WO2004093394A1 - Data relaying device, data relaying method, data relaying program, service selecting device, service selecting method, and service selecting program - Google Patents

Abstract

Deterioration of data quality is effectively prevented by selecting information on the loads on servers and paths depending on the information. A data relaying device (1) for relaying a request for connection from a client to one of servers is provided between the client and the servers. The data relaying device (1) comprises information collecting means (2) for collecting load information on the servers and the paths from the data relaying device (1) to the servers and service selecting means (5) for selecting the path from at least one of the servers to the server in response to the connection request of the client according to the load information collected by the information collecting means (2).

Description

 Description Data relay device, data relay method, data relay program, service selection device, service selection method, and service selection program

 The present invention relates to a data relay device or the like applied to an information communication system comprising a combination of a server, a client, and a network, and more particularly to a data relay device provided between a plurality of servers and a client, from a client to one of the servers. The present invention relates to a data relay device, a method, and a program for relaying a connection request. In addition, the present invention relates to a service selection device, a method, and a program for selecting at least one server from a plurality of servers and a route to the server. Background art

 With the advancement of information communication and information services in recent years, in these information communication systems and information service systems, the same information is deployed on multiple servers to improve the processing performance of the information. Server load balancing control is performed to distribute connection requests (hereinafter referred to as information requests) accessed by an address (for example, a virtual IP (VIP) address) to one of these multiple servers. In this load balancing control, especially in a system where the system is large and the geographical distribution of user sites and server sites is expanding, efficient distribution between servers and server sites deployed over a wide area It is necessary to perform load distribution. Here, it is desirable to effectively use resources and maintain service quality by selecting and allocating servers and routes from multiple server and route candidates based on policies and load information collected periodically. ing.

 Such a request is desired for the following system as shown in FIG.

 (a) Corporate information system

This system is a system that connects information system sites that are widely distributed. You.

 (b) Distributed IDC (Internet Data Center)

 This system connects regional IDCs and connects regional IDCs to center IDCs.

 (c) Carrier system

In FIG. 20, the above-mentioned distribution to the wide-area site server is performed by the GW (Gateway; data relay device) in FIG. On the other hand, the following is known as a conventional technique for constructing the above system.

(A) Wide area load distribution technology This technology distributes the load between servers and server sites deployed in a wide area. In general, a DNS round-mouth bin method in which DNS responses are distributed by sequentially turning them among a plurality of servers, and a method in which server loads are periodically collected and distributed to the location with the smallest load are used.

 In the DNS round mouth bin method, the DNS response to the server is simply returned in a predetermined order without considering both server load and route load. In the server load method, the destination server is determined only by the server load without considering the path load. In addition, no route can be selected in either method, and only a single route uniquely determined by the IP routing described below is used.

 (B) Traffic '' Engineering

 This technology monitors the route load at the network level and, when it is determined that the traffic is congested, distributes some of the load to alternate routes and distributes the load between routes. This technology optimizes only the path load, and does not control the server status or individual flows.

 (ゥ) IP routing

In this technology, routing information is exchanged between routers, and only the route that minimizes a single evaluation criterion (such as the number of hops) is used continuously. Generally, consider multiple routes do not do.

 (E) Q O S P F

 This technology relates to a routing protocol based on multiple criteria for QoS (quality of service). Although a plurality of evaluation criteria for the network are used, which route is selected is determined based on whether or not each evaluation criterion satisfies a certain condition (is within a predetermined range). In general, the solution is not uniquely determined, and you choose something that is not so bad.

 However, the above conventional technique has the following problems. Technology that selects services based only on routes, such as traffic engineering in (a) above, IP routing in (ゥ), or server load information only, as in (a) server load method. In the selected technology, a service is selected based on only one of the path load and the server load, so that the load of the other one of the path load and the server load increases in the selected service. If so, the service quality will be degraded. For example, if a service (server and route) is selected based only on server load information, the service will continue to be used even if the route is congested. Lig will occur. Therefore, resources cannot be effectively used by a technology that selects a service based on only the load information of the route or the load information of the server. In (e), service selection based on multiple evaluation criteria in QOSPF involves selecting services based on whether or not the conditions are satisfied for each evaluation criterion. There are no guidelines. Therefore, the service cannot be uniquely determined because the combination of the evaluation criteria cannot be simply compared. In addition, since the judgment is made for each evaluation criterion, the branch judgment and the amount of calculation are increased, and it takes time to converge.

Furthermore, in the above-described conventional technology, there is a possibility that the collected load information and the current situation may be different. In general, the load information collection interval is long (several seconds), so the load state may have changed at the intermediate time until the next load information is obtained. Is not accurate load information. The present invention has been made in order to solve the above-described problems. The present invention collects load information of one or more resources related to a server and a route, and then composites the plurality of load information. A data relay device, a data relay method, and a data relay program capable of judging and selecting a service (a server and a route) to effectively use resources and maintain service quality, and a service selection device, a service selection method, and It aims to provide a service selection program. Disclosure of the invention

 In order to solve the above-described problem, a data relay device according to the present invention is a data relay device that is provided between a plurality of servers and a client, and relays a connection request from a client to any of the servers. An information collection unit that collects load information on each of the server and a route from the data relay device to the server; anda server that receives the load information from the client based on the load information collected by the information collection unit. It is provided with at least one server and a service selection unit for selecting a route to the server in response to a connection request.

 According to such a configuration, after collecting load information of one or more resources related to the server and the route, the service (server and route) is selected by judging the plurality of load information in a complex manner. Effective use of resources and maintenance of service quality can be achieved.

 Further, in the data relay device of the present invention, the service selection unit calculates a remaining capacity ratio for each server and each route based on the load information, and calculates a product operation value of the obtained remaining capacity ratios. Is calculated for a plurality of combinations of each server and each route, and the at least one server and a route to the server are selected based on the calculated product operation value.

According to such a configuration, in order to select a service (server and route), a single evaluation criterion value (preference value) can be derived from a plurality of evaluation criteria based on load information. Comparative evaluation can be easily performed. With this reference value, it is possible to select options based on resource utilization rates such as servers and routes, and services can be allocated so that the resource utilization rates are equal. In addition, the superiority of options when combining multiple evaluation criteria can be easily compared based on a single reference value.

 Also, in the data relay device of the present invention, the service selection unit acquires load information or a change in the remaining capacity ratio of the predetermined server and the path to the server when the unit number of connection requests is received. The remaining capacity ratio when there is a connection request from the client is calculated based on the change, the remaining capacity ratio based on the load information already acquired, and the change in the number of connection requests. It is characterized by

 According to such a configuration, it is possible to predict the load state of the server and the route between the points for collecting the load information, and to perform the service selection based on the predicted value, thereby providing a more accurate service. Realization of allocation enables effective use of resources and maintenance of service quality.

 In the data relay device of the present invention, the service selection unit calculates a difference between a product operation value of the remaining capacity ratio before and after accepting a connection request, and based on the difference value, at least one server and the server It is characterized by selecting the route to it.

 According to such a configuration, for example, it is possible to extract an option that has the least effect on the reduction of the remaining capacity ratio of all resources by allocating a service, that is, an option that has the most surplus resources. . This configuration does not operate to make the utilization rate of each resource uniform, but if there is a difference in the absolute capacity between each resource, the absolute surplus capacity is selected as the highest In the data relay device of the present invention, the service selection unit may be configured to: load information collected at a first point in time, the number of connection requests at that time, and a second Based on the difference between the load information collected at the time and the number of connection requests at that time, the load information or the remaining capacity ratio of the predetermined server and the route to the server when the unit number of connection requests are received. Is calculated and obtained.

According to such a configuration, for example, when a single connection request (information request) is received, a decrease in the remaining capacity ratio of each resource is determined by feedback control based on a measurement result. It is possible to calculate, and more strict predictive control can be realized as compared with the case where it is held at a fixed value.

 Further, in the data relay device of the present invention, when calculating the remaining capacity rate, the service selecting unit acquires a resource utilization rate associated with the load information, and defines the resource utilization rate as a parameter Xi. The resource utilization rate is weighted, and the remaining capacity rate is calculated by 11 F (xi). According to such a configuration, it is possible to determine an option by weighting the utilization rate of each resource, and to increase or decrease the degree of influence when the evaluation standard is compared with other evaluation standards be able to. As a result, instead of treating all evaluation criteria equally, it is possible to make the influence of evaluation criteria that are to be prioritized when judging an option larger or to make the influence of evaluation criteria that are not to be prioritized appear smaller. Become. In addition, the data relay device of the present invention monitors the product operation value of the remaining capacity ratio for the combination of the server and the route, and when the monitored value falls below a predetermined value, among the resources of the combination. It is characterized by determining that congestion has occurred and switching the communication connection path to another path.

 According to such a configuration, when a path load increases due to the congestion of an already established communication connection due to the influence of another communication connection, the path of the relevant communication connection is not congested. By switching to a route, service quality can be maintained. Such autonomous control can be easily performed based on a single reference value calculated from multiple pieces of load information (server information, route information).

 Further, in the data apparatus of the present invention, the service selecting unit further selects a server and a route to the server based on policy information defined for service allocation. .

 According to such a configuration, for example, it is possible to prioritize the policy information, or to prioritize the load information within a predetermined policy range, so that the resource can be selected flexibly. Can be done.

In the data relay device of the present invention, the load information includes at least one of a CPU load ratio, a disk usage ratio, and the number of connections for the server. And the route may include at least one of a used bandwidth and a route delay, and the server includes a server load aggregating unit that aggregates load information from a plurality of application servers. It can be characterized by being configured with an SLB server that has it.

 Further, the present invention is a data relay method provided between a plurality of servers and a client, wherein the data relay method relays a connection request from a client to any one of the servers. An information collection step of collecting load information on each of the routes to the server; and at least one connection request from the client based on the load information collected by the information collection step. And a service selection step of selecting a route to the server.

 Here, the service selecting step is a step of calculating a remaining capacity ratio for each server and each route based on the load information, and a product operation value of the obtained remaining capacity ratio is calculated for each server and each route. And a step of selecting the at least one server and a route to the server based on the calculated product operation value. In addition, the service selecting step includes a step of acquiring a change in load information or a remaining capacity ratio of a predetermined server and a path to the server when a unit number of connection requests are received; A step of calculating the remaining capacity ratio when there is a connection request from the client, based on the remaining capacity ratio based on the load information that has already been acquired and the change in the number of connection requests. It is a feature.

 Further, the service selecting step includes a step of calculating a difference between the product operation values of the remaining capacity ratio before and after accepting a connection request, and at least one server and a route to the server based on the difference value. And a selecting step.

Further, in the data relay method of the present invention, the step of obtaining the change in the remaining capacity ratio includes: collecting load information at a first time and the number of connection requests at that time; and loading information at a second time. And the number of connection requests at that time. A predetermined server based on the difference between the load information and the number of connection requests collected at the time point and the load information and the number of connection requests collected at the second time point, when the connection request of the unit number is received. And a step of calculating a change in load information or a remaining capacity ratio along a route to the server.

 Further, in the service selection step, when calculating the remaining capacity rate, a resource usage rate associated with the load information is acquired, and a function F having the resource usage rate as a parameter X i is introduced. Weighting the resource utilization rate and calculating a remaining capacity rate by using 11 F (X i).

 Further, in the data relay method of the present invention, a step of monitoring a product operation value of the remaining capacity ratio for a combination of a server and a route; and, when the monitored value falls below a predetermined value, a combination of the combination. And determining that congestion has occurred in the resource and switching the path of the communication connection to another path.

 Further, the service selecting step further comprises a step of selecting a server and a route to the server based on policy information defined for service allocation.

 Further, the present invention is a data relay program provided between a plurality of servers and a client, and causing a computer to relay a connection request from the client to any of the servers, wherein the server comprises: An information collection step of collecting load information relating to each of the path from the data relay device to the server; and, based on the load information collected by the information collection step, for the connection request from the client, The program causes a computer to execute at least one server and a service selection step of selecting a route to the server. The present invention also relates to a service selection device for selecting at least one server from among a plurality of servers and a route to the server, wherein each of the plurality of servers and a route to the plurality of servers are related to each other. An information collection unit for collecting load information, and at least one server and a service selection unit for selecting a route to the server based on the load information collected by the information collection unit.

The present invention also provides at least one server from among the plurality of servers and the server. A service selection method for selecting a route in the information processing method, wherein an information collection step of collecting load information on each of the plurality of servers and a route to the plurality of servers;

And at least one server based on the load information collected by the information collecting unit and a service selection step of selecting a route to the server.

 Further, the present invention is a service selection program for causing a computer to select at least one server from among a plurality of servers and a route to the server, wherein the plurality of servers are connected to the plurality of servers. An information collecting step of collecting load information relating to each of the paths, and a service selecting step of selecting at least one server and a path to the server based on the load information collected by the information collecting unit. Is executed. In the embodiment of the present invention, the data relay device constitutes the service selecting device of the present invention, and its operation method constitutes the service selecting method of the present invention, and a program for causing a computer to execute the operation thereof is provided. Configure the service selection program. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a flowchart showing the operation principle of the present invention.

FIG. 2 is a functional block diagram of the data relay device to which the present invention is applied.

FIG. 3 is a block diagram showing the overall configuration of the embodiment.

FIG. 4 is a diagram showing the relationship between the server information table and the route information table. FIG. 4 (a) shows the server information table, and FIG. 4 (b) shows the configuration of the route information table. .

FIG. 5 is a diagram showing an example of a server information table.

FIG. 6 is a diagram showing an example of a route information table corresponding to the server information table of FIG.

FIG. 7 is a diagram showing another example of the server information table.

FIG. 8 is a diagram showing an example of a route information table corresponding to the server information table of FIG. FIG. 9 is a flowchart showing a first operation of the service selection unit 5A shown in FIG.

FIG. 10 is a flowchart showing a second operation of the service selection unit different from the operation shown in FIG.

FIG. 11 is a flowchart showing a third operation of the service selection unit different from the operations shown in FIGS. 9 and 10.

FIG. 12 is a flowchart showing a fourth operation of the service selection unit different from the operations shown in FIGS. 9 to 11.

FIG. 13 shows a weighting function used for obtaining the above-mentioned resource utilization rate.

FIG. 14 is a flowchart showing a fifth operation of the service selection unit different from the operations shown in FIGS. 9 to 12.

FIG. 15 is an overall block diagram showing one configuration example of the present invention in more detail as an embodiment.

FIG. 16 is a diagram showing a bucket format applied to each route shown in FIG.

FIG. 17 is a diagram showing an internal functional configuration of the data relay device 1 of FIG. FIG. 18 is a first specific example of a server information table and a route information table corresponding to the configuration of FIG.

FIG. 19 is a diagram showing a second specific example of the server information table and the route information table corresponding to FIG.

FIG. 20 is a diagram showing a system to which the present invention is applied, FIG. 20 (a) is a diagram showing a corporate information system, and FIG. 20 (b) is a distributed IDC (Internet Data Center). FIG. 20 (c) is a diagram showing a carrier system. BEST MODE FOR CARRYING OUT THE INVENTION

 Hereinafter, embodiments of the present invention will be described with reference to the drawings.

First, the principle of the embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a flowchart showing the principle of operation of the present invention, and FIG. 2 is a data diagram to which the present invention is applied. The function block diagram of the data relay apparatus is shown. The data relay device 1 shown in FIG. 2 includes an information collecting means 2 for collecting server load information (hereinafter referred to as server information) and route load information (hereinafter referred to as route information), and stores the collected server load information. A server information database ₃ to store the collected route load information, a route information database 4 to store the collected route load information, a service selecting means 5 to select a server and a route as a service based on the collected or stored information, (Resource) control means 6 for converting the destination address of the server, selecting a router, specifying a network, etc. to execute the provided service, and a data transmitting / receiving unit for transmitting / receiving between a client or server (not shown) Has eight.

 The operation of the data relay device 1 is shown in FIG. That is, in the data relay device 1, the information collecting means 1 periodically collects server information and route information (S1), and stores them in the server information database 3 and the route information database 4 as appropriate. Next, the service selection means 5 obtains the resource utilization rate of each resource (server and route) from the collected information sources (S2), and in response to the information request, determines the possible route / server combination for each resource. The product operation value (preference value) of the remaining capacity ratio is calculated (S 3), and the combination of the server and the route having the maximum preference value is selected (S 4). Then, the resource control means 6 controls the resources so that the service in the combination is executed (S5), transmits the information from the data transmission / reception unit 8 to the server selected according to the selected route, and Receives information according to the route selected from the server.

Next, Embodiment 1 of the present invention will be described. Fig. 3 is a block diagram showing the overall configuration of this system. This system includes a user client 11, various servers 12, 13, a data relay device 1 provided between the user client 11 and various servers 12, 13, and a plurality of connecting devices. It is provided with paths 14, 15, and 16. The data relay device 1 includes an information collection unit 2A, a server information database 3A, a route information database 4A, a service selection unit 5A, a resource control unit 6A, and a service selection unit for selecting a service in response to an information request. A classifying unit 7A for performing classification, a policy information database 9A storing policy information for performing classification, and a data transmitting / receiving unit 8A. The information collection unit 2A periodically collects server information and route information. Examples of the server information include a CPU usage rate and a memory usage rate, and examples of the path information include information on a bandwidth usage rate and path delay.

 The service selection unit 5A is based on the policy information (definition policy) stored in the policy information database or input as appropriate, and the status information (server information, route information) collected by the information collection unit 2A. Select an appropriate service (server, route) to be assigned to the information request (connection request) from the user client 11.

 The resource control unit 6A performs control (selection control) for selecting a server and a route in order to implement the selected (determined) service.

 In route selection control, label assignment and output interface selection for dynamically associating the received information request with the route of V LAN or IP-VP N are performed.

 In the server selection control, a virtual IP is converted into a real IP corresponding to each server, and the server information database 3A and the route information database 4A store server information and route information as an information table. These information tables are used to manage the server information of the distribution destination for the information request, and the route information to the server (if there are multiple routes to each server, multiple corresponding to them). The information collection unit 2A collects each information, writes the information in the information tables of these databases 3A and 4A, and refers to the information when the service selection unit 5A selects (determines) a service applicable to the information request.

 FIG. 4 is a diagram showing the relationship between the server information table and the route information table. FIG. 4 (a) shows the server information table, and FIG. 4 (b) shows the configuration of the route information table. FIG. 5 shows an example of a server information table, and FIG. 6 shows an example of a path information table corresponding to the server information table of FIG. FIG. 7 shows another example of the server information table, and FIG. 8 shows an example of the path information table corresponding to the server information table of FIG.

These server information tables exist for each request information (for example, for each virtual IP address corresponding to the request information), and hold a list of servers having the requested information and load information of the server. In Fig. 5, the CPU load rate 51, disk usage rate 52, This indicates that three types of information with 53 connections are collected.

 The route information table is pointed from each entry of the server information table as shown in FIG. 4 (b), and holds a list of routes to the server and load information of the route. FIG. 6 showing the route information table indicates that two types of information, that is, the used bandwidth 61 and the delay 62 are collected. In this example, multiple routes to the server are checked using the BGP (Border Gateway Protocol) route search.

 The difference between the server information table and the route information table shown in FIGS. 7 and 8 and the server information table and the route information table shown in FIGS. 5 and 6, respectively, is the CPU load. As information on the rate and disk usage, there is a point where the increment per unit request is a constant value or a rewritable value, and the connection information has load information as a change in the number of information requests since the most recent information collection. Is a point. That is, in the server information table shown in Fig. 7, the CPU load factor increment 51a, the disk usage factor increment 52a, and the change in the number of information requests since the most recent information collection were used as variable increments. The route information table shown in Fig. 8 has three types of information, 5 3a.In the routing information table shown in Fig. 8, the increase in bandwidth used 61a, the increase in route delay 62a, and the number of information requests since the latest information collection. It has three types of information, 6 3 a.

 (First operation)

FIG. 9 is a flowchart showing a first operation of the service selection unit 5A shown in FIG. In this example, the operation is performed using the server information table and the route information table shown in FIGS. Each time the data transmission / reception unit 8A receives an information request from the user client 11, the service selection unit 5A searches the server information table and the route information table for the information (S11). For each possible combination of server and route, the load information contained in each table is converted into a normalized utilization rate of the resource that takes a value between 0 and 1 (S12). For the server load information shown in Fig. 5 and the route load information shown in Fig. 6, the normalized resource utilization rate (resource utilization rate) is defined and calculated as follows. That is, when calculating the normalized usage rate for the server load information, the CPU load rate and the disk usage rate are used as they are as the normalized usage rate. The number of connections is obtained by dividing the measured number of connections by the maximum number of allowed connections. However, the measured connection If the number exceeds the maximum number of allowed connections, set to 1.

 Also, when calculating the normalized utilization rate for the path load information, the bandwidth used is obtained by dividing the measured bandwidth by the link speed, and the path delay is obtained by dividing the measured path delay by the maximum allowable path delay. It is required by However, if the measured path delay exceeds the maximum allowable path delay, it is set to 1.

 Server load information Normalized usage rate

 C P U load factor> Use as it is

 Disk usage> Use as is

 Number of connections Number of measured connections / Maximum allowable connections

 (However, if the measured number of connections exceeds the maximum allowable number of connections, this value is set to 1.) Route load information: Normalized utilization rate

 Bandwidth used → Measured bandwidth used / link speed

 → Measured path delay / Maximum allowable path delay

 (However, if the measured path delay exceeds the maximum allowable path delay, this value is set to 1.) Then, for each combination of server and path, the remaining capacity ratio of each resource is calculated as (1-1 normalized utilization rate) Is calculated. Next, the preference value P n for this combination is calculated as the product of the remaining capacity ratios of the full load information. Assuming that the normalized utilization rate of the load information i is X i, the preference value P n is represented by the following equation.

P n = (1-1) (1-X 2) · · · (1-X k) where X i is the normalized utilization rate of load information i. Then, Pn is compared for all combinations of (server, route), and the combination with the largest value is selected as the service to be assigned to the information request, and in order to control the actual resources, the combination shown in FIG. Start the resource control unit 6 A. If no combination satisfies the condition, the information request is rejected because there is no suitable resource. (Second operation)

 FIG. 10 is a flowchart showing a second operation of the service selection unit different from the operation shown in FIG. In this example, the operation is performed using the server information table shown in FIG. 7 and the route information table shown in FIG.

In this case, the service selection unit 5A searches the server information table (FIG. 7) and the route information table (FIG. 8) for the information every time an information request from the client is received (S21). For each possible combination of route and route, the predicted values of normalized resource utilization (values between 0 and 1) for the load information (obtained at the time of the most recent information collection) included in each table were collected most recently. It is calculated from the load information, the load increment per unit information, and the increment in the number of information requests (S22). Assuming that these values for the load information i are _X i, S xi, and m, the predicted value X i ′ of the normalized utilization rate of this load information including the new information request is expressed by the following equation. . In this example, the increment (δ X i) of each piece of load information per unit information request is calculated in advance as a constant value by actual measurement or the like, and the server information table and the route information table shown in FIGS. It shall be registered in.

X i = xi + (m + 1) X δ xi Using the value of the above equation, for each combination of (server, route), calculate the remaining capacity ratio of each resource by (1 _ predicted value of normalized utilization rate). Then, the preference value Pn 'for this combination is calculated as the product of the remaining capacity ratios of the full load information (S23). Assuming that the normalized utilization of the load information i is X i, the predicted value Pn ′ of the preference value is expressed by the following equation.

P n '= (1-X 1') (1-X 2 ') · · · (1-X k') where X i 'is a predicted value of the normalized utilization of the load information i. Then, P n ′ is compared for all combinations of (server, route), and the combination having the maximum value is selected as a service to be assigned to the information request (S 24). Then, in order to control the actual resources, the resource control unit 6A in FIG. 3 is started (S25Y, S26). If no combination satisfies the condition, The information request is rejected because there is no critical resource (S25N, S27). (Third operation)

 FIG. 11 is a flowchart showing a third operation of the service selecting unit different from the operations shown in FIGS. 9 and 10. In this example, the operation is performed using the server information table shown in FIG. 7 and the route information table shown in FIG.

In this case, each time the service selection unit 5A receives an information request from the user client, the service selection unit 5A searches the server information tape shown in FIG. 7 for the information and the route information table shown in FIG. Yes (S31). Thereafter, as described in the second operation (Fig. 10), all (server route) of the combination, the prediction value P _t of the preference value, including the new information request (= Pn ') It is calculated (S32). Next, a predicted value Pt-i of a preference value not including a new information request is calculated.

(S33). From these, all (server route) for the combination of a decrease in the estimated value of Purifu Arens value, P _t - - calculating a P _t (S 34). Then, the combination with the smallest difference value is selected as a service to be assigned to the information request (S35), and the resource control unit 6A in FIG. Apply (S 36Y, S 37). If there is no combination that satisfies the condition, the information request is rejected because there is no appropriate resource (S36N, S38).

 (4th operation)

 FIG. 12 is a flowchart showing a fourth operation of the service selection unit different from the operations shown in FIGS. 9 to 11. In this example, the operation is performed using the server information table shown in FIG. 7 and the route information table shown in FIG.

 In the fourth operation, the increment (δ χ ί) per unit information request (connection request) of each load information is given as a constant value in advance as described in the second operation (FIG. 10). Instead, it specifies a procedure for adaptively correcting based on load information collected periodically.

In this operation, each time new load information is collected, the server information table shown in FIG. 7 and the route information table shown in FIG. 8 for the information are searched (S41). Next, the remaining capacity ratio of each load information since the last load information collection The increase / decrease in the number of information requests is calculated (S42). Based on this value and the increment of the number of information requests, the increase / decrease of the remaining capacity ratio of each load information per unit request is calculated (S43). For example, for a given load information i, the load information value collected last time, the load information value collected this time, and the increment of the number of information requests are X (i-1), (i), and n, respectively. δ X i is represented by the following equation. ό i = vx (i-x (i-1)) / n The value calculated by the above equation is recorded in the server information table and the route information table (S44). This value is used as the load information increment.

 FIG. 13 is a diagram showing a weighting function used when obtaining the above-mentioned resource utilization rate. In other words, the curve shown here represents a function formula for calculating the value used in calculating the preference value instead of the actual resource utilization rate. After the load information value X i is converted by the weighting function y = F i (x), the preference value for the service combination is calculated. That is, the preference value Pn is represented by the following equation.

 Pn = (1 -F 1 (1)) (1 -F 2 (x 2)) (1-Fk (x k))

 Here, x i is the normalized utilization rate of load information i.

FIG. 13 shows an example of the weighting function F. Normally, use y = X (that is, leave the normalized utilization unchanged). When it is desired to greatly reflect the degree of influence of the load information, the function F can be realized by using an upwardly convex function y = X1 ^{/ 2} . When it is desired to reflect the degree of influence of the load information to a small extent, it can be realized by using a downwardly convex function y = X ² as the function F.

 (Fifth operation)

FIG. 14 is a flowchart showing a fifth operation of the service selection unit different from the operations shown in FIGS. 9 to 12. In this example, the operation is performed using the server information and route information table shown in FIGS. 5 and 6, or the server information table and route information table shown in FIGS. 7 and 8. In this case, each time the service selection unit 5A collects new load information, The server information table and the route information table for are searched (S51). Next, for the possible combination of the server and the route, the normalized utilization rate for the newly collected load information is calculated by the procedure shown in the operation of FIG. 1 described above (S52). Next, the remaining capacity ratio is calculated, and the preference value for this combination is calculated (S53). Then, it is determined whether or not the preference value is equal to or less than a predefined value (S54). If the preference value is equal to or less than the defined value (S54Y), it is determined that congestion occurs somewhere in the service. Then, a part of the route control of the already set connection is switched to another route (alternate route) to the same server (re-routing) (S55). Although this route control method is not described in detail in this patent, it is not necessary to change or select a label assigned for dynamically associating the information request with a VLAN or an IP-VPN route. This is realized by changing the output interface.

 (Example)

FIG. 15 is an overall block diagram showing one configuration example of the present invention in more detail as an embodiment. As shown in FIG. 15, an access network 100 (for example, a regional base network) accommodating the user client 11 is connected to the Internet 200 via the data relay device 1. On the other hand, the information system sites 300 and 400 where information is stored so as to be provided are connected to the Internet 200 via SLB (Session Load Balancer; server load balancer) servers 312 and 413, and internally, Distribute server load to application servers 301 to 30 m. The information providing unit in which the information is stored is virtualized in two stages: the data relay device 1 and the SLB servers 312 and 413. In FIG. 15, each IP address is indicated by # 1 to # 17. Then, the packet sent from the user client 11 to the VIP (virtual IP) # 2 (192.168.1.1) is converted by the data relay device 1 to an IP address for distribution, and the distribution destination SLB It is converted to the server's VIP # 6 (192.168.3.5 in the figure, SLB server 312 192.168.3.5), routed, and sent to SLB server 312. In response to this, the SLB server 312 converts the IP address for distribution from VIP # 6 (192.168.3.5), and the real IP of the distribution destination application server (in the figure, the application server 301 # 7 (192.168.5. 7)) and sent to the application server 301. The route is the next hop This is controlled by sending a packet by specifying the IP. There are two routes from the data relay device 1 to the SL B server 312, one via the router 201 (# 4: 192.168.2.2) and one via the router 202 (# 13: 192.168.2.3). By transmitting packets from the data relay device to # 4 (192.168.2.2) or # 13 (192.168.2.3) as the next hop, the route to be passed is identified. The route to the 35 server 413 (# 17: 192.168.4.6) is only via the router 203 (# 15: 192.168.2.4).

 Regarding collection of server load information, application server 301 (# 7: 192.168.5.7) to 3 Om (# 8: 192.168.5.8), 401 (# 10: 192.168.6.9) to 40n (# 11: 192.168.5.7) In 6.10), load measurement agents 301-a to 30m-a and 401-a to 40na that periodically measure the CPU load rate and disk usage rate will be provided. These load measurement agents include SNMP (Simple Network Management Protocol), which is a standard for network management, and UN

1 Implemented with a program that reads internal registers using s ar, the standard tool of the X-system OS. The application server periodically reports data measured by each application server to the SLB server. In response to this, the SLB server aggregates the load information obtained from the multiple application servers into information that indicates the entire site (by means of server load aggregation), and periodically sends it to the data relay device as load information that represents the entire site. Report appropriately. As a method of aggregation, a simple average is used for CPU load and disk usage, and a total is used for the number of connections.

C P U load factor → Simple average

 Disk usage → Simple average

 Number of connections → total The data relay device 1 has a server load collection unit 2A-1 and a route load measurement unit 2A-2 as the information collection unit 2A. For server load information, the server load collection unit

2A-1 collects information (aggregated load information) collected by the server load collecting units 312a and 413a of the respective SLB servers 312 and 413 as load information of the server sites 300 and 400. As for the path load, the path load measurement unit 2A_2 has a test packet generation unit 2A-2a and a response observation unit 2A-2b, and the test bucket generation unit 2A-2a determines the path load. Specify (specify the next hop IP in Fig. 15), send a test packet (typically an I CMP ECHO Request bucket), and respond

The response observing unit 2A-2b observes (delay and throughput), and calculates the bandwidth and delay of the specified route.

 FIG. 16 is a diagram showing a bucket format applied to each route shown in FIG. They are defined by destination IP, source IP, and next hop IP, respectively. The next hop IP is the IP of the device that throws the bucket directly. The next pop IP is used to send a MAC bucket to the MAC address for the next pop IP in the lower MAC (Media Access Cntrl) layer of the IP. This bucket format is used for the data exchange described in the above explanation. In FIG. 16, FIG. 16 (a) shows the case from the user client to the data relay device, FIG. 16 (b) shows the case from the data relay device to the router 201, and FIG. (C) shows the case from the router 201 to the SLB server 312, and FIG. 16 (d) shows the case from the SLB server 312 to the application server 301.

FIG. 17 is a diagram showing a functional configuration inside the data relay device 1 of FIG. The input packet is classified into a plurality of classes (packet 'queue) according to a classification policy defined in advance by the classification unit 7A. The service selection unit 5A selects a service (determines a server and a route) for each class of packets in accordance with a predefined service allocation policy and collected load information (server load and route load). The resource control unit 6A has a server control unit 6A_1 and a route control unit 6A_2, and the server control unit 6A-1 sends the transmitted destination IP to the destination IP for the determined server. Converted and passed to NAT (functional unit) 8A-1 of data transmission / reception unit 8A. The route control unit 6A_2 assigns the next hop IP corresponding to the route determined in the route selection and passes it to the routing 'stack (functional unit) 8A-2. Functional units such as NAT 8A-1 and Routing 'Stack 8A-2 perform specified operations. FIG. 18 is a first actual configuration example of a server information table and a route information table corresponding to the configuration of FIG. These correspond to the formats in Figures 5 and 6, respectively.

 FIG. 19 is a diagram showing a second actual configuration example of the server information table and the route information table corresponding to FIG. These correspond to Figure 7 and Figure 1 in Figure 8, respectively.

 Hereinafter, the operation of the configuration shown in FIG. 15 will be described in association with the first to fifth operations shown in FIGS. 9 to 12 and FIG.

 (First action)

 First, an operation (operation 1 of the embodiment) corresponding to the first operation described with reference to FIG. 9 will be described.

 An information request is sent to the data relay device in the format shown in Fig. 16 (a). The service selector searches the server information table and route information table shown in FIG. 18 for this information (# 2 (192.168.1.1) as VIP).

 From this, the combination of (server, route) for the information [hereafter, this is indicated by (SvrEntryID, RtEntrylD) in the table]

 (1, 1), (1, 2), (2, 3)

 It turns out that there are three. Here, “1” indicating the server indicates the SLB server 3 1 2 (# 6: 192.168.3.5), and “2” indicating the server is SLB f—4 1 3 (# 17: 192.168.4.6) "1" indicating a route indicates path _1 (router 201: # 4), "2" indicating a route indicates path-2 (router 202: # 13), "3 J indicating a route indicates path 1 (router 203: # 15)".

 For these three combinations, CPU load ratio (CpuLoad) / disk usage ratio (DiskLoad) / number of connections (ConnNum) are used as server load information, and used bandwidth (Bandwidth) / route delay (RtDelay) are used as route load information. ) Are the parameters considered in the service decision, and the normalized utilization rate of each resource is calculated as follows. The maximum allowable number of connections is 30,000 and the maximum allowable delay is 100 ms.

(For example, the CPU load rate is calculated as CpuLoad and the number of connections is calculated by ConnNu ra / (maximum allowable number of connections)) Calculation of normalized utilization

 Resource combination CPU load rate Disk usage rate Number of connectors a Bandwidth used Path delay (1, 1) 0.6 0.5 0.33 0.5 0.1

(1, 2) 0.6 0.5 0.33 0.1 0.001

(2, 3) 0.8 0.5 0.68 0.1 0.01 From the above normalized utilization rates, calculate the remaining capacity rate for each resource and the product operation value (= preference value) of all items for each combination as follows. The remaining capacity rate of each resource is 11 (normalized utilization rate), and the preference value is the product of the remaining capacity rates of all items in each combination.

 Calculation of the remaining capacity ratio preference value

Resource CPU load rate Disk usage rate Number of connections Bandwidth used Route delay Reference value combination

 (1, 1) 0.4 0.5 0.67 0.5 0.9 → 0.06030

(1, 2) 0.4 0.5 0.67 0.9 0.999 → 0.12048

(2, 3) 0.2 0.5 0.32 0.9 0.99 = 0.02851 Based on the above calculation result, select the combination (1, 2) that maximizes the preference value. Since this preference value is not 0, list it as a valid choice and pass it to resource control.

 (2nd action)

 Next, an operation (operation 2 of the embodiment) corresponding to the second operation described with reference to FIG. 10 will be described.

 An information request is sent to the data relay device in the format shown in Fig. 16 (a). The service selection unit searches the server information table and the route information table shown in Fig. 19 for this information (# 2: 192.168.1.1 as VIP).

 From this, as a combination of (server, route) for the information,

 (1, 1), (1, 2), (2, 3)

It turns out that there are three. For these three combinations, CPU load ratio (CpuLoad) / disk usage ratio (DiskLoad) / connection number (ConnNum) are used as server load information, and used bandwidth (Bandwidth) / route delay are used as route load information. (RtDelay) 2 One of them is a parameter to be considered in the service decision, and the predicted value of the normalized utilization rate of each resource (including the new information request) is calculated as follows. The maximum allowable number of connections is 30,000 and the maximum allowable delay is 100 ms (for example, the CPU load factor is calculated by CpuLoad + δCpuLoadX (ConnNum + ReqNum + l)).

 Calculating the predicted value of normalized utilization

 Resource combination CPU load rate Disk usage rate Number of connections 3 Bandwidth used Path delay

 (1, 1) 0.6402 0.5804 0.34 0.601 0.1101

(1, 2) 0.6402 0.5804 0.34 0.10202 0.0111

(2, 3) 0.8402 0.5804 0.6907 0.1201 0.0301 Based on the above-mentioned normalized utilization forecast, the predicted value of the remaining capacity ratio for each resource and the product operation value (= preference value) of all items are calculated as follows for each combination. Is calculated as follows.

 Calculation of estimated value of remaining capacity ratio and preference value Resource CPU load factor Disk usage rate Number of connections Bandwidth used Route delay Reference value Combination

 (1, 1) 0.3598 0.4196 0.66 0.399 0.8899 → 0.03538

(1, 2) 0.3598 0.4196 0.66 0.89798 0.9889 → 0.08848

(2, 3) 0.1598 0.4196 0.3093 0.8799 0.9699 → 0.01770 From the above calculation result, select the combination of (1, 2) that maximizes the preference value. Since this preference value is not 0, list it as a valid choice and pass it to resource control.

 (Third corresponding operation)

 Next, an operation corresponding to the third operation described with reference to FIG. 11 (operation 3 of the embodiment) will be described.

 An information request is sent to the data relay device in the format shown in FIG. 16 (a). The service selection unit searches this information (# 2: 192.168.1.1 as VIP) from the server information table and route information tape shown in Fig. 19.

 From this, as a combination of (server, route) for the information,

 (1, 1), (1, 2), (2, 3)

It turns out that there are three. For these three combinations, server load information CPU load rate (CpuLoad) / disk usage rate (DiskLoad) / number of connections (ConnNum), and bandwidth usage (Bandwidth) / path delay (RtDelay) as path load information and Barameta, the prediction value P _t of Preferences values including new information request, calculates as the procedure determined in the second corresponding operation. The maximum allowable number of connections is 30,000 and the maximum allowable delay is 100 ms.

 Similarly, the predicted value Pt-i of the preference value not including the new information request is calculated. From the above, Pw-Pt is calculated.

 The above data is described below for each combination.

 Predicted value and difference value of preference value

 Item combination Pt Pt-i Pt-i-Pt

 (1, 1) 0.03538 0.03552 0.00014

 (1, 2) 0.08848 0.08863 0.00015

 (2, 3) 0.01770 0.01774 0.00004

From the above calculation results, selects a combination of卩one P _t is the minimum (2, 3). Since this preference value is not 0, list it as a valid choice and pass it to resource control.

 (4th action)

 Next, an operation corresponding to the fourth operation described with reference to FIG. 12 (operation 4 of the embodiment) will be described.

 When the load information is collected periodically, the server information table and the routing information table shown in Fig. 19 are searched.

 Here, it is assumed that the newly collected load information is as follows. Server load

SvrEntrylD CpuLoad '(%) iskLoad' (%) ConnNum '

1 70 55 12000

2 82 55 21000 Route load of route (path'path-1)

RtEntrylD Bandwidth 'RtDelay' RtConnN m '6Mbps 12ms 2000 110Mbps 0.2ms 10000

Route load of route (path-3) RtEntrylD Bandwidth 'RtDelay' RtConnNum '

3 12Mbps 1.2ms 21000

 From this, the increment 1 (for example, if the CPU load is CpuLoad '= CpuLoad'-CpuLoad) from the load information (the value in Fig. 19) at the time of the previous load collection is as follows.

 Server load>

 SvrEntrylD dCpuLoad '(%) Z] DiskLoad' (%) ^ ConnNum '

1 10 5 2000.

2 10 5 479

<Route load of path (path 1 and path 1 2)>

 RtEntrylD Bandwidth 'RtDelay' ZlRtConnNum '

1 1Mbps 2ms 1000

 2 10Mbps 0.1ms 1000

<Route (path 1-3) path load>

RtEntrylD ^ Bandwidth 'lRtDelay' ZlRtConnNum '3 2Mbps 0.2ms 479 From this, the increase δ of each load information per unit request (for example, if CPU load, δ Cp uLoad' = ^ CpuLoad '/ ^ ConnNum') is as follows: Become. <Server load>

 SvrEntrylD δ CpuLoad '(%) Δ diskLoad' (%)

 1 0.005 0.0025

 2 0.021 0.0104

<Route load of path (path 1 and path 1 2)>

RtEntrylD δ Bandwidth 'δ RtDelay'

 1 0.001Mbps 0.002ms

 2 0.01Mbps 0.0001ms Route load of route (path-1 3)>

 RtEntrylD δ Bandwidth 'δ RtDelay'

 3 0.0042 Mbps 0.00042 These δ values are used as new load information increments and written into 5 CpuLoad, δ DiskLoad, δ Bandwidth, and δ RtDelay of the server information table and the route information table shown in FIG.

 (Fifth response operation)

 Next, an operation corresponding to the fifth operation described with reference to FIG. 14 (Example operation 5) will be described.

 When the load information is collected periodically, the server information table and the routing information table shown in Fig. 19 are searched.

 For each combination of (server, route), the predicted value of the regular usage rate of each resource is calculated by the procedure described in the operation 2 of the embodiment (however, there is no new information request).

Similarly, according to the procedure described in the operation 1 of the embodiment, the remaining capacity ratio is obtained from the normalized utilization rate, and the product operation value (= preference value) of all items is calculated for each combination. The calculated results are described below. Calculation of preference value>

 Combination one. Reference value

 (1, 1) 0.03552

 (1, 2) 0.08863

 (2, 3) 0.01774 Here, the threshold value (the defined value in step S54 in FIG. 14) for adjusting the resources of the route is set to 0.08. The meaning of the threshold value is a value for judging that a congestion state has occurred somewhere in the service of this combination if the preference value falls below this value (threshold value).

 When comparing the above preference value with this threshold value, when there are two routes (1, 1) and (1, 2), the preference value of (1, 1) falls below the defined threshold value. I have. In other words, it is determined that it is necessary to reflect the route load of the route (path 1) to the server 312 to the route (path 1) to the server 312.

 (Also, for (2, 3), the preference value is below the defined threshold value, but there is only one route, so the specified route switching process is not performed.)

 In the following, the operation for resource (resource) control for this is performed. The specified route is switched to the route (path-2) in session units. Industrial potential

 As described above, according to the present invention, these can be selected according to the load information of the server and the route, and deterioration of data quality can be effectively prevented. In addition, it is possible to derive a single evaluation criterion value from multiple evaluation criteria based on load information in order to select a service (server and route), making it easier to compare and evaluate options. become. Also, based on the collected load information and resource utilization prediction values, it is possible to predict the next state and select an appropriate server and route, thereby enabling predictive control. Therefore, these enable effective use of resources and maintenance of service quality.

Claims

The scope of the claims

1. A data relay device that is provided between a plurality of servers and a client and relays a connection request from a client to any of the servers,

 An information collection unit that collects load information on each of the server and a route from the data relay device to the server;

 A service selection unit that selects at least one server and a route to the server in response to the connection request from the client based on the load information collected by the information collection unit;

 A data relay device comprising:

2. In the data relay device according to claim 1,

 The service selection unit calculates the remaining capacity ratio of each server and each route based on the load information, and calculates a product operation value of the obtained remaining capacity ratio with each server. A data relay device that calculates a plurality of combinations with each route, and selects the at least one server and a route to the server based on the calculated integrated operation value.

3. In the data relay device according to claim 2,

 The service selection unit obtains load information or a change in the remaining capacity ratio for a predetermined server and a path to the server when a unit number of connection requests are received,

 Based on the change, the remaining capacity ratio based on the load information that has already been acquired, and the change in the number of connection requests, calculate the remaining capacity ratio when there is a connection request from the client. A data relay device.

4. The data relay device according to claim 3,

The service selection unit calculates a difference between a product operation value of the remaining capacity ratio before and after accepting a connection request, and based on the difference value, at least one server and the service A data relay device for selecting a route to a server.

5. The data relay device according to claim 3,

 The service selection unit is configured to calculate the unit number based on a difference between the load information collected at the first time and the number of connection requests at that time and the load information collected at the second time and the number of connection requests at that time. A data relay apparatus for calculating and acquiring load information or a change in a remaining capacity ratio for a predetermined server and a path to the server when the connection request is received.

6. The data relay device according to claim 2,

 When calculating the remaining capacity rate, the service selection unit acquires a resource utilization rate associated with the load information, and introduces a function F that uses the resource utilization rate as a parameter Xi to perform the resource utilization. A data relay device characterized in that the ratio is weighted and the remaining capacity ratio is calculated by 11F (xi).

7. The data relay device according to claim 1,

 The product operation value of the remaining capacity ratio for the combination of the server and the route is monitored, and if the monitored value falls below a predetermined value, it is determined that congestion has occurred in the resources of the combination, and another A data relay device for switching a communication connection path to a path.

8. In the data relay device according to claim 1,

 The service selecting unit further selects a server and a route to the server based on policy information defined for service allocation.

9. In the data relay device according to claim 1,

The load information includes at least one of a CPU load ratio, a disk usage ratio, and the number of connections for a server. A data relay device comprising at least one of the extensions.

10. The data relay device according to claim 1,

 The data relay device, wherein the server is configured by an SLB server having a server load aggregation unit that aggregates load information from a plurality of application servers.

11. A data relay method provided between a plurality of servers and a client, wherein the client relays a connection request from the client to one of the servers,

 An information collection step of collecting load information on each of the server and a route from the data relay device to the server;

 A service selection step of selecting at least one server and a route to the server in response to the connection request from the client based on the load information collected by the information collection step;

 A data relay method comprising:

1 2. In the data relay method according to claim 11,

 The service selection step includes: calculating a remaining capacity ratio for each server and each route based on the load information; and calculating a product operation value of the obtained remaining capacity ratio for each server. And calculating a plurality of combinations of the at least one server and each route, and selecting the at least one server and a route to the server based on the calculated product operation value. Relay method.

1 3. In the data relay method according to claim 12,

 The service selecting step includes a step of obtaining load information or a change in a remaining capacity ratio for a predetermined server and a path to the server when a unit number of connection requests are received;

Based on the change, the remaining capacity ratio based on the load information that has already been acquired, and the change in the number of connection requests, calculate the remaining capacity ratio when there is a connection request from the client. Performing the data relaying method.

14. In the data relay method described in claim 13,

 The service selecting step includes a step of calculating a difference between the product operation values of the remaining capacity ratios before and after accepting a connection request, and a step of selecting at least one server and a route to the server based on the difference value. A data relay method comprising:

15. The data relay method according to claim 13, wherein:

 The step of obtaining the change in the remaining capacity ratio includes: collecting the load information and the number of connection requests at that time at a first time; and collecting the load information and the number of connection requests at that time at a second time. Receiving the connection request of the unit number based on the difference between the load information and the number of connection requests collected at the first time and the load information and the number of connection requests collected at the second time. A data relay method comprising: calculating a change in load information or a remaining capacity ratio for a predetermined server and a route to the server in the above case.

16. In the data relay method according to claim 12,

 In the service selection step, when calculating the remaining capacity rate, a resource usage rate associated with the load information is acquired, and a function F having the resource usage rate as a parameter Xi is introduced to A data relay method comprising a step of weighting a rate and calculating a remaining capacity rate by 1−F (xi).

17. In the data relay method described in claim 11,

 Monitoring the product operation value of the remaining capacity ratio for the combination of the server and the route; and when the monitored value falls below a predetermined value, it is determined that congestion has occurred in the resources of the combination. And a step of switching a communication connection path to another path.

18. In the data relay method according to claim 11,

The service selection step further comprises a policy defined for service allocation. A data relay method, comprising: a step of selecting a server and a route to the server based on system information.

19. A data relay program, provided between a plurality of servers and a client, for causing a computer to relay a connection request from the client to the server or to the server,

 An information collection step of collecting load information on each of the server and a route from the data relay device to the server;

 A service selection step of selecting at least one server and a route to the server in response to the connection request from the client based on the load information collected in the information collection step;

 Data relay program that causes a computer to execute

20. In the data relay program according to claim 19,

 The service selecting step includes: calculating a remaining capacity ratio for each server and each route based on the load information; and calculating a product operation value of the obtained remaining capacity ratio for each server and each route. Characterized by causing a computer to execute, based on the calculated product operation value, a step of calculating a plurality of combinations of the above, and a step of selecting the at least one server and a route to the server. A relay program.

21. In the data relay program according to claim 20,

 The service selection step includes a step of obtaining load information or a change in a remaining capacity ratio for a predetermined server and a path to the server when a unit number of connection requests are received;

Based on the change, the remaining capacity ratio based on the load information that has already been acquired, and the change in the number of connection requests, the remaining capacity ratio is calculated when there is a connection request from the client. A data relay program characterized by causing a computer to execute the steps.

22. In the data relay program according to claim 21,

 The service selecting step includes a step of calculating a difference between the product operation values of the remaining capacity ratios before and after accepting a connection request; and a step of selecting at least one server and a route to the server based on the difference value. A data relay program characterized by causing a computer to execute the following.

23. In the data relay program according to claim 21,

 The step of acquiring the change in the remaining capacity ratio includes a step of collecting load information and the number of connection requests at that time at a first time, and a step of collecting load information and the number of connection requests at that time at a second time. Based on the difference between the load information and the number of connection requests collected at the first time and the load information and the number of connection requests collected at the second time, A data relay program for causing a computer to execute a step of calculating a change in load information or a remaining capacity ratio for a predetermined server and a path to the server when the server is received.

24. In the data relay program according to claim 20,

 In the service selection step, when calculating the remaining capacity rate, a resource usage rate associated with the load information is obtained, and a function F is set to use the resource usage rate as a parameter X i to introduce the resource usage rate. A data relay program characterized by causing a computer to execute a step of weighting a rate and calculating a remaining capacity rate by 1 _ F (xi).

25. In the data relay program according to claim 19,

Monitoring the product operation value of the remaining capacity ratio for the combination of the server and the route; and determining that congestion has occurred in the resources of the combination when the monitored value falls below a predetermined value. Switching a communication connection path to another path by a computer.

26. In the data relay program according to claim 19,

 The service selecting step further includes a step of selecting a server and a route to the server based on policy information prescribed for service allocation, and causing the computer to execute the step. Data relay program.

27. A service selection device for selecting at least one server from a plurality of servers and a route to the server,

 An information collection unit that collects load information on each of the plurality of servers and a route to the plurality of servers;

 A service selection unit that selects at least one server and a route to the server based on the load information collected by the information collection unit;

 Service selection device comprising:

28. A service selection method for selecting at least one server from a plurality of servers and a route to the server,

 An information collection step of collecting load information for each of the plurality of servers and a route to the plurality of servers;

 A service selecting step of selecting at least one server and a route to the server based on the load information collected by the information collecting unit;

 Service selection method.

29. A service selection program for causing a computer to select at least one server from among a plurality of servers and a route to the server, wherein the plurality of servers include a route to the plurality of servers. An information collection step for collecting load information for each of the

 A service selection step of selecting at least one server and a route to the server based on the load information collected by the information collection unit;

 A service selection program for causing a computer to execute the following.