JPH11338836A - Load distribution system for computer network - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve responsiveness to a service request from a client to a server. SOLUTION: This system is provided with a load monitoring means 21 by nodes for detecting the respective load levels of plural server nodes, optimum server retrieving means 40 for identifying the optimum server node based on the detected levels of loads out of plural server nodes, to which servers 22 and 23 capable of executing the service requested from a client node are connected, while grasping the kind of the service provided by the respective plural server nodes and the levels of the loads of respective server nodes detected by the load monitoring means 21 by nodes, and repeating control means 30 for repeating the service request from the client node to servers 22 and 23 and selectively sending the service request from the client node to the optimum server node.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a load distribution system for a computer network in which a plurality of server nodes and client nodes are formed, and at least one server is connected to each of the plurality of server nodes.

[0002]

2. Description of the Related Art For example, in a computer network to which a plurality of servers capable of executing an application program requested by a user are connected, a server which executes the application program is appropriately selected in accordance with a request of the user to thereby load the server. Can be dispersed.

If the processing is concentrated on a specific server and the load on the server becomes excessive, the processing is delayed and the real-time service is lost. By appropriately distributing various services executed in response to a user's request to a plurality of servers, it is possible to prevent processing concentration and avoid service delay. A conventional computer network load balancing system is configured as shown in FIG. This computer network includes a plurality of client nodes and a plurality of server nodes. Each server node has its own node load information management module that measures and manages its load level, and each client node has other node load information that collects and manages the load level of each server node. A management module is provided.

[0004] The other node load information management module of each client node receives load level data detected by all server nodes, respectively, and uses it to detect a server node with a small load.

[0005]

In a conventional computer network load distribution system, each server node needs to notify all client nodes of the load level data of the own node. Therefore, as the number of client nodes increases, the load on the server node and the load on the network increase.

[0006] Load level data is periodically transferred between the server node and the client node at predetermined time intervals. If the time interval is relatively large, control delay may occur. For example, if the load level of the server node rises rapidly, the load on the server node becomes excessive before the client node recognizes the increase in load, and processing is delayed because more service is allocated to the server node with excessive load. Occurs.

In a computer network in which a plurality of domains such as a local area network (LAN) are connected, server nodes belonging to different domains can be used. However, since the distance of the communication path between the domains is generally long, the required communication time is long. If a server node belonging to another domain is used in consideration of only the magnitude of the load, the response time from the server node becomes slow, and it is often impossible to predict the time until the response.

[0008] For an application in which an interactive process is performed between a person and a node, a responsiveness of several seconds to several tens of seconds is required for a server that provides a service. An object of the present invention is to improve responsiveness to a service request from a client to a server in a load distribution system of a computer network.

[0009]

According to a first aspect of the present invention, there is provided a load distribution system for a computer network, wherein a plurality of server nodes and a client node are formed, and at least one server is connected to each of the plurality of server nodes. In a load distribution system for a computer network, a node-by-node load monitoring unit that detects a load level for each of the plurality of server nodes, a type of service provided by each of the plurality of server nodes, and the node-by-node load monitoring unit include: By grasping the detected load level of each server node, an optimal server is selected from a plurality of server nodes to which a server capable of executing the service requested by the client node is connected, based on the detected load level. An optimal server search means for identifying a node; Relay control means for relaying a service request from the client node to the server and selectively transmitting a service request from the client node to the optimum server node identified by the optimum server search means. And

The node-specific load monitoring means detects a load level for each of the plurality of server nodes. The optimum server searching means executes the service requested by the client node by grasping the type of service provided by each of the plurality of server nodes and the load level of each server node detected by the node-specific load monitoring means. Based on the detected load level, for example, an optimum server node with the lowest load level is identified from among the plurality of server nodes to which the possible servers are connected.

The relay control means relays a service request from the client node to the server, and selectively sends a service request from the client node to the optimum server node identified by the optimum server search means. In the present invention, since the load level of each server node is centrally managed by the optimum server searching means, each server node does not need to notify the client node of load level data. Therefore, even if the number of client nodes increases, the processing of each server node for notifying the data of the load level does not increase, and the traffic on the network does not increase.

Further, since a service request from the client node is selectively transmitted to the server node having the detected minimum load level, high responsiveness to the service request can be obtained. According to a second aspect of the present invention, in the load distribution system of the computer network according to the first aspect, a service is allocated from the client node to the server for a server node in which an increase in load of a predetermined level or more is detected within a predetermined time. The present invention is characterized in that an assignment control means for temporarily prohibiting is provided.

In a case where the time interval for notifying the optimum server searching means of the load level from the node-specific load monitoring means is relatively large, if the load level of the server node rises rapidly, the increase in the load is determined by the optimum server searching. There is a possibility that the load on the server node becomes excessive before the means is recognized, and further service is allocated to the server node with the excessive load. Further, if the time interval for notifying the load level is reduced, the load required for the processing and the network traffic increase.

[0014] The allocation control means, for a server node in which an increase in load over a predetermined time within a predetermined time is detected,
The assignment of the service from the client node to the server is temporarily prohibited. Therefore, service allocation is prohibited before the load level becomes excessive for a server node whose load level tends to increase and the load level is likely to become excessive.

For this reason, even if the time interval at which the load monitoring means for each node notifies the optimum server search means of the load level is relatively large, the service is allocated to the server node having an excessive load. Can be prevented beforehand. According to a third aspect of the present invention, in the load distribution system of the computer network according to the second aspect, the assignment control unit determines that the node-specific load monitoring unit determines that the server node detects an increase in load of a predetermined level or more within a predetermined time period. The load level of the detected server node is forcibly changed to a predetermined maximum value.

For a server node in which an increase in load by a predetermined amount or more is detected within a predetermined time, the load level notified from the node-specific load monitoring means to the optimum server searching means becomes a maximum value, that is, an excessive level. Thus, a server node whose load increases rapidly is not recognized as an optimal server. The load distribution system for a computer network according to claim 4 includes a plurality of domains, each domain including a plurality of server nodes and a client node, and each of the plurality of server nodes includes at least one server. In a load distribution system of a connected computer network, a node-by-node load monitoring means for detecting a load level of each of the plurality of server nodes, a first communication time required for communication between a plurality of domains, and Communication time detecting means for detecting a second communication required time required for communication between a predetermined position and each server node, a type of service provided by each of the plurality of server nodes in a plurality of domains, and a load for each node. The load level of each server node detected by the monitoring means, and the communication time Based on the first communication required time detecting means has detected and a second time required for communications, from among the server nodes executable server service requested by the client nodes are connected 1
Optimal server search means for identifying one server node as an optimal server node, and a client node for relaying a service request from the client node to the server, and selectively for the optimal server node identified by the optimal server search means. And a relay control unit for transmitting a service request from the server.

The node-specific load monitoring means detects a load level for each of the plurality of server nodes. The communication time detecting means is a communication time detecting means for performing communication between a plurality of domains.
And a second communication time required for communication between a predetermined position in each domain and each server node. The optimum server searching means includes: a type of service provided by each of the plurality of server nodes in a plurality of domains;
A service requested by the client node based on a load level of each server node detected by the node-specific load monitoring means and a first communication required time and a second communication required time detected by the communication time detecting means. Is identified as an optimal server node from among the server nodes to which the server capable of executing is connected.

The relay control means relays a service request from the client node to the server, and selectively sends a service request from the client node to the optimum server node identified by the optimum server search means. The optimal server search means identifies the optimal server node from a plurality of domains connected to the network. Therefore, if there is no server node with a small load in the same domain as the client node requesting the service, another optimal server node searches for the optimal server node. A service can be assigned to a server node with a small load belonging to a domain.

The optimum server searching means identifies the optimum server node in consideration of not only the load level of each server node but also the first communication required time and the second communication required time. The service is not frequently assigned to server nodes belonging to a domain far away from even if the load level is small.

That is, if the load level is the same, the service can be preferentially assigned to the server node belonging to the domain closest to the client node, so that the load can be efficiently distributed. Claim 5 is Claim 4
In the load distribution system of a computer network described in the above, the distance of a communication path between the relay control means and each server node in each domain is set to be substantially the same, and the communication time detecting means is provided with the relay control means and each server. The second communication required time is detected based on a required round-trip time of a signal actually transmitted to and from the node.

The communication time detecting means detects the second required communication time based on the required round-trip time of a signal actually transmitted between the relay control means and each server node.
Since the distance of the communication path between the relay control means and each server node in each domain is almost the same, there is basically no difference in the communication time of each server node in the same domain. Therefore, the error included in the second communication required time in the same domain does not adversely affect the identification of the optimal server node.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The configuration and operation of a load distribution system according to an embodiment are shown in FIGS. This form corresponds to all claims. FIG. 1 is a block diagram showing a configuration example of one domain of a computer network including the load distribution system of this embodiment. FIG. 2 is a block diagram showing a configuration example of two domains of a computer network including the load distribution system of this embodiment.

FIG. 3 is a flowchart showing the contents of the load notification processing of the own node load information management module. FIG.
Is a flowchart showing the content of the search processing of the optimum server search module. FIG. 5 is a flowchart showing the contents of the distance measurement processing in the domain. FIG. 6 is a flowchart showing the content of the distance measurement process between domains. In this embodiment, the node-by-node load monitoring means, the optimum server searching means, and the relay control means correspond to the own node load information management module 21, the management node 40, and the relay device 30, respectively.

The assignment control means according to claims 2 and 3 corresponds to the own node load information management module 21.
The node-specific load monitoring means, the communication time detecting means according to claim 4,
The optimum server search means and the relay control means correspond to the own node load information management module 21, the own node load information management module 21, the management node 40, and the relay device 30, respectively.

The load distribution system of the present invention is implemented in a computer network as shown in FIGS. This type of computer network is composed of only one domain 50 (A) as shown in FIG. 1, or two or more domains 50 (A) and 50 () which can communicate with each other as shown in FIG. B), in some cases. In FIGS. 1 and 2, the symbols A and B in parentheses added to the reference numerals of the respective constituent elements indicate the domains to which they belong, and similarly, the numbers 1, 2, and 3 in parentheses indicate the nodes. . In the following description, these sections will be denoted by reference numerals of the respective constituent elements as necessary.

Referring to FIG. 1, the domain 50 includes a plurality of client nodes 10, a plurality of server nodes 20, a relay device 30, and a management node 40. This domain 50 is, in reality, a network such as a local area network (LAN) in which a number of nodes are connected to each other via a switch such as a hub or a router.

In FIG. 1, a plurality of client nodes 10, a plurality of server nodes 20, a relay device 30, and a management node 40 are connected to each other by a network as shown by dotted lines. The solid arrows shown in FIG. 1 and FIG.
2 shows a main signal flow in node assignment control.
Although many client nodes 10 and many server nodes 20 are actually connected to the network, FIGS. 1 and 2 show only a part of them.

The relay device 30 includes components similar to those of the exchange, and connects each component in the domain 50 and a plurality of domains 50 as necessary, thereby enabling communication between the connected components. I do. The relay device 30 includes a proxy module 31. This proxy module 31
For example, it is a program that can be executed by a computer built in the relay device 30 and realizes predetermined functions described later.

One or more servers 22 and 23 are connected to the server node 20 as needed. Server 22, 2
3 includes one or more computers built in the server node 20 and a server program of an application executed by the computer. In this example, servers 22 and 23 are provided with server programs of different types of applications classified by symbols α and β.

One or more clients 11 are connected to the client node 10 as needed. The client 11 includes one or more computers built in the client node 10 and a client program of an application executed by the computer.
When the client 11 uses a certain kind of application, any one of the server nodes 2 may be used in response to a request from the client program in the client 11.
The server program of the server 22 or 23 belonging to 0 starts. The server program of the server 22 or 23 executes a predetermined service and returns the result to the client 11.

In the domain 50 shown in FIG. 1, the server 22 for executing the application of the type corresponding to the symbol α is connected to a plurality of server nodes 20.
The service according to the request from the client 11 can be processed by any of the plurality of server nodes 20 (A, 1), 20 (A, 2), and 20 (A, 3). In this embodiment, the client 11 does not directly select any server node 20. In this embodiment, a service request from the client 11 is input to the relay device 30 in order to distribute the load, and the relay device 30 transmits the service request to the plurality of server nodes 20 (A, 1), 20 (A, 2), and 20 (A, 2).
One of (A, 3) is selected, and the service request is relayed to the server 22 belonging to the selected one server node 20.

Each server node 20 is provided with its own node load information management module 21 for managing the magnitude of the load inside the server node. The own node load information management module 21 is a control program executed by a computer built in the server node 20.
The management node 40 responds to a request from the relay device 30,
An optimum server node 20 in the domain 50 is determined, and the result is notified to the relay device 30. The management node 40 includes an optimum server search module 41 and a full load information management module 42. The optimal server search module 41 and the full load information management module 42
The control program is executed by a computer built in the management node 40.

The full load information management module 42 receives load information on all server nodes 20 in the domain 50 from the own node load information management module 21 of each server node 20 and manages it. The optimal server search module 41 includes a load information of each server node 20 (A) managed by the full load information management module 42 in the domain 50 (A) and a server node 20 managed by another domain 50 (B).
Based on the load information and the distance information of (B), an optimal server node is determined from the viewpoint of load distribution and real-time processing.

In the domain 50, the components are arranged such that the distance between the relay device 30 and each server node 20 is substantially the same. For this reason, the required communication time between the relay device 30 and each server node 20 is substantially the same. Therefore, the optimal server node in the domain 50 is determined to be the server node 20 with the smallest load. Of course, it is limited to the server node 20 provided with the server 22 or 23 that can execute the requested service.

For example, when one of the clients 11 sends a request for a process executable by the server (α) 22 to the proxy module 31 of the relay device 30, the proxy module 31 sends a load and real-time performance to the optimal server search module 41. Is requested to search for the optimal server (α) 22 from the viewpoint of (1). The optimal server search module 41 is an optimal server
When the search request for (α) 22 is received, the load information of the server node 20 to which the server (α) 22 belongs is requested to the full load information management module 42.

The full load information management module 42 always has its own node load information management module 2 of each server node 20.
When the load information of the server node is requested from the optimum server search module 41, the managed load information is returned to the optimum server search module 41. The optimal server search module 41
The server which receives the load information from the full load information management module 42 and is optimal in terms of load and real-time performance.
(α) 22 is detected, and the identifiers of the server node 20 and the server (α) 22 to which the (α) 22 belongs and information indicating the position thereof are returned.

The contents of each step of the load notification processing of the own node load information management module 21 provided in each server node 20 will be described with reference to FIG. Step S1
In step 1, the counter i and the value of the register X allocated to the internal memory of the computer executing this program are cleared.

In step S12, the magnitude of the load on the own node is obtained. Specifically, a calculation formula which is determined in advance by using the latest CPU utilization, memory utilization, and network load state (the number of clients connected to the node) detected for the computer connected to the own node as parameters Calculate the magnitude of the load from. In this example, the magnitude of the load is represented by being classified into Xm levels from 1 to a predetermined maximum value Xm. The range of the load level may be determined independently for each node. For example,
The minimum value of the load level may be changed to X1 (1 <X1 <Xm) according to the capacity of the CPU, the amount of memory, and the bandwidth of the network.

Step S12 is repeatedly executed at a constant time period. The magnitude of the load obtained in step S12 is determined by the value of the register X specified by the value of the counter i.
(i) is held. The value of the counter i is determined in step S12.
Is updated in step S18 or S11 each time is executed. In step S13, the contents of the register X (i) holding the result of step S12 are compared with the maximum value Xm.
The magnitude of the load detected in step S12 is the maximum value Xm
If it is less than the value, the process proceeds to step S14, and if the detected magnitude of the load is the maximum value Xm, the process proceeds to step S19.

In step S14, the value held by the register X (i-1) specified by (i-1) is compared with a predetermined threshold value Xa. That is, one cycle before step S12
Is compared with the threshold value Xa. If the value of the register X (i-1) is smaller than the threshold value Xa, step S
Proceed to step S16, otherwise proceed to step S15. In step S15, the difference between the value held by the register X (i) and the value held by the register X (i-1) is compared with a predetermined threshold value Xb. That is, the change amount of the load during one load detection period is compared with the threshold value Xb. The change amount is a threshold value Xb
If it is less, the process proceeds to step S16; otherwise, the process proceeds to step S20. For example, when the load increases rapidly, the process may proceed from step S15 to S20.

In step S16, the value held in the register X (i) is added to the register X to update the value of the register X. In step S17, the value of the counter i is set to (k−
Compare with 1). k is a constant. If the value of the counter i is less than (k-1), the process proceeds to step S18; otherwise, the process proceeds to step S21.

In step S18, the value of the counter i is updated. In step S19, the value held by the register Np is compared with the maximum value Xm. The value held by the register Np is the value of the load information previously notified to the full load information management module 42 in the process of step S22. Register N
If the value of p is less than the maximum value Xm, the process proceeds to step S20; otherwise, the process proceeds to step S11.

In step S20, the maximum value Xm is held in the register N. In step S21, the result of dividing the value of the register X by the constant k is held in the register N. In step S22, the value held in the register N is notified to the full load information management module 42 as the detected load level.

In step S23, the notified load level is stored in the register Np. Therefore, step S12
If the load level detected in step S12 is relatively small, the load level detected in step S12 is accumulated in the register X. Then, every time the load level is detected k times,
The average value of the accumulated values of the detected k data is calculated in step S
21. This average value is calculated for a predetermined period (for example, 1
Min) load level. This average value is notified to the full load information management module 42 at step S22 for each predetermined cycle.

However, if the load level detected in step S12 is equal to the maximum value Xm, and the load level notified to the full load information management module 42 in the previous step S22 is smaller than the maximum value Xm, step S13- S1
Since 9-S20-S22 is executed, the maximum value Xm is immediately notified to the full load information management module 42.

If the load level detected one cycle before in step S12 is equal to or greater than the threshold value Xa, and if the amount of change in the load level during one cycle of the load detection cycle is equal to or greater than the threshold value Xb, step S14- Since S15-S20-S22 are executed, the maximum value Xm is immediately notified to the full load information management module 42. That is, when a rapid increase in the load is detected, a maximum value Xm larger than the actually detected value is notified as the load level in order to temporarily prohibit the allocation of further processing to the server node 20. I do.

Since the own node load information management module 21 of each server node 20 executes the processing shown in FIG. 3, all the load information management modules 42 of the management node 40 have all the server nodes 20 in the domain 50. , Information indicating the level of the load is collected. Next, the contents of the search processing performed by the optimum server search module 41 will be described with reference to FIG. Domain 50
An example in which the client 11 belonging to (A) requests service allocation will be specifically described.

In step S31, the process waits until there is a search request. When the client 11 requests the proxy module 31 of the relay device 30 for service allocation, the proxy module 31 sends a search request to the optimal server search module 41 belonging to the same domain 50. Also, as shown in FIG. 2, when a plurality of domains 50 (A) and 50 (B) are connected and the optimum server search module 41 (A) belonging to the domain 50 (A) receives a search request, the optimum server The search module 41 (A) receives the optimum server search module belonging to the domain 50 (B) via the proxy module 31 (A) of the relay device 30 (A) and the proxy module 31 (B) of the relay device 30 (B). A search request is sent to 41 (B).

Upon receiving the search request, the process of the optimum server search module 41 proceeds from step S31 to S32.
In step S32, all server nodes 20 to which the server 22 (or 23) capable of executing the service is connected are identified based on the information of the identifier indicating the type of service requested by the client 11.

For example, when the client 11 requests allocation of a service represented by the symbol α in FIG.
The optimal server search module 41 acquires the identifiers of the server nodes 20 (A, 1), 20 (A, 2), and 20 (A, 3) including the server (α) 22 in step S32. Step S3
At 3, it is determined whether or not another domain is available. When a plurality of domains 50 (A) and 50 (B) are connected to each other as shown in FIG.
Proceed to.

In step S34, a search request is sent to the optimum server search module 41 belonging to another domain 50 together with an identifier indicating the type (α, β) of the server 22, 23 to be searched. Optimal server search module 4
The optimal server search module 41 (B) that has received the search request from the server 1 (A) also executes the processing in FIG. 4 in the same manner as the optimal server search module 41 (A).

In step S35, it is determined whether one or more server nodes 20 are detected in step S32.
When a plurality of server nodes 20 are detected, the process proceeds from step S35 to S36 in order to select an optimal server node 20 in the domain 50. In step S36,
A server 22 capable of executing a request from the client 11
The load information (load level) of all server nodes 20 to which (or 23) is connected is stored in the full load information management module 4.
Obtain from 2.

In step S37, one server node 20 having the minimum load level among all the server nodes 20 to which the server 22 (or 23) capable of executing the request from the client 11 is connected is determined. The server node 20 determined here is the optimum server node 20 in the domain 50. If there are a plurality of server nodes 20 having the minimum load level,
Choose one randomly.

When the optimum server search module 41 (A) requests the search to the optimum server search module 41 (B) of the other domain 50 (B) in step S34, the search of the optimum server search module 41 (B) is performed. Result is Step S38
Get in. The search result of the optimal server search module 41 (B) is obtained by the proxy module 31 (B) of the relay device 30 (B),
It is input to the optimal server search module 41 (A) via the proxy module 31 (A) of the relay device 30 (A).

The optimum server search module 41 determines the required communication time T1 between each server node 20 and the relay device 30 in the domain 50 and the domain 50 adjacent to each other.
(A), adjacent relay devices 30 (A), 30 (B) belonging to 50 (B).
The communication required time T2 during (B) is managed as distance information. In step S38, the other domain 50 (B)
From the optimal server search module 41 (B) belonging to the domain 50 (B), the identifier indicating one server node 20 (B) having the lowest load level among the server nodes 20 (B) belonging to the domain 50 (B). The data of the required communication time T1 (B) between 20 (B) and the relay device 30 (B) arrives at the optimum server search module 41 (A).

In step S39, the optimum server search module 41 (A) detects the optimum one server node 20 (A) in the domain 50 (A) detected in step S37 and the other server 50 (B). With respect to the selected one optimal server node 20 (B) and the like, the optimal one server node 20 is determined for the entire network in consideration of the load level and the distance information.

For one optimal server node 20 (A) in the domain 50 (A), the distance between the server node 20 (A) managed by the optimal server search module 41 and the relay device 30 (A) is set as the distance. The required communication time T1 (A) is applied. The distance between the server node 20 (B) of the other domain 50 (B) and the communication requirement between the server node 20 (B) received from the optimal server search module 41 (B) and the relay device 30 (B) is set. The result obtained by adding the communication time T2 between the relay devices 30 (A) and 30 (B) across the domain 50 managed by the optimal server search module 41 (A) to the time T1 (B) is applied.

When the server node 20 spanning a plurality of domains 50 is used, the required communication time T2 is obtained by accumulating the required communication time between the respective domains 50. In step S39, an optimum one server node 20 is selected from the plurality of server nodes 20 detected in the plurality of domains 50 using a predetermined calculation formula.
Identify. For example, the server node 2 of another domain 50 (B) is compared with the server node 20 (A) of the domain 50 (A).
Even if the load level of 0 (B) is small, if the communication time T2 between the relay devices 30 (A) and 30 (B) is long, the server node 20 in the domain 50 (A) is not required. (A) is identified as the optimal server node.

In step S40, for example, the search result is returned to the proxy module 31 that has requested the search. In the case of the optimal server search module 41 (B) belonging to another domain 50 (B) requested to be searched by the optimal server search module 41 (A), the request source optimal server search module 41 (A) in step S40. Relay device 30
The result is notified via (B) and 30 (A).

The proxy module 31 of the relay device 30 that receives the search result from the optimum server search module 41 in response to the search request allocates the service from the client 11 to the optimum server node determined by the optimum server search module 41. So that the request is relayed.

The server node 20 in the domain 50
The required communication time T1 between the communication device and the relay device 30 is obtained by the process shown in FIG. The process illustrated in FIG. 5 is executed, for example, when the power of the server node 20 is turned on. The processing of FIG. 5 will be described below. In step S51, the server node 20 refers to the time of the clock circuit of the computer built in the server node 20 to set the current time t.
Detects 11. Immediately thereafter, predetermined data having a small data length is transmitted from the server node 20 to the relay device 30 in the domain 50 in step S52.

When the data transmitted from the server node 20 in the domain 50 is received, the processing of the relay device 30 proceeds from step S61 to S62. Step S62
Then, the received data is immediately transmitted to the server node 2 of the transmission source.
Return to 0 and send. Upon receiving the reply data from the relay device 30, the server node 20 proceeds from step S53 to S54. In step S54, the current time t12 is again referred to by referring to the time of the clock circuit built in the computer.
Is detected.

The server node 20 calculates the required communication time T1 based on the following formula in step S55. T1 = (t12−t11) / 2 The server node 20 sends the determined required communication time T1 to the management node 40 in the next step S56.
The required communication time T1 is managed by the optimum server search module 41 of the management node 40.

On the other hand, the relay device 3 spanning the domain 50
The required communication time T2 between 0 (A) and 30 (B) is obtained by the processing shown in FIG. The process shown in FIG. 6 is executed, for example, when the power of the relay device 30 is turned on. FIG.
Will be described below. The relay device 30 (A)
In step S71, the current time t21 is detected by referring to the time of the clock circuit of the computer incorporated therein. Immediately thereafter, in step S72, predetermined data having a small data length is transmitted from the relay device 30 (A) to the relay device 30 of the other domain 50.
(B).

The processing of the relay device 30 (B)
When the data transmitted from (A) is received, the process proceeds from step S81 to S82. In step S82, the received data is immediately transmitted back to the transmission source relay device 30 (A). Upon receiving the return data from the relay device 30 (B), the relay device 30 (A) proceeds from step S73 to S74. In step S74, the current time t22 is detected again by referring to the time of the clock circuit built in the computer.

The relay device 30 (A) calculates the required communication time T2 based on the following formula in step S75. T2 = (t22−t21) / 2 The relay device 30 (A) compares the determined required communication time T2 with the management node 40 in the domain 50 in the next step S76.
Send to (A). The required communication time T2 is managed by the optimum server search module 41 (A) of the management node 40 (A).

Since the same processing is executed in the domain 50 (B), the optimum server search module 41 (B) can acquire and manage the required communication time T2.

[0068]

According to the present invention, since the load level of each server node is centrally managed by the optimum server search means, each server node needs to notify the client node of the load level data. Absent. Therefore,
Even if the number of client nodes increases, the processing of each server node for notifying the data of the load level does not increase, and the traffic on the network does not increase.

Further, since a service request from the client node is selectively transmitted to the server node having the minimum detected load level, high responsiveness to this service request can be obtained. In addition, by providing an assignment control unit for temporarily prohibiting the assignment of a service to a server node whose load increases rapidly, the load level of a server node which is likely to be excessively high is provided. Service allocation is prohibited before it becomes too large. For this reason, even if the time interval for notifying the optimum server search unit of the load level from the node-specific load monitoring unit is relatively large, it is possible to prevent the service from being assigned to the server node with an excessive load. .

In an environment where a plurality of domains can be used, not only the load level of each server node detected by the node-specific load monitoring means, but also the first communication required time and the second communication time detected by the communication time detecting means. By providing the optimum server search means for identifying the optimum server node in consideration of the required communication time, the load can be appropriately distributed in consideration of the response from the server.

Further, by setting the distance of the communication path between the relay control means in each domain and each server node to be substantially the same, a slight error included in the detected required second communication time is allowed. Since it is possible, it is easy to detect the second required communication time. Further, by detecting the second required communication time based on the required round-trip time of the signal actually transmitted between the relay control means and each server node, the second required communication time can be obtained relatively easily. Can be

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration example of one domain of a computer network including a load distribution system according to an embodiment.

FIG. 2 is a block diagram illustrating a configuration example of two domains of a computer network including the load distribution system according to the embodiment;

FIG. 3 is a flowchart showing the contents of a load notification process of a local node load information management module.

FIG. 4 is a flowchart showing the contents of a search process of an optimum server search module.

FIG. 5 is a flowchart illustrating the content of a distance measurement process in a domain.

FIG. 6 is a flowchart illustrating the content of a distance measurement process between domains.

FIG. 7 is a block diagram illustrating a configuration example of a conventional load distribution system.

[Explanation of symbols]

 Reference Signs List 10 client node 11 client 20 server node 21 own node load information management module 22, 23 server 30 relay device 31 proxy module 40 management node 41 optimum server search module 42 full load information management module 50 domain

Claims (5)

[Claims] 1. A load balancing system for a computer network, wherein a plurality of server nodes and a client node are formed, and at least one server is connected to each of the plurality of server nodes. A node-based load monitoring unit that detects a load level; a type of service provided by each of the plurality of server nodes; and a load level of each server node detected by the node-based load monitoring unit. An optimal server search unit for identifying an optimal server node based on a detected load level from a plurality of server nodes to which a server capable of executing a service requested by the node is connected; and from the client node to a server. Relay the service request of A load balancing system for a computer network, comprising: relay control means for selectively transmitting a service request from the client node to the optimum server node identified by the server search means. 2. A load distribution system for a computer network according to claim 1, wherein a service is allocated from the client node to the server for a server node in which an increase in load of a predetermined level or more is detected within a predetermined time. A load distribution system for a computer network, comprising an assignment control unit for temporarily inhibiting the load. 3. The load distribution system of a computer network according to claim 2, wherein said allocation control means is configured to: for each of the server nodes for which an increase in load of a predetermined level or more is detected within a predetermined time period, said node-specific load monitoring means. A load distribution system for a computer network, wherein a detected load level of a server node is forcibly changed to a predetermined maximum value. 4. A computer network comprising a plurality of domains, each domain having a plurality of server nodes and client nodes, and each of the plurality of server nodes being connected to at least one server. In the distributed system, a node-by-node load monitoring means for detecting a load level for each of the plurality of server nodes; a first communication time required for communication between a plurality of domains; a predetermined position in each domain; Communication time detecting means for detecting a second required communication time required for communication with the server, type of service provided by each of the plurality of server nodes in a plurality of domains, and each server detected by the node-specific load monitoring means. The node load level, the first communication required time detected by the communication time detecting means, and the Optimal server search means for identifying one server node as an optimal server node from server nodes to which a server capable of executing a service requested by the client node is connected, based on the required communication time of the client node; Relay control means for relaying a service request from the node to the server and selectively transmitting a service request from the client node to the optimum server node identified by the optimum server search means. Computer network load balancing system. 5. The load distribution system for a computer network according to claim 4, wherein a communication path distance between said relay control means and each server node in each domain is set to be substantially equal, and said communication time detecting means is provided. And detecting the second required communication time based on a required round-trip time of a signal actually transmitted between the relay control means and each server node.