the destination equipment of the destination of the message control target device of another control device In the case of, a message generated by each of the plurality of control devices executing the predetermined process from each of the plurality of control devices including a control unit for transferring the message to the other control device. The amount of the first message transmitted from each of the plurality of control devices to the device to be controlled by each of the plurality of control devices, and the second message transferred from each of the plurality of control devices to another control device. The amount of messages, the receiver that receives, and
A first ratio of the amount of the first message to the total amount of messages generated by the predetermined processing by each of the plurality of control devices, or a second of the amount of the second message. Based on the ratio, for each of the plurality of control devices, the control device having a higher first ratio or a lower second ratio requires a larger amount of the predetermined processing to be distributed to the control device. A processing unit that determines the weight used for distributing the request for the predetermined processing, and a processing unit that determines the weight.
Information processing device equipped with. Each of the plurality of control devices executes a plurality of types of predetermined processes to generate a message.
The receiving unit receives from each of the plurality of control devices the amount of the first message and the amount of the second message for each of the plurality of types of predetermined processes.
The processing unit obtains the first ratio or the second ratio for each of the plurality of types of predetermined processes executed by each of the plurality of control devices, and the plurality of the plurality of predetermined processes for each of the plurality of types. Determines the weights used to distribute the requirements to each of the controllers in
The information processing device according to claim 4. The processing unit determines an upper limit value of a required amount that can be distributed to each of the plurality of control devices, and based on the first ratio or the second ratio and the upper limit value of the required amount, the first The control device having a higher ratio of The weight used for distributing the request to each of the plurality of control devices is determined so as not to exceed the upper limit value of each of the control devices.
The information processing device according to claim 4 or 5. The receiving unit, from each of the plurality of control devices, further receives the respective requirements of the double several predetermined process each of the plurality of control apparatus has received,
In the linear programming method, the processing unit is a predetermined process of type i (type of 1 ≦ i ≦ processing) among the plurality of types of predetermined processing executed by the control device j (1 ≦ j ≦ the number of control devices). _{Σ j} Xij (Σ _j :), where the first ratio or the second ratio with respect to (number of The addition related to j) is set as the upper limit of the required amount that can be processed by the control device j, and the addition related to Σ _i Xij (Σ _i : i) is received by each of the plurality of control devices for a predetermined process of the type i. The required amount of the predetermined processing of the type i allocated to the control device j that maximizes or minimizes the objective function ΣAij * Xij (* is a multiplication symbol) under the constraint condition that is the sum of the requested amounts. The value of Xij is acquired, and the ratio of the acquired value of Xij to the sum of the requested amounts received by each of the plurality of control devices for a predetermined process of the type i is assigned to the plurality of control devices j. Determined as the weight used for sorting the requirements for the predetermined processing of the type i.
The information processing device according to claim 6. The receiving unit, from each of the plurality of control devices, further receives the respective requirements of the double several predetermined process each of the plurality of control apparatus has received,
The processing unit has the plurality of types based on the upper limit of the amount of demand that can be processed by each of the plurality of control devices and the total amount of requests to the plurality of control devices for each of the plurality of types of predetermined processing. For each of the predetermined processes of the above, the control device j (1) corresponding to the first ratio of the highest value among the first ratios for each of the plurality of types of predetermined processes executed by each of the plurality of control devices. The first process of assigning a request for a predetermined process of type i (1 ≤ i ≤ the number of types of predetermined processes) to (≦ j ≦ the number of control devices) is performed, and the control device j executes the first process of the type i. When the first ratio for a predetermined process is set to the first ratio Aij, the required amount assigned to the control device j is the required amount for the first ratio Aij in descending order of the first ratio. When the upper limit of the amount is reached, the first process is repeatedly executed for the remaining control devices excluding the control device j until the total number of requested amounts of each of the plurality of types of predetermined processes becomes 0. The ratio of the required amount of each of the plurality of types of predetermined processes assigned to each of the plurality of control devices is used to distribute the request to each of the plurality of control devices for each of the plurality of types of predetermined processes. Decide as,
The information processing device according to claim 6. For a control device whose determined weight is lower than a predetermined threshold value, the processing unit determines to stop a predetermined process corresponding to the determined weight by the control device corresponding to the determined weight.
The information processing device according to any one of claims 4 to 8. The processing unit determines an upper limit value of the required amount of each of the plurality of control devices as a value having a predetermined allowable range.
The information processing device according to any one of claims 3, 7, and 8. The receiving unit further receives from each of the plurality of control devices the requested amount of each of the plurality of types of predetermined processes received by each of the plurality of control devices.
The processing unit sets the upper limit of the required amount of each of the plurality of control devices and the sum of the requested amounts of each of the plurality of types of predetermined processes received by each of the plurality of control devices of the plurality of control devices. Divide by the number of units,
The information processing device according to any one of claims 3, 7, and 8. The processing unit determines an upper limit value of the required amount of each of the plurality of control devices according to the processing performance of each of the plurality of control devices.
The information processing device according to any one of claims 3, 7, and 8. The computer
A first receiving unit that receives a request distributed to its own device among the requests distributed among a plurality of control devices, and a predetermined process is executed according to the request distributed to the own device to generate a message. and, if the equipment of the destination of the message is a device of the control target of its own device transmits the message to the device of the destination equipment of the destination of the message of another control device
In the case of a device to be controlled, a control unit that transfers the message to the other control device, and a storage unit that stores information on the device to be controlled for each of the plurality of control devices.
For each of the plurality of control devices, the weights used for distribution of requests for the predetermined processing, demand of the predetermined processing is distributed number of that equipment to the control target to the many control device as the control device become so many, to determine the load sharing control method. The computer
A first receiving unit that receives a request distributed to its own device among the requests distributed among a plurality of control devices, and a predetermined process is executed according to the request distributed to the own device to generate a message. and, if the equipment of the destination of the message is a device of the control target of its own device transmits the message to the device of the destination equipment of the destination of the message control target device of another control device In the case of, a message generated by each of the plurality of control devices executing the predetermined process from each of the plurality of control devices including a control unit for transferring the message to the other control device. The amount of the first message transmitted from each of the plurality of control devices to each of the controlled devices of the plurality of control devices, and the second message transferred from each of the plurality of control devices to another control device. The amount of messages, and received,
A first ratio of the amount of the first message to the total amount of messages generated by the predetermined processing by each of the plurality of control devices, or a second of the amount of the second message. Based on the ratio, for each of the plurality of control devices, the control device having a higher first ratio or a lower second ratio requires a larger amount of the predetermined processing to be distributed to the control device. Determines the weights used to distribute the requirements for the predetermined process.
Load balancing control method. The processing unit
Select multiple instrumentation placed al one or more of equipment, each of the plurality of control apparatus is a control object,
When the control device for the selected equipment and the controlled object is changed to another control device, the amount of messages that are transferred between a plurality of control devices is reduced, and, from said each of the plurality of control devices the conditions under which the amount of messages sent to a plurality of equipment includes that uniform among the plurality of control devices to determine whether the selected equipment satisfies,
Determining that the selected equipment is if the condition is satisfied, it changes the control device to be controlled to the selected equipment to another control device,
The information processing device according to any one of claims 1 to 12. The processing unit
Wherein the plurality of instrumentation placed et al, to generate a plurality of case information indicating a case of selecting one or more equipment,
For each of the generated plurality of case information, the case of changing the controller to be controlled the selected equipment to another control device, to determine the weights, it determined the predetermined processing based on the weight was sorting and simulate the request, the amount of messages that are transferred between the plurality of control devices, said plurality of control of the amount of messages sent to a plurality of equipment to be controlled from each of the plurality of control devices Obtain the index value, which indicates the variation between devices, and
As the condition, one or one selected in the extracted case information is extracted from the case information in which the index value indicating the variation is equal to or less than a predetermined threshold value and the amount of messages transferred between the control devices is the smallest. It determines to change the control unit to be controlled a plurality of equipment to another control device,
The information processing device according to claim 15. Each of the plurality of control devices executes a plurality of types of predetermined processes to generate a message.
Wherein the processing unit, the plurality of types of the predetermined process, the first type message volume which is transmitted to a plurality of equipment of the controlled object from each of the plurality of control devices is equal among the plurality of control devices acquires, selecting the first type of acquired one or more equipment received message intensive upper predetermined processing of the message, said plurality of instrumentation placed et the acquired one or more equipment Generate case information
The information processing device according to claim 16.

Images