JP2003016044A - System for load distribution, method therefor and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system for load distribution, a method therefor and a program capable of performing a load distribution by using an operation agent and avoiding setups at every operation terminal in advance even in a parallel task and sequential processing. SOLUTION: A main process controlling an application and the operation agent execute an application function, a cyclic agent traveling over a network performs load measurement of terminals 10-16 except 13 on a network, when the terminals 10-16 except 13 become a predetermined status, the operation agent is transmitted to other terminals 10-16 except 13 in the network based on the result of the measurement to bring the operation agent into execution of the application function.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a load balancing system, and more particularly to a load balancing system, a load balancing method and a program for causing a work agent to execute a part of an application function.

[0002]

2. Description of the Related Art As prior art, Japanese Patent Laid-Open No. 9-16534
In the publication, an agent system is provided between a distributed process processing method and its device client and server,
Disclosed is a technique for distributing and executing processing requests to a plurality of server processes according to the load and resource status of each computer in the network. Also, for parallel tasks, a parallel task control list is created, and the agent sends the task to another process based on this to process it.

Further, Japanese Patent Laid-Open No. 2000-76172 discloses a system in which the execution position of a process requested by a distributed system client can be dynamically changed to either the client side or the server side.

[0004]

However, in a terminal in an environment composed of a plurality of devices connected by a network, for example, when executing an application having a server function, the terminal capacity is required for a plurality of client requests. There was a problem of lack of sufficient CPU power. Also,
Even when the terminal is a portable information terminal, there are many other cases where there is a concern about power consumption, or when the CPU power is insufficient.

To solve the above-mentioned problem of insufficient CPU power, conventionally, in Japanese Patent Laid-Open No. 9-16534, a plurality of processing engines are prepared and the requests are distributed so that the load is not concentrated on a specific processing engine. , Load balancing. In addition, the processing terminal becomes unresponsive when the load exceeds the processing capacity. Therefore,
In order to avoid this, even if the increase in load is not a chronic phenomenon but a temporary phenomenon, it is necessary to prepare a plurality of engines in each work terminal in advance assuming the case where the load becomes maximum.
However, this has led to increased costs for purchasing multiple terminals and increased work such as setup. In addition, JP-A-2000-
With regard to 76172, since the execution position is moved only between the client and the server, neither the server nor the client can handle the heavy load.

From the above, it is an object of the present invention to perform load distribution by sending work agents even in parallel tasks and sequential processing, and a load distribution system and a load distribution method that do not require setup for each work terminal in advance. To provide a program.

[0007]

A load balancing system, a load balancing method and a program according to the present invention are a system composed of a plurality of terminals connected by a network, and a control means for controlling an application, A work agent movable in the network executes the application function, a load agent that circulates in the network measures the load on a terminal on the network, and the terminal is in a predetermined state. At this time, there is provided means for transmitting the work agent to another terminal of the network based on the result of the load measurement and causing the work agent to execute the application function.

The application function executing means of the work agent executes a part of the application.

Further, the predetermined state of the terminal is characterized in that the load of the terminal is increased or the remaining power of the terminal is below a predetermined value.

Further, the work agent is sent out,
It becomes a terminal that holds the load measurement information and is configured with a terminal that receives and executes the work agent. According to an instruction from the terminal that holds the load measurement information, another terminal connected to the network Load distribution is performed by transmitting the work agent and causing the work agent to execute a part of the application function.

Further, a load balancing system is characterized in that a plurality of terminals for receiving and executing the work agent are connected to the network.

[0012] A load distribution method comprising a plurality of terminals connected by a network, wherein a main process step is provided as a control step for controlling an application,
A step in which a work agent movable in the network executes the application function; a step in which a cyclic agent that circulates in the network measures a load on a terminal on the network; and the terminal enters a predetermined state. When the load measurement result is transmitted to the other terminal of the network based on the result of the load measurement,
And a step of causing the work agent to execute the application function.

When a load distribution system composed of a plurality of terminals connected by a network is used, a main process process as a control process for controlling an application and a work agent movable in the network are used as the application process. A process for executing a function, a process for measuring a load on a terminal on the network using a patrol agent that patrols the network, and a process for measuring the load on the network based on the result of the load measurement when the terminal is in a predetermined state. It is a program including instructions for transmitting the work agent to another terminal and causing the work agent to execute the application function.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

First, FIG. 1 shows a LAN according to an embodiment of the present invention.
The device configuration when used in (Local Area Network) is shown.

As shown in FIG. 1, a plurality of terminals connected to the network, that is, a server (not shown), a personal computer (hereinafter PC) 10, a workstation (12-13), a printer 14, a storage device. 15
Etc., a terminal with application execution function, router 1
It is composed of 6 etc.

Here, the configuration in the case where there are a plurality of terminals that perform the same application (for example, 11-13) is also shown, but it is assumed that there are a plurality of server terminals.

Although not shown, a personal digital assistant,
A configuration is also conceivable in which a mobile phone or the like connects to the network by wireless communication such as Bluetooth. In both cases,
A terminal that executes an application sends an agent to another device on the network and causes the agent to execute a part of the application function, thereby distributing the load of the CPU of the terminal.

Next, FIG. 2 shows the configuration of the application process 24 included in the terminal having the application execution function. As shown in FIG. 2, when the load on the CPU of the terminal becomes heavy, the load measurement agent 30 is sent to the network by the device processing capacity and load list creation process 22, and thereafter, the devices on the network are predetermined. It voluntarily travels along any route and reports load information to the process. With the configuration of FIG. 1 described above, the device to be connected can be acquired from the routing table of FIG. 3 (described later). Then, in the case of using a wireless connection such as a portable information terminal, it is possible to check the connected device by broadcasting to the network.

Further, the application process 24 is
It is composed of a main process 20 and a work agent 26. The main process 20 mainly controls the processing. The work agent 26 is grouped by function, has various methods, and provides a service by being called by the main process 20. The work agent 26 is movable and moves to another device having a light load via the network when the load on the server becomes heavy.

Then, the main process 20 is executed as shown in FIG.
Since the address of the work agent 26 is obtained from the reference table of 1 and the method is called, the processing logic does not change when the work agent 26 is local or remote.

Next, FIG. 3 shows a routing table that can be obtained from the router. In the case of the configuration of FIG. 1 described above,
The device processing capacity and load list creation process acquires the destination IP address of the device and device in the range of the administrator from this table. Even in the case of wireless connection of a mobile information terminal or the like, it is easy to obtain the address of the connected device. The address information of these connected devices is attached to the load measuring agent 30 as a list.
Circulates among the devices on the network according to the list information.

Next, FIG. 4 is a list of equipment processing capacities and loads reported by the load measuring agent 30.
It is used to determine the transmission destination of the work agent 26 when the CPU load of each terminal becomes heavy or when it is desired to suppress power consumption. In the case of FIG. 1 described above, the transmission destination of the work agent 26 is determined in consideration of the CPU load and the processing speed of the candidate device. The condition of a device that retains the above power may be added to the determination.

Next, FIG. 5 is a reference table for managing the transmitted work agent 26. here,
The work agents 26 are classified into objects 1 to 4. Object references can be used to cause local and remote objects to perform call processing. Also, “transmitted” indicates that the transmission is completed and the remote object can be used. CPU load
It is assumed that the greater the value, the greater the effect of the remote agent on load reduction.

In FIG. 5, objects 1 to 3 remain in the local server, and object 4 shows a state in which transmission to the remote has been completed. Then, by using the reference address of pc1_obj4, the remote method is used for the object 4.

Next, FIG. 6 is a flow for creating a device processing capacity and a load list. This process is executed as a low-priority process independent of the application process on the terminal capable of executing the application. First, as shown in FIG. 6, the load of the own terminal is measured in S1, and if the load is heavier than a predetermined load, the processes from S2 onward are performed. A migration destination list of the load measurement agent 30 is created in S2. Then, in S3, the agent is transmitted to the top address of the list.

After this, the agent voluntarily logs for a certain period of time, measures the load value, etc., reports the result to its own terminal, and moves to the next address in the list. After that, this is repeated. Next, in S4, this load measurement agent 3
Take in reports from 0. Then, in S5, it is determined whether or not the report is from the load measuring agent 30, and if the report is not from the load measuring agent 30, the process returns to S4.
Go to 6. In S6, the processing capacity and load list table is updated. Next, in S7, it is determined whether there is a transmitted object in the reference table. If not, the process returns to S4, and if there is a transmitted object, the process proceeds to S8. In S8, a report is sent from the load measurement agent 30 at regular intervals, so that the CPU load information in the reference table is updated each time.

Next, in S9, the load on the terminal is judged, and if the load is light for a certain period of time, the process proceeds to S10 and the activity of the load measuring agent 30 is stopped. And
In step S11, the load measurement agent 30 stop flag is turned on. This flag can be referred to by the application process as well, and the application process determines that the load on its own terminal is not heavy, and all processing is performed using the local work agent 26. On the other hand, if it is determined that the load is heavy in S9, the process proceeds to S12, it is determined whether the load measurement agent 30 is stopped, and if the load measurement agent 30 is stopped, the load measurement agent 30 is instructed to start moving in S13. To restart, and then in S14, the load measurement agent 30 stop flag is set to O.
Set to FF. On the other hand, if the measurement agent 30 has not stopped in S12, the load measurement agent 30 in S4
Return to the process of fetching the report from. The device processing capacity and load list here, and the reference table are used to determine the delivery destination of the work agent 26 in the later application process.

Next, FIG. 7A is a processing flow of the load value measuring agent. When this is transmitted by the application terminal and becomes executable, first, in S22, the computer name is read, and in S24, information such as available memory, network I / O usage status, and remaining power is acquired. When the computer name is acquired, the processing speed information (MIPS) is transmitted to the application terminal.
Value). Next, in S26, a log is taken at regular intervals and the load on the processor is obtained as an average value. Subsequently, in S28, the results are collectively reported to the application terminal, and in S30, the process moves to the next device in the move list. After that, S22 to S30 are repeated. If the device cannot be moved even after several retries, the address of the device is removed from the transfer list and the application terminal is also notified. Based on this result, the device processing capacity and load list shown in FIG. 4 and the reference table shown in FIG. 5 are updated.

Next, FIG. 7B is a processing flow of the stop and restart commands received from the device processing capacity and load list creation process. First, these commands are received by interruption, and the commands are determined after reception. In S32, it is determined whether it is a movement stop command, and if it is a movement stop command, S
In step 34, it is determined whether the load value measuring agent is moving. When the load value measuring agent is moving, the process proceeds to S36 and the process is stopped and terminated. On the other hand, if it is not the movement stop instruction in S32, the process proceeds to S38, and it is determined whether or not the movement start instruction. If it is not a movement start command, the process ends. If it is a movement start command, the process proceeds to S40, and it is determined whether or not it is stopped. If it is not stopped, the process ends. If it is stopped, the process proceeds to S42, starts the process, and then ends.

Next, FIG. 8 shows a processing flow of the application process. First, in the flow of FIG.
At 50, the load value of the own terminal is measured. Then, the process proceeds to S52, and it is determined whether the load is heavy. If the load is heavy, S
In step 54, the agent is transmitted to another device. The procedure for transmitting an agent to this other device is shown in FIG.

In FIG. 9, by checking the load value of the reference table, the work agent having a large transmission effect is selected in S80. If the work agent selected in S80 has not been transmitted, the process proceeds to S88 in order to perform the transmission process after S88. On the other hand, if it has been transmitted, the process proceeds to S84, the load value of the transmission destination is checked, and if the load is large, it is necessary to send a copy of the same work agent to another device, so the transmission process from S88 is performed. If the load of the transmission destination is small in the determination of S84, it is not necessary to transmit and the work agent already transmitted is used.

Then, in S88, a device on the network having a sufficient memory, network I / O, power, etc. is selected. Then, in S90, the processing speed * (1-processor load) is calculated from the devices selected in S88 to determine the device having the highest CPU capability. Then, in S92, the work agent is transmitted to the device having the highest CPU capability. Then, the process proceeds to S94, the transmission completion is confirmed, and if the transmission is completed, S96
Then, the reference table such as the address of the object reference and the transmitted flag setting is updated.

Returning to FIG. 8A again, in S56, the request queue from the client is taken out, and if there is no request queue, the process returns to S50. If there is a request queue, S
Proceeding to 58, the process is taken out from the process request queue and the process is performed. Then, the process proceeds to S60, where a new processing process is created and the request is processed.

Next, FIG. 8B shows the processing of the process. First, in S62, the address of the reference table is acquired. Then, the process proceeds to S64, the load of the own terminal is checked, and if the load is light, the reference of the work agent is locally set (S68). On the other hand, if the load is heavy,
In S66, the addresses of all work agents are acquired. At this time, if the object has not been transmitted in the reference table, the local object reference is obtained, and if it has been transmitted, the remote reference is obtained. When there are multiple transmission objects, the object reference for the device with a light processor load at the transmission destination is acquired. This can be done by comparing the processor load values, which are described in the reference table of FIG.

Then, the process proceeds to S70, where the request processing from the client is performed, but since the address of the work agent has already been determined in S66 and S68, the case where the work agent is local to the server and the other terminal Processing logic is the same as when moving to. Then, the completion of the process is determined in S72. If not completed, the process returns to S70. When the process is completed, the result is returned in S74, the process is deleted in S76, or the process is pooled in the system and terminated.

Next, FIG. 10 shows the processing forms of the application process and the work agent. As shown in FIG. 10, the application process asks the server 40 to call a method of the work agent 41 to execute a part of the function, and is told the processing result.
This is the CPU originally required for the application execution terminal
It consumes resources such as time and memory,
The load on the application terminal is reduced. Also,
In the server 40, when there are requests from a plurality of clients, the processes are proceeding with parallel processing, and the load is reduced in each process, and as a result, the response time is shortened.

Next, FIG. 11 shows the configuration of a plurality of server terminals. The first server (server 1) operates with the same configuration and procedure as described above. When there are second and subsequent servers (server 2), as shown in FIG. 11, the device processing capacity and load list creation process and the load measurement agent 30 are deleted from the server 2, and the main process 34 The work agent 36 does not execute the device processing capacity and load list creation process flow shown in FIG. 6, and the work agent 3 is managed by the management table shared with the server 1 by another terminal connected to the network.
6 is sent to cause the work agent 36 to execute a part of the application function, thereby distributing the load.
Further, the configuration of the server 1 and the server 2 can be performed by the configuration at the time of setting up the terminal.

The server 2 may have the request from the client be distributed from the server 1 or may be another independent access point. Further, the processing flow of the main process 20 is the same as the flow described for the server 1 except that the management table (tables in FIGS. 3, 4, and 5) is not local but refers to the server 1. That is, when the load on the server 2 is not large, the local work agent 26 is used to process the request from the client. When the load increases, the management table of the server 1 is referenced to use the work agent 26 that is moving remotely, or a copy of the work agent 26 is transmitted to another device. Further, when the device load of the transmitted work agent 26 is heavy, another device is searched for and a copy of the work agent 26 is transmitted. Further, since the management table may be simultaneously accessed by a plurality of application execution terminals, simultaneous access does not occur by using an exclusive control mechanism such as locking and unlocking. This can be done by calling a system call of the OS, but it is not described in the flow because it is common in the multi-process mechanism. A series of these procedures is the same as the flow of the server 1.

Note that in this system, the agent is transmitted to another terminal and executed, so security must be taken into consideration when performing processing relating to the password and encryption / decryption. In this regard, Java (registered trademark)
The agent's encryption technology discussed in Section 3.1.1 may be used. Further, since there is an envelope protection technique in the conventional technique, it can be used. This is protected by scrambling the application file and executed while decompressing, so it can be protected during agent transmission / reception and execution.

As described above, by using the embodiment of the present invention, one processing engine is decomposed into a main process for controlling and a movable work agent, and the work agent is moved to another device when access is concentrated. This avoids concentration of load. With this method, the load can be distributed and the response can be improved by moving the work agent to another device except when the access is extremely concentrated. In addition, it is possible to use it for load distribution even if it is not necessarily a terminal. For example, a work agent can be sent to a device such as a printer or a FAX machine equipped with a CPU to distribute the load. Also, if the load on the entire device connected to the network is not heavy when the load on the terminal increases, we will provide a device that can handle it without increasing the number of processing engines by letting the terminal with a light load process. You can do it. Therefore, when the load increases, the load can be distributed and the response can be improved without increasing the cost of purchasing a new terminal and man-hours such as software setup. In addition, the load can be distributed even in the sequential processing, and it is possible to deal with the case where the load on both the server and the client is heavy.

Further, even when power consumption is concerned in a portable information terminal or the like, power consumption can be saved by requesting work to another terminal having a stable power source, and further, the same application processing is performed. When configured with multiple terminals, information can be shared with each other and cooperative operation can be performed, and load distribution can be effectively performed.

[0043]

EFFECTS OF THE INVENTION By using the present invention, it is possible to provide a load distribution system in which parallel tasks and load distribution can be performed by sending work agents even in sequential processing, and no setup is required for each work terminal in advance. it can.

[Brief description of drawings]

FIG. 1 is a diagram showing a device configuration when an embodiment of the present invention is used in a LAN.

FIG. 2 is a diagram showing a configuration of an application process included in a terminal having an application execution function according to the embodiment of the present invention.

FIG. 3 is a diagram showing a routing table that can be acquired from the router according to the embodiment of the present invention.

FIG. 4 is a diagram showing a device processing capacity and a load list reported by a load measurement agent according to the embodiment of the present invention.

FIG. 5 is a diagram showing a management table of a transmitted work agent according to the embodiment of the present invention.

FIG. 6 is a flowchart showing a flow of creating a device processing capacity and a load list according to the embodiment of the present invention.

FIG. 7 is a flowchart showing a processing flow of a load value measuring agent and a processing flow of a stop / restart command received from a device processing capacity and load list creation process according to an embodiment of the present invention.

FIG. 8 is a diagram showing a processing flow of an application process according to the embodiment of the present invention.

FIG. 9 is a diagram showing a procedure for transmitting an agent according to the embodiment of the present invention.

FIG. 10 is a diagram showing processing forms of an application process and a work agent according to the embodiment of the present invention.

FIG. 11 is a diagram showing a configuration of a plurality of server terminals according to the embodiment of the present invention.

[Explanation of symbols]

10, 11, 12 ... Terminal, 13 ... Server terminal, 15 ... Disk device, 14 ... Printer, 16 ... Router, 20, 3
4 ... Main process, 22 ... Load list creation process, 2
4 ... Application process, 26, 36, 41 ... Work agent, 30 ... Load measuring agent

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] December 18, 2001 (2001.12.
18)

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be amended] Claims

[Correction method] Change

[Correction content]

[Claims]

[Procedure Amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0008

[Correction method] Change

[Correction content]

The means for executing the application function by the work agent executes a part of the application.

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0010

[Correction method] Change

[Correction content]

Further, the work agent is sent out,
A terminal for holding information of the load measuring receiving said working agent is composed of a terminal that Ru is executed in accordance with an instruction from the terminal for holding information of the load measuring, the other terminals connected to the network Load distribution is performed by transmitting the work agent and causing the work agent to execute a part of the application function.

[Procedure amendment 4]

[Document name to be amended] Statement

[Name of item to be corrected] 0024

[Correction method] Change

[Correction content]

Next, FIG. 5 is a reference table for managing the transmitted work agent 26. Here, the work agents 26 are objects 1 to 4
It is classified into. Object references can be used to cause local and remote objects to perform call processing. In addition, transmission has been completed,
Indicates that the remote object can be used. The load is C
It is assumed that the larger the value of PU consumption, the greater the effect of reducing the load by the remote agent.

[Procedure Amendment 5]

[Document name to be amended] Statement

[Name of item to be corrected] 0027

[Correction method] Change

[Correction content]

After this, the agent voluntarily logs for a certain period of time, measures the load value, etc., then reports the result to its own terminal, and moves to the next address in the list. After that, this is repeated. Next, in S4, the load measuring agent 30
Capture reports from. Then, in S5, it is judged whether or not the report is from the load measuring agent 30, and if it is not the report from the load measuring agent 30, the process returns to S4,
If the report is from the load measuring agent 30, the process proceeds to S6. In S6, the processing capacity and load list table is updated. Next, in S7, it is determined whether there is a transmitted object in the reference table. If not, the process returns to S4, and if there is a transmitted object, the process proceeds to S8. In S8, a report is sent from the load measurement agent 30 at regular intervals, so the processor load information in the reference table is updated each time.

Claims (7)

[Claims] 1. A system comprising a plurality of terminals connected by a network, wherein the control means controls an application, and a work agent movable in the network executes the application function. A means for measuring the load of a terminal on the network using a patrol agent that patrols the network; and when the terminal is in a predetermined state, the work agent is transmitted to another terminal of the network based on the result of the load measurement. And a means for causing the work agent to execute the application function. 2. The load balancing system according to claim 1, wherein the work agent executes the part of the application by the application function executing means. 3. The load according to claim 1, wherein the predetermined state of the terminal is a state in which the load on the terminal is increased or a state in which the remaining power of the terminal is less than or equal to a predetermined value. Distributed system. 4. A terminal configured to send out the work agent and hold the load measurement information and a terminal to receive and execute the work agent, according to an instruction from the terminal holding the load measurement information. ,
The load balancing system according to claim 1, wherein the work agent is transmitted to another terminal connected to the network, and the work agent executes a part of an application function to perform load balancing. 5. The load balancing system according to claim 1, wherein a plurality of terminals for receiving and executing the work agent are connected to the network. 6. A load balancing method comprising a plurality of terminals connected by a network, the control step of controlling an application, and the step of a work agent movable in the network executing the application function. And a step of measuring a load of a terminal on the network using a patrol agent that patrols the network, and when the terminal is in a predetermined state, the other terminal of the network is connected to another terminal of the network based on a result of the load measurement. Transmitting a work agent and causing the work agent to execute the application function. 7. A control process for controlling an application when using a load distribution system composed of a plurality of terminals connected by a network, and a process for a work agent movable in the network to execute the application function. A process of measuring a load of a terminal on the network using a patrol agent that patrols the network; A program including instructions for transmitting a work agent and causing a computer to perform a process of causing the work agent to execute the application function.