JPH05143559A - Load distribution method in distributed processing system - Google Patents

Abstract

PURPOSE:To make it possible to perform an automatic load distribution without any manual work by performing an enquiry of a loading state to all service computers existing in a system when each sub-computer is connected with the service computers and connecting the sub-computers with the lightest load service computer based on the response. CONSTITUTION:When each sub-computers 21 to 24 are connected with service computers 11, 12, each sub-computer 21 (22/23/24) executes a load state enquiry program 2b. At first, a load enquiry packet which the address of the sub- computer and a request service name are set is prepared. Next, the load enquiry packet is transmitted as a multi-address notification for all the service computers 11, 12 by a broadcast mode to a network 3 via a network interface 2a and the state becomes the standby state of a response packet. Thus, the processing of a service is performed by automatically selecting the lightest load service computer connected on the network 3 and by connecting with this.

Description

Detailed Description of the Invention

[Object of the Invention]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a load balancing method in a distributed processing system.

[0002]

2. Description of the Related Art Conventionally, there is a distributed processing system including a plurality of service computers, a plurality of sub computers, and a network connecting the plurality of service computers and the plurality of sub computers. In this distributed processing system, when each sub-computer requests the service of the service computer, one of the plurality of sub-computers is selected and connected to it, and the service is provided from this connected service computer. Is becoming

By the way, in such a system, when connecting a sub computer to one service computer,
When connecting to any service computer, the load states among the service computers are biased. Therefore, it is necessary to consider the load state of each service computer when connecting, though the work is troublesome.

Therefore, in general, an operator calculates the number of sub-computers connected to the service computer based on the number of service computers and the number of sub-computers existing on the network so that each sub-computer is equalized. When the address of the service computer of the connection destination is set in the storage device on the computer and each sub computer requests the service, it is connected to the service computer of the address set in the storage device and it does it. .. In this way, load is distributed by allocating sub-computers to service computers.

However, this conventional load balancing method has a problem that the allocation becomes difficult and complicated as the number of computers on the network increases.

Particularly, as the number of sub-computers increases, the address setting work also increases, and the work becomes significantly complicated.

Further, when the system configuration is changed, such as the increase or decrease of the number of computers or the change of the performance of each service computer, etc., in the system whose allocation has been decided once, the allocation based on the number of the above-mentioned computers is newly added. There is also the problem of having to do some work.

[0008]

As described above, the conventional load balancing method requires a lot of labor, and the degree of the trouble is worse as the system is changed and the system is expanded. In other words, it can be said that it hinders the growth of the system, and drastic improvements are required.

The present invention has been made in view of the above-mentioned problems of the prior art. The object of the present invention is to automatically perform load distribution without human labor.
It is to provide a load balancing method in a distributed processing system.

[Structure of Invention]

[0011]

According to a load balancing method in a distributed processing system of the present invention, when each of a plurality of sub-computers connects to any service computer, the sub-computer inquires of the load status to the plurality of service computers. Is sent to the network as a response to the inquiry about the load status, and the sub-computer that generated the broadcast notification sends the load notification to the network based on the load status notification. It is characterized in that the service computer with the lightest load is obtained from the service computers that have issued the load status notification, and is connected to this service computer with the lightest load.

[0012]

According to the present invention, when each sub-computer is connected to the service computer, the service computers in the system are inquired about the load status toward all the service computers existing in the system, and the service computer with the lightest load is based on the response. The load is distributed by connecting to.

Therefore, since the sub-computers are connected to the service computer so as to be evenly allocated, the load can be automatically distributed without the need for manpower. As a result, the manual work is a constant simple work regardless of the size of the system, for example, a connection command is issued.
It can be as simple as

Further, each time the sub-computer is connected to the service computer, the service computer to be connected is determined to be the computer with the lightest load. Therefore, even if a system change such as the addition of a service computer occurs, the load is The distribution is done automatically without any human intervention.

[0015]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a flow chart showing a load status inquiry program of a sub-computer and a load status check program of a service computer for realizing a load balancing method according to an embodiment of the present invention, and FIG. 2 is a service of the embodiment shown in FIG. FIG. 3 is a functional block diagram of the computer and the sub computer, and FIG. 3 is a functional block diagram of the entire system.

First, FIG. 3 shows an example of a system in which two service computers are shown by reference numerals 11 and 12 and four sub computers are shown by reference numerals 21, ..., 24. Therefore, each of the sub computers 21 to 24 is connected to any of the two service computers 11 and 12 to receive the service.
3 is a network connecting these computers.

In FIG. 2, the service computer 1 comprises a network interface 1a, a load status check program 1b and a service program 1c, and the sub computer 2 comprises a network interface 2a, a load status inquiry program 2b and a service request program 2c. Have

In this configuration, each sub computer 21-2
4 connects to either of the service computers 11 and 12, first, each sub computer 21 (/ 22/23/2)
4) executes the load status inquiry program 2b. This load status inquiry program 2b is shown in FIG. 1A, and is started by an operator's connection instruction or the like. First, in step ST201, a load inquiry packet in which the address of the sub-computer and the requested service name are set is created, and in the next step ST202, the load inquiry packet is broadcasted to all service computers 11 and 12 in broadcast mode. It is sent to the network 3 via the network interface 2a. Then, the standby state for the response packet in steps ST203 and ST204 is entered.

The service computers 11 and 12 are shown in FIG.
The load status check program 1b shown in 1 is always executed in parallel with other processing programs. First, in step ST101, a load inquiry packet is waited for a certain period of time, and then in step ST102, the presence or absence of an inquiry is confirmed. If not, the process returns to step ST101 to repeat those processes. Then, via the network interface 1a, the sub computer 21 (/ 22/23/2
When the load inquiry packet from 4) is received, the determination in step ST102 is YES and step S
At T103, the requested service name in the received packet is compared with the service name of the service program 1c, and it is checked whether they match, that is, whether the requested service can be provided. Then, in step ST104, as a result of checking the requested service name, if the requested service cannot be provided, the determination is NO, and the process returns to step ST101. Also,
As a result of the check, if the service can be provided, the determination in step ST104 is YES and step 1
Move to 05. In this step ST105, a load response packet in which the number of sub-computers already connected to the service computer 11 (/ 12) self and its own address are set is assembled. Next, in step ST106, the load response packet is sent to the network 3 via the network interface 1a toward the inquiry source sub-computer 21 (/ 22/23/24). Then, the process returns to step ST101.

The sub computer 21 (/ 22/23/24) which has sent the load inquiry packet,
In ST204, the load response packet is in a standby state. That is, in step ST203, the load response packet is received and stored. Next step ST
At 204, it is checked whether or not a fixed time has passed since the state of waiting for a response packet, and if the fixed time has not passed, the process returns to step ST203 to continuously wait for a response packet. As a result of the determination in step ST204, if the fixed time has passed, the reception of the response packet is terminated at this point. In other words, the services that can be provided differ depending on the service computer, and it is expected that if the service computer is busy and unable to respond, or if the number of service computers changes, the number of responses returned is uncertain. It does not wait forever and cuts in time.

When the reception of the response packet is terminated, then in step ST205, the load state in the packet of the service computer that has made a response is checked to find the one with the lightest load. At this time, when the load states are the same, the service computer to be connected is determined by a predetermined priority level. Then, in step ST206, the service computer is connected to the obtained service computer and the processing is passed to the service request program 2c. After that, the service request program 2c and the service program 1c in the service computer are exchanged.

In such a configuration, the sub computer 2
When new connection processing is performed in the order of 1, 22, 23, and 24, the result is as follows.

First, the sub computer 21 is connected to the service computer 11 having a higher priority level because the number of sub computers connected to the service computer 11 and the number of sub computers connected to the service computer 12 are both “0” and the loads are equal. Process the service.

Next, in the sub computer 22, the number of sub computers connected to the service computer 11 is “1”, the number of sub computers connected to the service computer 12 is “0”, and the load of the service computer 12 is small. Since it is light, it is connected to the service computer 12 and processes the service.

The sub computer 23 is connected to the service computer 11 and performs service processing because the number of connected sub computers of the service computer 11 and the sub computer 12 is "1" and the loads are equal.

In the sub computer 24, the number of sub computers connected to the service computer 11 is "2", the number of sub computers connected to the service computer 12 is "1", and the service computer 12 is lighter in load. Because of this, the service computer 12 is connected to perform service processing.

As a result, the number of sub-computers connected to each of the service computers 11 and 12 is "2", and the loads are equalized.

As described above, the sub computer automatically selects the service computer with the lightest load among all the service computers connected to the network 3 and connects to the service computer to process the service. By doing so, the load of the service computer on the network 3 is evenly distributed.

Further, as described above, the service computer 1
Two sub computers 21 and 23 are connected to the service computer 1
When two sub-computers 22 and 24 are connected to 2, it is assumed that the number of service computers has increased by one and the number of sub-computers has increased by two.

In this case, the load condition is the service computer 1
1 is “2”, service computer 12 is “2”, and new service computer is “0”. Since this new service computer has the lightest load, the sub computer that first issued the request connects to that new service computer. Will process the service.

Then, the number of sub computers connected to the service computer 11 is "2", the number of sub computers connected to the service computer 12 is "2", and the number of sub computers connected to the new service computer is "1". Since the new service computer still has the lightest load, the sub computer requesting the second request will be connected to this new service computer.

Therefore, the number of sub-computers connected to each service computer is "2", and even if a service computer is newly installed, the load is automatically equalized without any human intervention.

[0034]

As described above, according to the present invention, when each sub-computer is connected to the service computer, a query of the load status is made to all the service computers existing in the system, and based on the response, The load is distributed by connecting to the service computer with the lightest load, so the sub computers are connected so that they are evenly allocated to the service computer, so the load is automatically loaded without human intervention. Dispersion can be performed. As a result, the manual work can be limited to a certain simple work regardless of the size of the system, for example, a simple work such as issuing a connection command.

Furthermore, each time the sub-computer is connected to the service computer, the service computer to be connected is determined to be the computer with the lightest load. Therefore, even if a system change such as the addition of a service computer occurs, the load is not changed. The distribution is done automatically without any human intervention.

[Brief description of drawings]

FIG. 1 is a flowchart of a load status check program of a service computer and a load inquiry program of a sub computer that realizes a load distribution method according to an embodiment of the present invention.

FIG. 2 is a functional block diagram of a service computer and a sub computer according to the embodiment shown in FIG.

FIG. 3 is a functional block diagram of the entire system.

[Explanation of symbols]

11, 12 Service computer 1a Network interface 1b Load status check program 1c Service program 21-24 Sub computer 2a Network interface 2b Load inquiry program 2c Service request program 3 Network ST201, ST202 Sub computer load inquiry processing ST101-ST106 Service computer Load state response process ST203 to ST206 Sub computer connection process

Claims (1)

[Claims] 1. A plurality of service computers, a plurality of sub-computers each of which is connected to any one of the plurality of service computers to receive the service thereof, the plurality of service computers and the plurality of sub-computers. In a distributed processing system consisting of a network connecting the plurality of sub-computers, when each of the plurality of sub-computers connects to any of the plurality of service computers, the sub-computer broadcasts as a query of the load status to the plurality of service computers. To the network, the plurality of service computers send a load status notification to the network as a response to the inquiry about the load status, and the sub-computer that has generated the broadcast notification loads the load based on the load status notification. Find the service computer with the lightest load among the service computers that issued the status notification And a load balancing method in a distributed processing system, wherein the load balancing method is connected to the lightest load service computer.