JPH10154075A - Address management method of moving agent and moving agent device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make each moving agent cooperate and manage the latest address (identifier and an environmental name) of a moving agent which freely moves among environments on a computer network. SOLUTION: Moving agents 105 to 108 have a fixed agent which does not move for their ancestor and have an absolute path name which consists of the address of a fixed agent and a label string which tracks only labels of a parentage from the fixed agent and reaches their own moving agents the themselves. A name server function 114 manages absolute paths and addresses 118 to 120 of agents which become its own direct parent and child with a label named attached to them. When it moves between environments, a communication function 113 notifies agents which become its direct parent and child of a new address which is acquired in an environment after movement. When a communication which uses a certain address fails, an identifier and an environmental name which corresponding to the label name are successively inquired of other agents along the absolute path name of the communication object, the latest address is acquired and communication of performed again.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention belongs to the field of implementing network services such as e-mail, workflow management, electronic shopping, and travel reservation with a mobile agent, and particularly relates to a mobile agent address management method and a mobile agent device.

[0002]

2. Description of the Related Art As a method of performing remote programming in a computer network to which a plurality of computers are connected, an apparatus called a mobile agent which moves between different computers while executing a program while executing the program is disclosed in Japanese Patent Application Laid-Open No. Hei 7 (1995) -107. It has been conventionally proposed in, for example, JP-A-182174 (hereinafter referred to as Document 1).

In a mobile agent device (sometimes simply referred to as a mobile agent), the operation of a computer process is stopped while the computer process is running, and all procedures and states are transferred to another computer. By reconstructing the process equivalent to that before the transfer and restarting the execution, the computer process can be moved between computers. This moving computer process is called a mobile agent. The mobile agent freely moves around on the network as needed, uses various resources on the network, and executes various processes given while communicating with other mobile agents.
By using a mobile agent, a device composed of two distributed modules that require a large amount of communication between them, etc., one of which moves to the same computer as the other and performs communication, thereby reducing the amount of communication on the network At the same time, stable and high-speed communication can be performed by the device via an unstable and low-speed line such as a wireless line. In addition, since it is possible to access resources existing in various places by one process, it is possible to freely design a device without being restricted by the distributed arrangement of resources on a network.

[0004] In order for a user or a mobile agent to communicate with a certain mobile agent, it is necessary to specify a communication target by some method. In order to specify a communication target, information for searching for and specifying a communication target is required. In the mobile agent device described in Document 1, only the location information of the target is specified as such information, and communication occurs only between agents in the same environment. I can't. In a general computer system, a name is used as such information, and a name management device is used to associate the information with an entity to be communicated. A name is a sequence of one or more characters that is created so that one name always determines one agent as information that uniquely determines the communication object.

In a situation where there are a large number of mobile agents on a large computer network to which a large number of computers are connected and they move freely around the network, the name uniquely determines the agent on this large computer network. There is a need. In such a case, it is extremely difficult to manage the entire network with only one name management device and maintain a one-to-one correspondence between names and entities of mobile agents. As a solution to such a problem, a method has conventionally been used in which a name management device is configured by a plurality of naming systems, and a compound name obtained by combining the names determined by the individual naming systems is used as a unique name. . JP-A-5-27
Japanese Patent No. 4274 (hereinafter referred to as Document 2) provides an apparatus and a method for configuring a name management apparatus using an arbitrary number of different naming systems as a name management method in an object-based distributed system. This allows a compound name to be constructed by a number of heterogeneous naming systems and to be easily decomposed.

When defining a device in which a plurality of mobile agents are associated with each other and communicate with each other, it is necessary to embed information necessary for specifying an object in the definition. In the mobile agent device described in Document 1, the target is defined based on information on the location of the target. Since the location information is unique on the network, different agents having the same definition of the target will always communicate with the same target. Also, in the case where a target is specified using a unique name on a network that is generally performed, similarly, different agents having the same information as the definition of the target always communicate with the same target.

[0007]

When communication is performed between mobile agents, since a communication target always has the possibility of moving, the method of specifying the target only by its address guarantees the success of specifying the target. Can not do. FIG. 22 shows a state in which, when the target is specified by the environment name, the communication fails when the communication target moves to a different environment.

[0008] Therefore, an object of the present invention is to provide a communication system in which even if the address changes due to movement of a communication target,
The purpose is to ensure that the latest address can be obtained.

One way to achieve such an object is to
It is common to provide a name server to obtain the address of the entity from the name, but if you want to manage such information using a single name server, you must assign a unique name on the network. And it is difficult to avoid contradictions such as duplication of names. In addition, the load is concentrated on the name server.

It is another object of the present invention to provide each mobile agent with a partial name server function,
Each mobile agent cooperates with each other to realize one name server as a whole, thereby distributing the load.

In the name management device described in the above document 2, a unique name on the network is constituted by a compound name of a plurality of different naming systems, and means for decomposing the name is provided. If it is applied to the address management, the name server will always be queried when communicating with the target. In the address management of the mobile agent, it is sufficient to make an inquiry only after the movement has occurred, and it is wasteful to inquire every time. Further, in the mobile agent device described in Document 1, communication between mobile agents having different locations cannot be performed in the first place.

Therefore, another object of the present invention is to prevent wasteful inquiring of the address to be moved every time.

Further, in the conventional method, the information for specifying the agent always indicates the same object regardless of the interpreter, so that it is impossible to specify the communication object based on the relative relationship between the agents. For example, if the definition "call the parent" is included in the definition, the parent to be communicated with should originally be different depending on the agent who is the interpreter of the definition. Such an interpreter cannot specify a different communication target, and all agents having such a definition communicate with an agent having a unique name of “parent” on a name server. I have.

Therefore, still another object of the present invention is to allow an agent to be specified based on a relative relationship with the user.

[0015]

A mobile agent address management method according to the present invention is a mobile agent address management method capable of moving between arbitrary environments on a computer network composed of a plurality of environments. a) As a condition for generating a mobile agent, a condition that an ancestor always has a fixed agent having no mobile function is given, and an identifier and an environment name of the fixed agent of the ancestor are given to the mobile agent generated under the condition. And an absolute path name consisting of a label sequence from the fixed agent of the ancestor to the parent agent by tracing only the label of the parent-child relationship, and (b) for each mobile agent, the direct parent, Label the information consisting of the absolute path, identifier, and environment name of the child agent (C) For the mobile agent that has moved between environments by itself, the new agent and environment name obtained in the environment after the movement are notified to the direct parent and child agents, and these direct agents are notified. The parent and child agents should always be able to manage the latest identifier and environment name of the moved mobile agent,
(D) If the mobile agent fails to communicate with the communication target using the identifier and the environment name, the mobile agent directly contacts the inquiry destination agent in order from the fixed agent determined by the identifier and the environment name in the absolute path name of the communication target. And specifying the label name of the child agent and inquiring about its identifier and environment name, and finally obtaining the current latest identifier and environment name of the communication target whose communication failed. I do.

Since the identifier of the fixed agent and the environment name are not changed, communication to the fixed agent always succeeds. Also, each mobile agent manages the latest identifiers and environment names for direct parent and child agents. Therefore, even when communication using a certain identifier and environment name fails, the search for the latest identifier and environment name of the communication target using the absolute path name always succeeds. For this reason,
Even when the address changes due to the movement of the communication target, the latest address can be reliably obtained.

Further, the present invention is characterized in that, when communication with a communication target is successful, an identification name, an environment name and an absolute path are obtained from the communication target and stored in the own agent. This prevents waste of inquiring the address of the movement target every time.

Further, according to the present invention, in the processing of the above (d), the own agent and the communication target that has failed to communicate have a common ancestor, and the path length of the absolute path of the communication target that failed communication is greater than the path length of the communication target that failed communication. If the path length of the alternative path that traces from the own agent to the common ancestor and traces the subsequent path from the middle of the absolute path to be communicated is shorter, label names are sequentially specified along the path of the alternative path. And inquiring about the identifier and the environment name, and finally obtains the current latest identifier and environment name of the communication target in which the communication has failed. Thereby, the process of searching for the identifier and the environment name of the communication target can be performed more efficiently.

Still further, according to the present invention, when a communication target is designated by a relative path which is a label string indicating a route from the own agent to the communication target agent, the labels are sequentially arranged along the path of the relative path. The system is characterized in that an identifier and an environment name are inquired by designating a name, and finally an identifier and an environment name of a communication target are obtained, and communication with the communication target is performed using the obtained identifier and environment name. As a result, it is possible to specify an agent based on a relative relationship with the user.

Further, the mobile agent device of the present invention has an agent main function for controlling the original operation of the agent such as information search, electronic shopping, reservation, an agent generating function for generating another mobile agent, and a mobile device for moving to another environment. Function, a communication function that interacts with other agents, and a name server function that retains information about other agents. From a fixed agent or another mobile agent, the ancestor has a fixed agent that does not have a mobile function. An absolute mobile agent device that is generated and includes an identifier and an environment name of an ancestor fixed agent, and a label sequence from the ancestor fixed agent to only the parent-child relationship label and arriving at its own agent. Has a path name and the name server Function is immediate parent, the full path of the agent that is a child, the information consists of the identifier and the environment name, a database for managing with a label,
A control function of notifying an inquiry source of an identifier and an environment name in information corresponding to the label name in the database in response to an inquiry specifying a label name from another agent, wherein the communication function is When the own agent moves between the environments by the function, the new identifier and the environment name obtained in the environment after the movement are notified to the direct parent and child agents managed in the database. When receiving the latest identifiers and environment names from the direct parent and child agents, the control unit reflects them to the database through the control function, and communicates with the communication target using the identifiers and environment names. If the direct communication function to perform and the communication by this direct communication function fails, the identifier and environment name in the absolute path name of the communication target are used. In order from the circle fixed agent,
A search path that specifies the label name of the direct child agent of the inquiry destination agent and inquires about its identifier and environment name, and finally obtains the current latest identifier and environment name of the communication target whose communication failed. It has a developing function.

In the mobile agent device having such a configuration, when the mobile agent moves between environments, the communication function becomes new with respect to the agent directly acting as a parent and child of the own agent in the environment after the movement. Notify the identifier and environment name,
The parent and child agents update their own databases. When the direct communication using the identifier and the environment name fails, the search path expanding function searches for the latest identifier and the environment name from another agent using the absolute path name of the communication target for which the communication has failed. Therefore, communication can be performed by attempting communication again using the retrieved identifier and environment name.

Further, the name server function is characterized in that an identification name, an environment name and an absolute path of a communication target which has succeeded in communication are stored if they are not stored in the database.

Still further, the communication function has a common ancestor for the own agent and the communication target that has failed to communicate,
In addition, it is determined whether or not the path length of the alternative path that traces from the own agent to the common ancestor and traces the path from the middle of the absolute path of the communication target is shorter than the path length of the absolute path of the communication target that failed to communicate. A search path minimizing function for examining and selecting the alternative path when the path is short is provided, and when the alternative path is selected, the search path expanding function executes the label name in order along the path of the alternative path. And inquiring about the identifier and the environment name, and finally obtaining the current latest identifier and environment name of the communication target in which the communication has failed.

Further, when the communication target is specified by a relative path, which is a label string indicating a route from the own agent to the communication target agent, the search path expanding function is performed along the path of the relative path. A configuration in which an identifier and an environment name are queried in order by designating a label name in order, and finally an identifier and an environment name of a communication target are obtained, and communication with the communication target is performed using the direct communication function by using the obtained identifier and environment name. It is characterized by having.

[0025]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention. Reference numerals 101, 102, 103, and 104 denote environments where agents on the computer network exist, and reference numerals 105, 106, 107, and 108 denote mobile agents existing on those environments.

As an embodiment, the environments 101 to 104 are realized by a computer system. Another embodiment is a virtual computer including a plurality of computers.

Each of the mobile agents 105 to 108 has, as exemplified in the mobile agent 105, information 110 relating to itself, in addition to an agent main function 109 for performing an agent's original operation such as information search, shopping, and reservation. An agent generation function 111 that has an identifier identifying itself as one in the environment, a current environment name, an absolute path of its own, and generates another mobile agent, a migration function 112 for moving to another environment, another Communication function 11 for interaction with mobile agent
3. It has a name server function 114 that holds information on other agents.

The identifier in the information 110 is information that is managed for each environment and uniquely identifies an agent in one environment. Every agent has only one identifier. As the identifier, an arbitrary identifier such as a sequence of one or more characters, a sequential number, a directory name and a file name of a file system, and the like according to the management method of the agent entity for each environment is used.

All mobile agents can be uniquely identified by an identifier and an environment name. However, the identifier and the environment name change when the agent moves between environments. That is, when the mobile agent moves to a new environment, the mobile agent acquires an environment name relating to the environment and an identifier that uniquely identifies the agent within the environment. On the other hand, the absolute path in the information 110 does not change even if the mobile agent moves between environments. This absolute path will be described later.

The name server function 114 has a database 117 and a database control function 115 therein.
Information 118, 119, and 120 in the database are information for specifying other mobile agents, each having a target identifier, an environment name, and an absolute path, and managed by attaching a label. Labels must uniquely identify other mobile agents within one mobile agent, but different mobile agents can assign the same label to different mobile agents, or conversely, assign different labels to the same mobile agent. You can also attach. The label expresses "how an agent recognizes and manages another agent" and can be said to express the relationship between an agent and another agent. This relationship is one-way. Information 118, 11 exemplified in FIG.
For 9 and 120, "parent", "child 1",
Labeled "Child 2".

The transfer function 112 realizes transfer of its own agent between environments by copying a process to a remote location, continuing its execution, and extinguishing its own process.

The communication function 113, as shown in FIG.
21, direct communication function 122, search path expansion function 123,
It has a search path minimizing function 124 and a receiving unit 125. The control unit 121 calls the direct communication function 122 and the search path development function 123 according to the information provided for specifying the communication target at the time of communication transmission. Search path expansion function 1
23 calls a search path minimizing function 124 to improve the efficiency of the search when possible. The indirect communication function is provided by the control unit 1
21, direct communication function 122, search path expansion function 123,
This is realized by the search path minimizing function 124.

The agent main function 109 is a part created by an agent designer for each agent, and defines the entire operation of the agent, such as information search, workflow management, and electronic shopping. The agent main function calls the agent generation function 111, the movement function 112, and the communication function 113 in a predetermined format as necessary, and controls the operation of the agent.

In FIG. 1, all the mobile agents hold their own mobile agent (hereinafter referred to as parent agent) and information regarding the mobile agent (child agent) generated by themselves in a name server. I do. First, when the mobile agent 105 is generated by the mobile agent 106, information 118 about the parent agent 106 is appropriately labeled in the database 117 of the name server function 114 of the mobile agent 105, and the name server control function 115
Registered by. Also, the agent main function 109 of the mobile agent 105 is
Is called to generate the mobile agents 107 and 108 as their own child agents, the agent generating function 111
17 shows information 11 about child agents 107 and 108
9 and 120 are registered with appropriate labels. Similarly,
The mobile agent 106 stores information about the mobile agent 105 as a child agent.
07 and the mobile agent 108 register information on the mobile agent 105 as a parent agent with an appropriate label in its own name server.

In addition to the parent-child relationship, the mobile agent can manage an identifier, an environment name, and an absolute path of any other mobile agent with a proper label and manage it by the name server. Such an agent can specify another agent only by specifying a label from the agent basic function. Also, if this label is made public to other agents, other agents can obtain information on the target agent by inquiring this agent by designating the label.

When an agent moves, the moving agent obtains a new identifier and environment name in the destination environment, and notifies the other mobile agents having a parent-child relationship of the new identifier and environment name at the same time as the movement. Agent main function 109 of mobile agent 105 is mobile function 1
Upon invocation of 12, the mobility function 112 gets the new identifier and environment name and notifies all parent and child agents 106, 107, 108 of the new information via the communication function 113. Upon receiving the notification, the mobile agents 106, 107, and 108 update the database of each name server through the control function. This allows all mobile agents to retain the exact identifiers and environment names of other mobile agents that have a direct parent-child relationship.

As described above, since all mobile agents always hold the correct identifiers and environment names of other mobile agents having a direct parent-child relationship, communication with them is always successful. However, since the exact identifier and environment name of the mobile agent other than the direct parent-child relationship are not managed, the communication using the identifier and the environment name for the mobile agent other than the direct parent-child relationship is not always successful. This is because the communication target is moving between environments, and the identifier and the environment name may have changed. For this reason, a special mechanism for obtaining the latest identifier and environment name of the mobile agent other than the direct parent-child relationship is required. Hereinafter, this point will be described.

In the present invention, an agent called a fixed agent is provided as a special mobile agent having no mobile function. The fixed agent has the same function as the mobile agent, except that the fixed agent does not move. Then, there is a constraint that all mobile agents must be generated so as to always have fixed agents as ancestors. This restriction does not apply to fixed agents.

Due to the above restrictions, all mobile agents have a hierarchical parent-child relationship with fixed agents as ancestors as shown in the example of FIG. The example of FIG.
01, its child agent 202, and its child agents 203 and 204, 204
5 shows a hierarchical parent-child relationship composed of Nos. 5 and 206. 20
7, 208, 209, 210, and 211 indicate that the agent 201 is 202, 202 is 203, and 20
2 is a label attached to 204, 204 is assigned to 205, and 204 is a label attached to 206. In addition, all agents having a parent agent are referred to the parent agent as "pa
Suppose that it has a label “rent”.

For an arbitrary mobile agent, the shortest path to the mobile agent by tracing only the label of the parent-child relationship from the fixed agent of the first ancestor of the mobile agent's parent-child relationship is defined as the absolute path of the mobile agent. Call. For example, the absolute path of the mobile agent 206 is the sum of the identifier and environment name of the fixed agent 201 and labels 207, 209, and 211 in the middle.

This absolute path is invariable even when individual mobile agents move between environments. The identifier and environment name of the fixed agent are not changed, and the correct identifier and environment name of the child agent are held by the parent agent. Therefore, if each mobile agent holds the information on the absolute path of the communication target, it can use the information to acquire the latest identifier and environment name information of the communication target. That is, for example, in FIG.
The operation in which the mobile agent 203 acquires the latest identifier and environment name of the mobile agent 206 based on the absolute path of the mobile agent 206 can be performed as follows.

The mobile agent 203 communicates with the fixed agent 201 using the identifier and environment name of the fixed agent 201 in the absolute path, and
Request expansion of. Since the identifier of the fixed agent 201 and the environment name are unchanged, this communication always succeeds. Since the fixed agent 201 manages the correct identifier and environment name of the child agent 202 in its own name server, the fixed agent 201 extracts the identifier and environment name in the information corresponding to the label 207, and
Notify 3. The mobile agent 203 is a fixed agent 201
The mobile station 202 communicates with the mobile agent 202 using the identifier and the environment name notified from the mobile terminal 202, and requests expansion of the label 209. In the mobile agent 202, the child agent 2
Since the correct identifier and environment name relating to the label 04 are managed by its own name server, the identifier and environment name in the information corresponding to the label 209 are extracted and notified to the mobile agent 203. The mobile agent 203 is a mobile agent 202
The mobile station 204 communicates with the mobile agent 204 using the identifier and the environment name notified from the mobile terminal 204, and requests the expansion of the label 211. In the mobile agent 204, the child agent 2
Since the correct identifier and environment name relating to 06 are managed by its own name server, the identifier and environment name in the information corresponding to the label 211 are extracted and notified to the mobile agent 203. Thereby, the mobile agent 203 can know the latest identifier and environment name of the mobile agent 206.

On the other hand, relations other than parent-child relations (information managed by labeling other than parent-child in the name server)
Does not update its information (identifier, environment name) even if it moves. For this reason, if the target moves, erroneous information will be stored. However, it is possible to always obtain accurate information by storing the absolute path information and following this.

In this embodiment, when a certain mobile agent communicates with another mobile agent, in addition to the specification using an identifier and an environment name, the specification using an absolute path, the specification using a label, and the specification using a relative path are possible. It is. That is, 1. Label 2. 2. Identifier and environment name Absolute path 4. The communication target can be specified by one of the relative path methods. The specified information is processed by the control unit 121 of the communication function, and a necessary function is called.

Here, both the absolute path and the relative path are collectively called a search path. The search path is information that specifies a route to reach a target agent from a certain agent by using a relationship between many agents. One example of a search path expression is “((identifier environment name) label 1 label 2... Label N)”.
The absolute path in the search path is, as described above, the shortest search path that reaches the mobile agent by tracing only the label of the parent-child relationship from the fixed agent of the first ancestor of the mobile agent's parent-child relationship. Yes, a relative path is a search path that is not an absolute path.

A search path is defined from another viewpoint with reference to FIG. FIG. 3 is a flowchart of an algorithm that receives an arbitrary identifier, an environment name, and a label string and outputs an identifier, an environment name, or failure. S1 and S2 are inputs, and S9 or S10 is an output. Hereinafter, the algorithm will be described. First, an arbitrary agent identifier and environment name are designated as an initial inquiry destination in S1, and an arbitrary label string is given in S2. Next, in S3, the label name to be expanded is set at the head of the label string, and in S4, the name server function of the agent set as the inquiry destination is inquired about the expansion of the label set in S3. If the inquiry succeeds, the information (identifier, environment name) of the deployed agent is returned as a result, and if the inquiry fails, the failure is returned as a result. The inquiry was successful in S5,
If the label string is not empty in S7 even after the head of the label string is removed in S6, the result obtained in the inquiry in S8 is set as the next inquiry destination, and the process returns to S3. This is S5
Is repeated until the query fails or the label string becomes empty in S7. If the query fails in the middle, a failure is output in S9, and if the label string becomes empty, the result obtained in the last query is output in S10. I do. A search path is defined as follows using this algorithm.
It is assumed that in a certain computer network, there is one piece of information including an identifier, an environment name, and a label string, and one agent specified by another identifier and an environment name. Here, when the information is given as an input of the algorithm, and an identifier and an environment name for specifying the agent are output as a result, the information is called a search path of the agent. Note that the length of the search path is defined as the number of necessary inquiries, that is, the number of labels included in the search path.

FIG. 5 is a flowchart of the operation of the control unit 121 when communicating with another mobile agent. If a label is first specified in S11, the label is expanded by the name server function 114 in S15, that is, information in the database 117 corresponding to the label is searched, and the obtained environment name, identifier and absolute path are stored in S15.
Used from 12 onwards. If an incorrect label is given, deployment fails in S16 and communication fails. If the identifier and the environment name are given in S12, the direct communication function 122 is called in S17, and if this is successful, the communication is successful. In S24, the result is output and the processing ends. On the other hand, if the direct communication cannot be performed or has failed, the absolute path is checked in S13, and if given, the search path development function 123 is called in S19 with the absolute path as the search path. If the absolute path has not been given and the relative path has been given in S14, the relative path is set as the search path in S2.
When 0, the search path development function 123 is called. If the search path has been successfully expanded in S21, the result is used in S22.
To communicate directly. If the communication is successful, the communication is successful in S24. If the communication is unsuccessful, the communication failure is output in S25.
If neither the absolute path nor the relative path is given, or if the development fails, further search becomes impossible, and a communication failure is output in S25, and the processing ends.

When the designation by the identifier and the environment name is performed, the direct communication function 122 of the communication function 113 directly connects to the target environment and the mobile agent thereon. That is, the direct communication function 122 accesses the specified environment, checks whether an agent having the specified identifier exists, and if so, passes the communication content to the receiving unit of the communication function of the agent and succeeds. If not, it returns failure as a result. In the specification by the identifier and the environment name, the target can be specified without the search cost as long as the target mobile agent does not move. However, if the target moves after obtaining the information of the identifier and the environment name, the communication fails.

In the specification by the absolute path, in order to obtain the correct identifier and environment name of the target, inquiries must be made in order from the fixed agent of the target ancestor to the target mobile agent by following the parent-child relationship. Search cost is higher than direct specification by environment name. However, unlike the direct specification by the identifier and the environment name, the search by the absolute path always succeeds and can communicate with the target.

The specification by the relative path is described in 3. Requires the same search as the specification by the absolute path, so the same search cost is required. Furthermore, since a search using a relative path includes a relationship other than a parent-child relationship, it is not always the case that the target is successfully specified. However, in the specification using the relative path, it is possible to specify an object based on a wide range of relative relationships using the knowledge of the agent and other agents around the agent and the relationship between agents.

FIG. 6 is a flowchart showing the operation of the search path expanding function 123. The search path expansion function 123
Based on the specified search path, the communication target identifier and environment name are searched as follows. First, in S30, the environment name and the identifier included in the search path are set as the inquiry destination agent, and in S31, the label to be queried is set at the head of the label string of the search path. Then, in S32, the name server function of the inquiry destination agent is inquired about the label development, and if successful, the result obtained in S35 is set as the next inquiry destination, and the label to be inquired is set in S36 as the next label in the label sequence. Repeat from
If all the labels have been successfully developed in S34, the last obtained result is output in S37, and the process ends. If the development fails in S33 before the last label is reached, a failure is output in S38 and the process ends. FIG. 7 is an explanatory diagram showing the operation of the search procedure.

By the way, in the case of the absolute path search, if the agent to be searched and the agent to be communicated have a common ancestor, the number of inquiries required for the search can be reduced depending on the relationship with the object. For this reason, in the present embodiment, the search path expanding function 123 calls the search path minimizing function 124 before starting the search, and attempts to minimize the search path if possible. Agents with a common ancestor
Two methods are possible: a method of searching according to an absolute path, and a method of tracing from oneself to the first common ancestor and tracing thereafter from the middle of the absolute path. Since both methods follow only the parent-child relationship, both search possibilities are guaranteed.

Here, it is assumed that the length of the search path X is represented by | X |. The absolute path of the searching agent itself is A,
Let B be the absolute path of the agent to be searched, and C be the absolute path of the longest absolute path among the common ancestors of the two agents. Since the number of inquiries required to retrieve the remainder of B from C is the same in either method, the number of inquiries required to inquire of C for accurate information becomes a problem. The number of searches required to search C is as follows: when following the absolute path | C |, when following the parent agent with itself as the search start position | A
Since |-| C |, | A |-| C |> as shown in FIG.
When | C |, that is, when | A |> 2 | C |, the search is performed according to the absolute path B, and | A | − | C | ≦ | as shown in FIG.
In the case of C |, that is, when | A | ≦ 2 | C |, if the search is performed by tracing the parent agent with itself as the search start position, the number of inquiries required to search for the same agent is reduced. be able to.

The search path minimizing function 124 attempts to minimize the search path according to the above method. FIG. 10 is a flowchart showing the operation of the search path minimizing function 124. First, in S51, two absolute paths to be compared are referred to as “A
= Absolute path of own ”and“ B = Absolute path of target ”and check in S52 whether they have a common ancestor.If they have a common ancestor, in S53,“ C = A, common ancestor of B ” Among them, it is assumed that the absolute path length is the longest, and | A |> 2 | C | is checked in S54. In the case of | A | ≦ 2 | C |, “own identifier and environment name” + “| A |-| C
The | labels indicating the parents] + “the remaining label sequence obtained by removing C from B” are output as the path to be searched in S55. | If the ancestors are not common or they are common, | A |> 2 |
In the case of C |, the absolute path B of the target which is the original search path is output as it is in S56. By using the output of the search path minimizing function as an input of the search path expanding function, the search path can be minimized.

In the communication object managed in the name server by the label, if the communication with the object succeeds in some way, the identifier, the environment name, and the absolute path are obtained from the communication object, and the name is stored in the name server. Update the applicable information in. Since the name server stores the target identifier, environment name, and absolute path, if communication is successful even once, the target is identified by the low-cost identifier and environment name as long as the target does not move. If this is the case, the target can be specified by the absolute path, so that an efficient target search can be performed. If the target is specified by a relative path, which is not always guaranteed, once communication with the target is successful, the identifier, environment name, and absolute path obtained at that time are saved and the target is specified. I do. This makes it possible to minimize the search cost.

[0057]

【Example】

Embodiment 1 An embodiment of the present invention will be described with reference to FIG. FIG. 11 shows a relationship diagram of the agent at a certain moment. 302,
303, 304, 305, 306, 307, 309, 3
10, 311 and 312 are mobile agents, and 30
1 and 308 are fixed agents. The solid arrow in the figure represents the generation relationship, and the label 3 attached to the arrow is shown.
Reference numerals 13 to 322 represent labels assigned to the child agent by the parent agent. It is also assumed that all child agents have labeled the parent agent "parent". 323 is Agent 3
Reference numeral 10 denotes a label attached to the agent 305. This is a relationship other than the parent-child relationship.

It is also assumed that the identifiers and environment names of the agents 301 to 312 at a certain moment in FIG. 11 are given as shown in FIG.

At this time, the absolute path of each mobile agent is “((ancestor identifier ancestor environment name) label 1 label 2
... Label n) ”is determined as shown in FIG.

Here, each agent holds information on an agent having a direct parent-child relationship with itself. For example, the mobile agent 305 includes the agents 302 and 30
6, 6307 is stored in the database in the name server.

Information on agents having a relationship other than the parent-child relationship is also stored. For example, the mobile agent 310 includes the agents 309 and 3 having a parent-child relationship.
11 and agent 30 in a relationship other than parent-child relationship
15 is stored in the database in the name server.

Here, it is assumed that the mobile agent 305 on the environment “Env1” moves to the environment “Env4” as shown in FIG. At this time, the mobile agent 305 moves to the environment “Env4”, acquires a new identifier “ag4” on the environment “Env4”, and
become. The mobile agent 401 is a process equivalent to the mobile agent 305 except for the identifier and the environment name. Thereafter, the mobile agent 401 stores the new identifier and environment name in the agent 302,
306, 307, and updates the data relating to the user in the name servers 302, 306, 307.

The notification at the time of movement is given only to agents having a direct parent-child relationship. For example, the mobile agent 310 holds information on the mobile agent 305 in the name server, but information on the mobile agent 401 which is a new entity of the mobile agent 305 is not notified to the mobile agent 310. When the mobile agent 310 communicates with the mobile agent 305, it is possible to directly connect to the target by specifying the identifier “ag1” and the environment name “Env1” until the mobile agent 305 moves. From the mobile agent 310 after the mobile agent 305 has moved to become the mobile agent 401, the mobile agent 305
The first communication to (401) begins with the old identifier "ag
The connection is attempted by specifying the environment name “Env1” and the environment name “Env1.” At this time, the first connection fails because the mobile agent 305 does not already exist. (40
1) Absolute path "((Controller Env1)
Secretary scheduler) "
The correct information of the mobile agent 305 is obtained by the following procedure.

1. The name server of the fixed agent 301 having the identifier "Controller" and the environment name "Env1" is inquired with the label "Secretary" to obtain accurate information of the mobile agent 302. 2. The obtained information on the mobile agent 302, that is, the identifier “ag2” and the environment name “Env1” are designated, and “sch
Inquiry with the label of “eduler” to obtain accurate information of the mobile agent 401. 3. Update the part of the name server of the mobile agent 310 relating to the mobile agent 305 to the information of the mobile agent 401 as shown in FIG. Connect to the mobile agent 401 by specifying the identifier “ag4” and the environment name “Env4”.

Thereafter, as long as the mobile agent 401 does not move, it is possible to directly connect to the mobile agent 401 using the newly acquired information identifier “ag4” and environment name “Env4”.

Embodiment 2 Another embodiment will be described with reference to FIG. 501,50
2, 503, 504 and 505 are class definitions of a group of agents for executing a certain task. Reference numeral 501 generates child agents 502 and 503, and manages information by attaching labels to “Task1” and “Task2”, respectively. Generated agents 502 and 5
03 generates 504 and 505 as child agents, respectively, and attaches a label “Subtask” to them to manage information. Agents 502, 503, 5
04 and 505 operate as independent agents, however, during operation, the agent 502 and the agent 503 mutually operate, and the agent 504 operates as the agent 5
03, the agent 505 is the agent 50
4 is communicated.

In order to define an agent group that performs such an operation, the agents 502, 503, 504, 5
05, it is necessary to provide information for specifying each communication target. By specifying a relative path as this information, a relative relationship is described. Relative paths all take the form of a search starting from oneself and following the parent. More specifically, if oneself is represented by "()", each communication target can be represented as shown in FIG.

As an example in which a plurality of instances are generated based on the definition of the same class group, consider an example in which this agent group is periodically generated by a specific agent. The relationship between such agent instance groups is as shown in FIG. An agent 601 periodically (for example, once a month) generates an instance of an agent group, and 612 and 613 are generated instance groups. 602 and 607 are 501, 603 and 608
Is 502, 604 and 609 are 503, 605 and 61
0 is 504, and 606 and 611 are instances of 505, and have relative paths related to the communication target given in the class definition.

The agents 603 and 608 of the same class use the common relative path “(() parent Task”
However, since the search start position is itself, the communication target 603 is 604, 6
08 is different from 609. Similarly, 604,
Relative paths of 605, 606 and 609, 610, 6
11 have the same relative path, but have different targets for actual communication. As described above, the agent instance groups 612 and 613 all have information on a common communication target, but the actual communication is performed in the same instance group and can act as a completely unrelated agent group.

Conversely, if specified, it is possible to communicate with agents included in different instances.
For example, if the agent 611 included in the instance group 613 attempts to communicate with the agent 605 included in the instance group 612 in order to check information on the task of the last month, if only the rule of the label that the 601 assigns to its child agent is known. , The target can be specified. In this embodiment, if it is known that "601 uses the name of the month as a label", "Jully" attached to 607 is obtained from its own absolute path, and the label "602""June" and inferring "(()" as information indicating 605 as a result.
parent parent parent June
Task1 Subtask) "is obtained, and this can be used to actually specify 605.

For communication beyond such a group of agents, the common ancestor 601 does not require any special processing for mediation. The same can be said for all the other agents. If each agent clarifies the relationship with the other agent in the form of a label, then 611 becomes the relative relationship with the communication target 605 "the parent's own It is possible to identify and communicate with the target object only by knowing the information "child agent of child, Task1 of child, Task1 of child of June, parent of the parent, and specifying it in the form of a relative path.

Embodiment 3 Although the foregoing embodiment deals with an agent group having a common ancestor, a similar communication target can be specified even with an agent group having no common ancestor.

An embodiment of the present invention will be described with reference to FIG. 703, 704, 705, 706, 707, 70
8, 709 and 710 are all agents having no parent-child relationship with each other. 701 and 702 are task processing groups constructed according to the contract net protocol. 703 is a manager of the group 701;
4,705,706 are contractors who have contracted with 703. Similarly, 707 is the manager of group 702,
708, 709, and 710 are contractors who have contracted with 707.

Although the manager and his contractor initially do not have any information about each other, the manager broadcasts his / her identifier, environment name and absolute path at the time of the task presentation, and the contractor also bids his / her own identifier, at the time of bidding. By notifying the manager of the environment name and the absolute path, information for identifying each other in communication is obtained, and a task processing group is configured. In response to the information obtained in this manner, the manager sends information on each contractor to “contractor”.
1 "," contractor2 "," contrac
The contractor manages the information related to the manager by attaching the label “manager”.
03 and 707 receive communication of the other party's task presentation and recognize each other as managers having similar tasks,
Labeling each other with "same _- trade" and managing the information.

In a group of agents having such a relationship, if the contractor 704 knows that the manager knows another manager who has a similar task,
Reference numeral 704 denotes a relative path representing a relationship of “an agent who has contracted with a similar manager of his / her manager”. (() Manager same _- trade contractor1) (() manager same _- trade contractor2) (() manager same _- trade contractor3) ( By specifying information such as () manager same _- trade contractor4) ……, it is possible to communicate with agents 708, 709, and 710 in a similar situation to the user. However, “(() manager same _- tradecontrac
This communication fails because the agent indicated by tor4) "does not exist. By performing such communication, 704 searches for an agent who has contracted a subtask similar to itself, and easily cooperates with the agent. Also, since this relationship is completely symmetric, the opposite agent 708 and the like can specify 704, 705, 706 can be communicated.

[0076]

As described above, according to the present invention, the following effects can be obtained.

Even if the address (identifier and environment name) changes due to the movement of the communication target,
The latest address can be reliably obtained. The reason is,
This is because the latest identifier and environment name can be searched from another agent by the absolute path.

The load does not concentrate on a specific device with respect to the address management of the mobile agent. The reason is,
Instead of using unique names on the network for name management of mobile agents, identifiers are managed for each environment and labels are managed for each agent, and each mobile agent itself has a partial name server function, and each mobile agent has a name server function. Cooperate with each other to realize one name server as a whole.

There is no waste in inquiring the address of the movement target every time. The reason is that, if the communication with the communication target succeeds by any method, the identification name, the environment name, and the absolute path are acquired from the communication target and stored in the own agent.

It is possible to specify a communication target based on the relative relationship between agents. The reason is that instead of a unique name on the network, the communication target can be specified by a relative path, which is a label string indicating the route to reach the communication target agent from the own agent, combining labels managed for each agent. Because. As a result, a group of agents with the same interaction
It can be created from the same agent definition. Further, at the same time, it is possible to specify an object beyond such an agent group.

In the configuration in which the search path is minimized, the search for the address to be communicated can be performed more efficiently.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration example of a communication function.

FIG. 3 is a flowchart illustrating an algorithm for defining a search path.

FIG. 4 is a hierarchy diagram of a parent-child relationship showing an example in which all mobile agents are generated with fixed agents as ancestors.

FIG. 5 is a flowchart illustrating an operation example of a control unit of a communication function when communicating with another mobile agent.

FIG. 6 is a flowchart illustrating an operation example of a search path expanding function.

FIG. 7 is an explanatory diagram showing an operation of a procedure for searching for an identifier and an environment name of a communication target based on a specified search path.

FIG. 8 is an explanatory diagram of a search path minimizing function.

FIG. 9 is an explanatory diagram of a search path minimizing function.

FIG. 10 is a flowchart showing an operation of a search path minimizing function.

FIG. 11 is a diagram illustrating an agent group used in the description of the first embodiment.

FIG. 12 is a diagram showing an identifier and an environment name of the agent group of FIG. 11;

FIG. 13 is a diagram illustrating an absolute path of each agent in FIG. 11;

14 is a diagram showing an example of information held by a mobile agent 305 in FIG. 11 in a database in a name server.

FIG. 15 is a diagram showing an example of information held by a mobile agent 310 in FIG. 11 in a database in a name server.

FIG. 16 shows a mobile agent 30 on an environment “Env1”.
FIG. 5 is a diagram showing a state in which the environment No. 5 moves onto an environment “Env4”.

FIG. 17 is a diagram showing an example of updating a part relating to the mobile agent 305 of the name server of the mobile agent 310 itself.

FIG. 18 is a diagram illustrating an agent group used in the description of the second embodiment.

FIG. 19 shows agents 502, 503, and 5 in FIG.
It is a figure showing the example of the relative path of the communication object given to 04,505.

FIG. 20 is a diagram showing a plurality of instances generated from the definition of the same class group.

FIG. 21 is a diagram illustrating an agent group used in the description of the third embodiment.

FIG. 22 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

101 to 104 Environment 105 to 108 Mobile agent device 109 Agent main function 110 Information including an agent's own identifier, environment name and absolute path 111 Agent generation function 112 Mobile function 113 Communication function 114 Name server function 115: Name server control function 117: Database 118-120: Information on other agents 121: Communication function control unit 122: Communication function direct communication function 123: Communication function search path expansion function 124: Communication function search path Shortest function 125 ... Receiver of communication function 201 ... Fixed agent device 202-206 ... Agent device 207-211 ... Label 301 ... Fixed agent device 302-307 ... Agent device 308 ... Fixed agent device 309- 312: Agent device 313 to 323 Label: 401: Mobile agent device 501 to 505: Agent device definition class 601 to 611: Agent device instance 612, 613: Agent instance group 701, 702: Task processing group 703 to 710: Agent device

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code FI H04L 12/26

Claims (8)

[Claims] 1. An address management method for a mobile agent capable of moving between arbitrary environments on a computer network composed of a plurality of environments.
As a condition for generating a mobile agent, a condition that an ancestor always has a fixed agent that does not have a mobile function is given. For a mobile agent generated under the condition, an identifier and an environment name of the fixed agent of the ancestor and An absolute path name composed of a label sequence from the fixed agent of the ancestor to only the parent-child relationship label and arriving at the own agent is given. (B) For each mobile agent, a direct parent, child and The information composed of the absolute path, identifier, and environment name of the agent is managed by assigning a label name. (C) For a mobile agent that has moved between environments by itself, new information obtained in the moved environment is obtained. The direct parent / child agent is notified of the identifier and the environment name, and the moved parent / child agent side is moved. You should be prepared to always manage the latest of the identifier and the environment name of the agent, (d)
If the mobile agent fails to communicate with the communication target using the identifier and the environment name, the child agent is a direct child agent of the inquiry destination agent in order from the fixed agent determined by the identifier and the environment name in the absolute path name of the communication target. A mobile agent characterized in that the label and the environment name are specified and the identifier and the environment name are queried, and finally the current latest identifier and the environment name of the communication target whose communication has failed are obtained. Address management method. 2. The communication method according to claim 1, wherein when the communication with the communication target is successful, the identification name, the environment name and the absolute path are obtained from the communication target and stored in the own agent. Mobile agent address management method. 3. In the processing of (d), the own agent and the communication target that has failed to communicate have a common ancestor, and the own agent has a path length of the absolute path of the communication target that has failed communication. If the path length of the alternative path that traces from the path to the common ancestor and thereafter from the middle of the absolute path to be communicated is shorter, the label name is specified in order along the path of the alternative path, and the identifier and 3. The address management method for a mobile agent according to claim 1, further comprising the step of inquiring about an environment name and finally obtaining a current latest identifier and an environment name of the communication target in which the communication has failed. 4. When a communication target is specified by a relative path, which is a label string indicating a route from the own agent to the communication target agent, a label name is sequentially specified along the path of the relative path. 4. The method according to claim 1, wherein the identifier and the environment name are inquired to obtain the identifier and the environment name of the communication target, and the communication with the communication target is performed by using the identifier and the environment name. Mobile agent address management method. 5. An agent main function which controls the original operation of the agent such as information search, electronic shopping, reservation, etc.
An agent generation function for generating another mobile agent, a movement function for moving to another environment, a communication function for interacting with another agent, and a name server function for holding information on other agents,
A mobile agent device generated from a fixed agent or another mobile agent so as to have a fixed agent without a moving function as an ancestor, wherein an identifier and an environment name of the fixed agent of the ancestor and a parent-child relationship from the fixed agent of the ancestor are provided. And the name server function includes a direct parent and child absolute path, an identifier, and an environment. A database that manages information consisting of names with label names, and inquires of identifiers and environment names in information corresponding to the label names in the database in response to inquiries from other agents that specify label names. A communication function based on the movement function. When the own agent moves between environments, a new identifier and environment name obtained in the environment after the movement are notified to the direct parent / child agent managed in the database. When the latest identifier and environment name are notified from the parent and child agents, the control unit reflects the latest identifier and environment name to the database through the control function, and communicates directly with the communication target using the identifier and environment name. If the communication function and the communication by this direct communication function fail, specify the label name for the direct child agent of the inquiry destination agent in order from the fixed agent determined by the identifier in the absolute path name of the communication target and the environment name To query the identifier and the environment name, and finally the current and latest identification of the communication target in which the communication has failed. A mobile agent device having a search path expanding function for obtaining child and environment names. 6. The name server function according to claim 5, wherein an identification name, an environment name, and an absolute path of a communication target that has successfully communicated are stored in the database if they are not stored in the database. Mobile agent device as described. 7. The communication function according to claim 1, wherein the own agent and the communication target that has failed to communicate have a common ancestor, and the local agent has a common ancestor that is smaller than the path length of the absolute path of the communication target that has failed communication. A search path minimizing function for checking whether or not the path length of an alternative path following the ancestor of the alternative path from the middle of the absolute path of the communication target is shorter. When a path is selected, the search path expanding function inquires an identifier and an environment name by sequentially specifying a label name along the route of the alternative path, and finally, the communication fails. 7. The mobile agent device according to claim 5, wherein the mobile agent device has a configuration for obtaining a current latest identifier and an environment name of a communication target. 8. The search path expanding function is provided, when a communication target is specified by a relative path which is a label string indicating a route from the own agent to the communication target agent, along a path of the relative path. A configuration in which an identifier and an environment name are queried in order by designating a label name in order, and finally an identifier and an environment name of a communication target are obtained, and communication with the communication target is performed using the direct communication function by using the obtained identifier and environment name. 8. The mobile agent device according to claim 5, wherein the mobile agent device has a mobile agent device.