JPH0728732A - Interprocess communication method - Google Patents

Abstract

(57) [Summary] (Modified) [Purpose] Performing tuple space type inter-process communication that can be accessed by a name with a tree structure in a safe and efficient manner. [Configuration] A request analysis / transfer / processing unit 500 that accesses a tuple space by a tree-structured name, and tuple space access is controlled from the process identifier assigned to the tuple, and an effective access method is set by setting the effective time. A tuple space management unit 700 provided and a network control unit 600 that creates and manages a duplication of a tuple space by communicating with a tuple space management unit of another node via a network. [Effect] In a tuple space type communication mechanism in which multiple tuple spaces exist, it is possible to specify tuples flexibly by accessing tuple spaces by tree structure names. In addition, tuple access control and multiple tuple spaces can be specified. Copying on nodes can improve security and reliability.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an interprocess communication system for transmitting information between a plurality of distributed processes using a shared data area called a tuple space and structured data called a tuple. is there.

[0002]

2. Description of the Related Art In a multiprocessor system having a configuration in which a plurality of computers are connected by a network, it is an important issue to realize high-speed data processing centered on parallel processing. Various proposals have been made regarding methods and communication methods. For example, in the Linda calculation model in the United States, a shared data area called a tuple space is virtually realized in a multiprocessor system, and information is transmitted between processes through input / output of structured data called a tuple for this shared data area. Since tuples are commonly accessed in parallel by a plurality of processes, efficient parallel processing can be realized.

The basic operation of communication between tuple spaces will be described with reference to FIG. First, communication between processes is performed by inputting a tuple into a tuple space “out”, extracting a tuple “in”, and reading a tuple “rea”.
The tuple extracted by the pattern matching between the tuple type and structure existing in the tuple space and the argument type and structure of "in" and "read" is performed by the "d". Fig. 32 (a) shows the tuple space. To (“aaa”, “bbb”, “ccc”), and (4
It shows that two tuples of "abc" exist first. In this state, process A and process B
Are out (“abc”) and out (“xy”, respectively.
Consider the case where a tuple is input by z ″, 20).
From process A, select [OUT, <string>, “ab
The message "c"] is transmitted to the tuple space. Here, OUT is a message identifier indicating an out request, <string> is information indicating the tuple type and structure, and
“Abc” is the actual data information. Similarly, from the process B, [OUT, <string, int>, “x
yz ", 20] is sent to the tuple space.
As shown in FIG. 32B, the tuple space is registered with these tuples, and new ("abc") and ("xy" are registered.
The tuple of z ″, 20) is registered. Next, process C
For tuple space in (char * buf,
Consider a case where a tuple of a sequence of a character string and an integer is requested by int x). From process C [I
N, <string, int>] message is sent. Then, the tuple space retrieves ("xyz", 20) corresponding to such a tuple and returns it to the process C as a response. At this time, the tuple is deleted from the tuple space at the same time, as shown in FIG. Next, referring to FIG.
Read (int y, ch for tuple space
Consider a case where a tuple of an array of an integer and a character string is requested to be read (different from the fetch operation) by ar * buf). From process D [READ, <int, s
message> is sent. Then, the tuple space retrieves (4, "abc") corresponding to such a tuple and returns it to the process D as a response. However, in this case, the tuple is not deleted from the tuple space. In addition, in FIG. 32C, the process E is in (int
Consider a case where a tuple of an integer and a sequence of integers is requested to be fetched by m, int n). The process E sends a message [IN, <int, int>]. However, the tuple space does not have such a tuple, and therefore the response processing to the process E is blocked (held). This block state is released when the tuple corresponding to this request is input, and a response is returned to the process E.

However, in the Linda calculation model, since there is only one tuple space, there is a risk of erroneous operation when tuples collide between different processes. Further, there is a problem that when the number of tuples increases, the tuple search time existing in the tuple space also increases accordingly. Further, when considering the modularity of the program, there is a problem that information cannot be structured in a calculation model in which all tuples are visible to all processes. On the other hand, some proposals have been made regarding the structuring of tuple spaces. For example, the Cellula model developed at Kyushu University and Japanese Patent Laid-Open No. 4-287121 tried to realize modularization of a program, collision avoidance of tuples, and improvement of tuple search time by generating multiple tuple spaces. It is a thing.

FIG. 33 shows an embodiment of Japanese Patent Laid-Open No. 4-287121. FIG. 33A is a block diagram of an embodiment, in which some processor elements are connected by a network, a shared memory is connected to the network, and a tuple space is realized on the shared memory. Has been done. FIG. 33B shows an example of the internal structure of the tuple space. Where #ORG, #A, #B, #
C is a tuple space. Associating with the basic concept of the Linda calculation model, it is sufficient to consider the case where only the tuple space of #ORG exists. The tuple space #C has a relationship of inheriting the tuple of #ORG according to the definition of FIG. That is, it means that a tuple existing in an area within #ORG but not included in another tuple space can be operated by accessing #C. By defining a tuple space with a limited scope in this way, it is possible to improve the efficiency of tuple search because there is no need to search tuple spaces that are not related, and collisions between tuples are reduced. is there.

[0006]

Since the conventional tuple space type communication is configured as described above, a plurality of tuple spaces are actually created on the network, and the tuple space type communication is effectively and highly reliable. When trying to realize communication,
There were the following problems.

Problem 1: In order to perform tuple space type communication based on a plurality of tuple spaces, it is necessary to define the relationship between the tuple spaces and to specify the tuple space using the relationship. For example, when hierarchically associating "a tuple space is a child of another tuple space" with a grouping specification method of "all children of a tuple space" or "parent of a tuple space" An indirect designation method is required, but such a flexible designation method was not considered. Further, even when tuple spaces dispersed on the network are designated, there is no flexible designation method for designating the existing position on the network.

Problem 2: In the conventional tuple space communication, security has never been taken into consideration when inputting or taking out tuples. As a result, the tuples that have been put into the tuple space are visible to all processes that are searching the tuple space, which is a security problem.

Problem 3: In the conventional tuple space communication, the tuple data transfer process of reading the tuple is necessary even when it is desired to know only whether or not the tuple has been input. Therefore, there is a problem that efficiency must be sacrificed particularly when writing an application program that operates by using a tuple input from another process as a trigger.

Problem 4: There may be tuples in the tuple space that have not been fetched by the process but have been accumulated. This imposes an unnecessary load on the tuple space management mechanism and wastes the storage capacity. In addition, when this is managed by the client, there is a problem that it takes time and effort to inquire that the tuple has been taken out and to take out the tuple when the tuple exists.

Problem 5: When a tuple whose data information size is larger than the type and structure information is input to the tuple space, the type and structure information has already arrived in the tuple space, Since the data information has not arrived, it is not included in the tuple search target, and there is a problem that the processing of the process is blocked for a long time when trying to retrieve it.

Problem 6: Even though the tuple space is divided in the past, since each divided tuple space is one in the system, a process communicating using a certain tuple space Has a problem that communication cannot be established if a failure occurs in the tuple space management mechanism.

Problem 7: Conventionally, in the tuple space, only one tuple was allowed to be thrown in one tuple throwing operation, so that the communication cost required to throw a plurality of tuples was increased, There is a problem that interprocess communication cannot be performed efficiently.

Problem 8: Conventionally, with respect to the tuple space, only one tuple extraction or read operation is allowed in one tuple extraction and read operation. Therefore, there is a problem that the communication cost required for extracting or reading out a plurality of tuples increases, and communication between processes cannot be performed efficiently.

Problem 9: Conventionally, when the retrieval result for the tuple space or the reading process cannot be obtained within a certain period of time, or the retrieval process is not completed within a certain period of time and no response is returned. It was not possible to handle exception handling etc.

The present invention has been made in order to solve the above-mentioned problems, and provides a tuple space type communication system for performing communication using a plurality of tuple spaces.
(1) By giving a method that defines the relationship between tuple spaces, it is possible to specify a flexible tuple space, (2) improve security that is a problem in tuple space communication, and (3) input tuples. It enables efficient operation even for applications that use triggers. In addition, (4) it is possible to save the trouble of managing a wasteful area of the tuple space due to tuples existing in the tuple space for a long period of time, and (5) search processing prior to the arrival of data in the tuple search in the tuple space. (6) The purpose is to improve the reliability of management of dispersed tuple spaces. Furthermore, (7) it is possible to input a tuple in which a plurality of tuples are connected, (8) it is possible to take out and read out a plurality of tuples, and at the same time, (9) it is possible to operate a tuple with a designated time. It is intended to realize a communication mechanism.

[0017]

In the inter-process communication mechanism according to the present invention, a means for analyzing a tuple space by using a name having a tree structure, which is associated with a hierarchical relationship of tuple spaces or a location on the network. And a means for accessing tuple spaces dispersed on the network based on the analysis result.

Further, the process for accessing the tuple space is provided with a mechanism for giving a specific grouping and an ID for identifying the group, and the specific ID
Is provided, and a tuple space access management means for controlling access to a tuple of the tuple space using the ID is provided.

Further, means for pre-registering the type and structure information of the target tuple with respect to the tuple space management mechanism, and notifying the process whether the tuple exists or has been input. A means is provided for the tuple space management mechanism.

Also, the existence period in the tuple space can be set in a specific field of the tuple in the tuple space communication, and means for interpreting the disappearance time set in the tuple from the data structure of the field and the disappearance time elapsed information are used. Means for deleting the tuple designated by the above.

When a large amount of data is contained in one tuple, or when it becomes necessary to place many tuples in the tuple space at the same time, the type information and structure information required for tuple retrieval are referred to as data information. Separately, a tuple division transmission means to be sent first and a tuple retrieval means for recognizing the divided tuples and performing retrieval processing in advance without waiting for arrival of data information are provided.

In addition, when one tuple space exists in a plurality of nodes, a tuple space management means for managing the management information about the copy, and a management information transfer for notifying the tuple space management information existing in a certain node to another node. A communication means is provided which enables communication with the tuple space management means on another node when a failure occurs in communication between the means and the tuple space management means on the own node.

Further, a tuple concatenation means for collectively converting a plurality of tuples into one message, and a message analysis means for judging whether the message sent to the tuple space is a normal single tuple or a concatenated tuple. A message dividing means for decomposing a message when a plurality of tuples are linked from the analysis result by the analyzing means and a tuple space managing means for actually registering the tuple in the management table are provided.

Further, the format of the request for the process to specify the number of tuples in response to the tuple extraction request, the tuple operation interpretation means for interpreting the format, and the tuple space for actually extracting a plurality of tuples from the tuple space. A management means, a tuple connecting means for collecting a plurality of tuples into one tuple, and a tuple dividing means for analyzing and dividing the tuples connected on the process side are provided.

In addition, a request format for setting a time limit for a request for fetching or reading a tuple from a tuple space, and request interpreting means for interpreting the format to obtain the time limit set for the tuple space operation. , A timer managing means for measuring the time limit of a request by a timer and sending the time limit to the request managing means when the time limit has passed, and when the notification from the timer managing means is received, the processing of the corresponding unprocessed request is terminated. A request management means for returning information indicating that the time limit has been exceeded to the requesting process is provided.

[0026]

In the inter-process communication mechanism according to the present invention, the tuple space is identified by using a tree-structured name that is associated with the hierarchical relationship of the tuple space and the existing position on the network. It is possible to flexibly specify the hierarchical relationship and the location.

Further, a process of accessing the tuple space is given a specific ID by using a mechanism for giving a specific grouping and an ID for identifying the group.
And the ID is used to control access to the tuple space tuple, so that the security function of the tuple space communication can be improved.

Also, a mechanism for registering the type and structure information of the target tuple to the tuple space management mechanism and notifying whether the tuple exists or has been input should be added to the tuple space management mechanism. Thus, it becomes easy to write an application program that operates by using the input of a tuple as a trigger.

Further, the existence period of the tuple space can be set in a specific field of the tuple space in the tuple space communication so that the tuple space management mechanism can detect the expiration of the tuple space. Therefore, the tuple left as dust is discarded. By doing so, it is possible to effectively use the storage area of the tuple space and reduce the labor of management.

When one tuple contains a large amount of data, or when many tuples need to be placed in the tuple space at the same time, the type information and structure information necessary for tuple retrieval are referred to as data information. Separately, the tuple space management unit performs the search processing without arriving at the arrival of data information, so tuple search can be started without depending on the data transfer time, and efficient communication can be performed. Is possible.

Further, when a copy of a tuple space is created on a plurality of computers, one tuple space is managed by a plurality of management mechanisms, and when a change occurs in a copy of one tuple space, the copy exists. Since I transferred it to all computers and kept the state the same,
The same result can be obtained regardless of which copy the process accesses, and even if a tuple space fails, stable communication function can be maintained by accessing another normal tuple space, and the failure occurs. The reliability with respect to the occurrence can be improved.

When it is necessary to put a plurality of tuples in the tuple space, a plurality of tuples are connected to one another,
Since a plurality of tuples can be transmitted to the tuple space by one-time insertion processing, efficient communication can be performed with a reduction in communication cost of tuple transmission.

Further, when it is necessary to extract or read out tuples having the same type and structure by connecting from the tuple space, the required number of tuples can be defined in the method of defining the request for extraction or read out. In addition, the tuple space management mechanism makes it possible to retrieve tuples according to the required number of tuples, or to concatenate them with a read operation to make one message, so multiple tuples can be retrieved with one request. It becomes possible to reduce the communication cost due to the tuple extraction and read processing, and the efficient communication becomes possible.

In addition, since a method for setting and requesting an expiration date for a tuple operation, a processing mechanism for interpreting the expiration date, and a mechanism for monitoring the expiration date are provided, when the expiration date expires. It will be returned to the issuer of the request, and it will be possible to easily return the tuple retrieval result and the retrieval result that have been searched within the expiration date, and to set exception processing for expiration of the expiration time.

[0035]

【Example】

Example 1. Example 1 of the present invention will be described with reference to FIGS. 1 is a block diagram showing interprocess communication according to a first embodiment of the present invention. In FIG. 1, a plurality of computers 100 are connected by a network 400, and a process 200 is operating on each computer. A tuple space communication mechanism 300 that provides tuple space type inter-process communication is operating on each computer.

FIG. 2 is a detailed block diagram of the process and tuple space communication mechanism 300 in FIG. Here, the message processing unit 260 is a part that controls communication between the process 200 and the tuple space communication mechanism 300. The process 200 uses the tuple space name 205 as a target,
A tuple 210 to be input into the tuple space 730, an operation identifier 220 indicating an operation on the tuple space 730,
Tuple operation additional information 230 as optional information
Holding The message processing unit 260 on the process 200 side assembles these pieces of information and sends them as a message 240 to the tuple space communication mechanism 300, and also cuts out the processing result 250 and the received tuple 210 from the received message 240.

When the message processing unit 260 on the tuple space communication mechanism 300 side receives the message 240 from the process 200, the received information is subjected to request analysis / analysis.
It is sent to the transfer / processing unit 500. Request analysis / transfer / processing unit 5
In 00, the node in which the tuple space 730 exists is identified based on the information of the tuple space name 205, and if it exists in its own node, the tuple space management unit 70
The tuple space management unit returns the processing result to the process 200 by using the message processing unit 260. The request analysis / transfer / processing unit 500 analyzes the tuple space name 205 and returns tuple space 730.
If it is determined that the node exists on another node, the request is transferred to the network controller 600 on the other node using the network controller 600 of the own node. Network control unit 600 on the other node that received the request transfer
Executes the transferred request by the tuple space management unit 700 of its own node and returns the result to the requesting network control unit 600. The request source network control unit 600 analyzes the processing result by request analysis / transfer / processing unit 500.
Then, the request analysis / transfer / processing unit 500 sends the processing result to the process 200 through the message processing unit 260.

FIG. 3 shows the message 24 in the first embodiment.
0 is a configuration example of 0. The message 240 is
Header information 241, such as the serial number of the message, process information 242 indicating from which process the tuple space operation type 243, option information 244 attached to the operation, tuple space name 2 to be operated
45, the type and structure information 246 of the tuple to be operated, and the actual tuple data or processing result information 247 area.

FIG. 4 shows a syntax example of the tuple space name definition in the first embodiment. A description based on this syntax enters the tuple space name area 245 of FIG. In this syntax, tuple spaces have a tree-structured name syntax based on the hierarchical structure in which the tuple spaces are managed. That is, the names of tuple spaces located at the upper level of the hierarchy are described in order, separated by delimiters.
Further, the tuple space has the concept of a current tuple space (“.”) And its parent tuple space (“..”), and can show a relative relationship with respect to the tuple space of interest. Further, as a regular expression, "/ ABC / XYZ / *" can be specified for a specification such as "all tuple spaces below that tuple space".

FIG. 5 shows a processing flow of the request analysis / transfer processing unit 500 in the first embodiment. First, when a tuple space processing request is received from the message processing unit 260 (S501), the tuple space name is analyzed (S502). Here, the node whose tuple is managed is extracted from the tuple space name, and it is determined whether the tuple space is a local tuple space or a remote tuple space (S503). Whether it is local or remote can be determined by, for example, making an agreement such that the first path of the tuple space name represents the node name, or preparing a tuple space position information DB separately.

When it is determined that the name is the local tuple space name, the request is sent to the local tuple space management unit (S504), and the processing result is returned to the message processing unit (S505). Also, if it is determined that the tuple space name is a remote tuple space,
The network control unit 600 sends the tuple space management unit to the remote (S506), and the processing result is sent to the message processing unit (S505).

FIG. 6 is a processing flow of the network control unit 600 in the first embodiment. The network control unit determines whether the request is from the local request analysis / transfer / processing unit 500 or the request from the remote network control unit (S601 to S603). Local tuple space management unit 70
0 for processing (S604), and the processing result is returned to the remote network control unit (S605). If the processing request is a request from the local request analysis / transfer / processing unit, it is processed by the remote network control unit (S6).
06), request processing to the remote tuple space management unit, and analyze the processing result locally and transfer it.
It is returned to the processing unit (S605).

FIG. 7 is a processing flow of the tuple space management unit 700 in the first embodiment. When the tuple space management unit receives the request (S751), the tuple space name is analyzed (S752), and if it is a regular expression, the tuple space name is expanded (S753), and the tuple space name is checked ( (S754), the tuple space is actually accessed (S755). Also,
If it is an access to the tuple space management itself, after performing it (S756), the processing result is finally returned to the request source (S757).

Example 2. A second embodiment of the present invention will be described based on FIGS. 8, 9 and 10. FIG. 8 is a configuration diagram of the process structure, the process ID management mechanism, and the tuple space communication mechanism 300 according to the second embodiment. Here, the message processing unit 260 uses the process 200 and the tuple space communication mechanism 3 as in the first embodiment.
Control communication between 00s. The process 200 is a process I managed by the process ID management mechanism 800.
D is received when the process is created, and it is set in the message as process information. The process ID management mechanism manages the managed process ID in the process ID table 810. FIG. 9 shows an example of the process ID table.

When the message processing unit 260 on the tuple space communication mechanism 300 side receives the message 240 from the process 200, the received information is subjected to request analysis / analysis.
It is sent to the transfer / processing unit 500. Request analysis / transfer / processing unit 5
In 00, the node in which the tuple space exists is specified based on the information of the tuple space name, and if it exists in its own node, the tuple space management unit 700 is requested to perform processing, and the result is processed as a message. Return to process 200 using section 260. If tuple space exists on another node as a result of parsing tuple space name,
The request is transferred to the network control unit 600 on another node by using the network control unit 600 of the own node. Upon receiving the request transfer, the network control unit 600
The request transferred by the tuple space management unit 700 of its own node is executed and the result is returned to the requesting network control unit 600. Requestor network controller 6
00 returns the processing result to the request analysis / transfer / processing section 500, and the request analysis / transfer / processing section 500 sends the result to the process 200 through the message processing section 260.

The tuple space management unit 700 holds the access right of the tuple space as an access management list 740. The tuple space access management list 740 is changed when there is a tuple space creating process or a tuple space access right changing process.
Then, when an access request to the tuple space is received, the validity of the access request is checked based on this access management list 740, and a processing request for which access is not permitted is rejected. FIG. 10 shows an example of the access management list.

For example, as shown in FIG. 9, the process ID
The process having 56789 as the tuple space ID 5234 managed by the access management list of FIG.
Consider accessing a tuple space with 6.
The process sends an access request to tuple space communication mechanism 3
When the request is transmitted to 00, the request finally reaches the tuple space management unit 700 via the message processing unit 260 and the request analysis processing unit 500. The tuple space management unit extracts the process user ID 61 and the group ID 356 from the process ID table of FIG. 9 as the process information added as the access request. Therefore, the tuple space I of the access management list in FIG.
When the access right of the tuple space defined by D is 52346 is checked, it is described as rwrw-. From this, it can be seen that the access is permitted to the process whose user ID or group ID matches the tuple space, and the access is not permitted to the other processes. By the way, the user ID and group ID of the tuple space are 11 and 129, respectively.
And the process user ID and group ID are each 6
It can be seen that it is rejected because it is different from 1 and 356.

Example 3. The third embodiment of the present invention will be described with reference to FIGS. Figure 11
Is the tuple space management unit 70 in the third embodiment of the invention.
0, and other elements of the tuple space communication mechanism are
For example, assume the configuration of the first embodiment. Here, the tuple space management unit 700 analyzes the tuple space requirement /
A processing unit 701, a tuple space management table 710,
A tuple input / output unit 702, a tuple management table 703, a tuple space area 720, a notification tuple type / structure information management unit 708, and a notification tuple type / structure information list 707. Next, operations of the tuple space request analysis / processing unit, the tuple input / output unit, and the notification tuple type / structure information management unit in the third embodiment will be described based on the processing flow of FIG. The request analysis / processing unit 701 refers to the tuple space management table 710 from the tuple space name included in the message and searches for the tuple space ID (S771). Here, FIG. 13 is an example of the tuple space management table, in which the name of the tuple space and its ID are associated with each other. Next, identify whether the message is a "tuple space I / O request (tuple space communication request)" with normal data, or a "notification processing request" that notifies only information whether there is a tuple without data. Yes (S772 to S7
73). Then, if it is a normal “tuple space communication request”, the tuple space ID and its processing content are sent to the tuple space input / output unit 702. The tuple input / output unit refers to the tuple management table 703 created corresponding to the tuple space ID, and performs tuple input / output processing (S).
778). FIG. 14 shows an example of the tuple management table. If the request is input of a tuple into a tuple space, the information is also sent to the notification type / structure information management unit 708 (S780). If there is a tuple having the same type and structure as the tuple entered in the notification type / structure information list managed by the notification type / structure information management unit (S781),
The process waiting for the tuple insertion notification is taken out from the list and a notification message is sent (S782).

On the other hand, if the message is a "notification processing request" that simply notifies the presence / absence of a tuple, the tuple space ID is retrieved (S774) and then the tuple space ID management unit 708 for notification is informed of the tuple space ID. And request a processing request. This notification tuple type / structure information management unit 708
Then, it is checked whether or not the requested tuple exists in the requested tuple space (S775), and if it exists, it is immediately notified that the requested process exists (S777). If there is no tuple, the notification tuple type / structure information management unit stores the message in the notification tuple type structure information list 707 (S776). Then, when a new tuple is input, the list is checked, if the type / structure information matches, the result is notified, and the stored message is deleted from the notification tuple type structure information list. . FIG. 15 shows an example of the notification tuple type / structure information list.

Example 4. A fourth embodiment of the present invention will be described based on FIGS. 16 and 17. FIG. 16 shows a tuple space management unit 700 according to the fourth embodiment of the invention, and the other elements of the tuple space communication mechanism are:
For example, assume the configuration of the first embodiment. In this embodiment, the tuple space management unit 700 includes a tuple space request analysis / processing unit 701 and a tuple space management table 71.
0, tuple input / output unit 702, tuple management table 70
3, a tuple space area 720, a tuple extinction time management unit 711, and an extinction time management table 712. FIG. 17 is an example of the disappearance time table 712 managed by the tuple disappearance time management unit 711 according to the fourth embodiment of the invention. The tuple space request analysis / processing unit 701 searches a tuple space from the tuple space management table, requests the tuple input / output unit for actual processing, and performs tuple input / output processing. The tuple input / output unit 702 registers the tuple ID and its disappearance time in the tuple disappearance time management unit 711 when the tuple disappearance time is set when the tuple is input. The tuple annihilation time management unit 711 manages the tuples registered from the tuple input / output unit and their annihilation times using an internal timer mechanism, and when a tuple ID for which the annihilation time has elapsed is detected, the tuple ID is tuple ID. Notify the input / output unit and delete it.

Example 5. A fifth embodiment of the present invention will be described with reference to FIGS. FIG.
Is a tuple space management unit 70 according to the fifth embodiment of the invention.
0, and other elements of the tuple space communication mechanism have the configuration of the first embodiment, for example. Here, the tuple space management unit 700 includes a tuple space request analysis / processing unit 701, a tuple space management table 710, a tuple input / output unit 702, a tuple management table 703, a tuple space area 720, a tuple type / unstructured data list. 7
It is composed of 13. The tuple space request analysis / processing unit 701 searches a tuple space from the tuple space management table, requests the tuple input / output unit for actual processing, and performs tuple input / output processing.

FIG. 19 shows an example of the message format in the fifth embodiment of the invention. FIG. 19A shows the type / structure information of the divided message, and FIG. 19B shows the data information of the divided message. Is shown. Here, in the operation option, the message is structure information (tuple type,
It is divided into structure (size, etc.) and data information (actual data such as numerical value and character string), and a flag indicating which of them is set.

In FIG. 18, the operation option added to the message is analyzed by the tuple input / output unit 702, and the following processing is performed when the tuple is input. When the tuple structure information and the tuple data information are separately divided and input when the tuple is input, (1)
If the tuple input processing request that has arrived is “divided data information”, it has already arrived and is managed in the tuple management table based on the process information (process ID) assigned to the tuple and its message information (message ID). It is checked whether or not it matches a tuple having only the structure information (in FIG. 20, the data area is described as “not yet arrived”). If there is a matching tuple, the data information is described in the structure information. FIG. 20 shows an example of the tuple management table 703 according to the fifth embodiment of the present invention. Here, the tuple ID 12347 indicates that the tuple ID has not arrived yet. On the other hand, if there is no matching tuple process ID and message ID, it is stored in the tuple type / structure information unarrival data list 713. Next, (2) if the arriving tuple input processing request is “divided structure information”, the process ID and message are displayed as to whether “data information” has already arrived in the tuple type / structure information unarrival data list. The ID is confirmed, and if it arrives, the information is added and registered in the tuple management table 703.
If such data information does not exist, it is entered and registered in the tuple management table (FIG. 20) in the data area. FIG. 21 is an example of a tuple type / unstructured data list in the fifth embodiment of the invention.

Example 6. A sixth embodiment of the present invention will be described based on FIGS. 22 and 23. Figure 2
2 is a configuration diagram of a process and a tuple space communication mechanism 300 according to the sixth embodiment of the present invention. Here, the message processing unit 260 is a unit that controls communication between the process 200 and the tuple space communication mechanism 300, as in the first embodiment. In the request analysis / transfer / processing unit 500, (1) when the tuple space creation request includes a copy designation, information on which node (computer) to place the copy tuple space in the copy management table 520 is provided. Record. Then, a tuple space is created in the recorded node. FIG. 23 shows an example of the copy management table. Here, the tuple space called / ABC is
It is shown that it exists above node1 and node2. Further, (2) when there is an access request to the tuple space, the copy management table is searched from the tuple space name to check whether or not the tuple space has a copy. When a tuple space having a copy is accessed, the same processing request is simultaneously transferred to another node via the network control unit 600 if the copy exists in another node. It For example, when the copy management table in node 1 is as shown in FIG. 23, the access to / ABC is also executed in node 2. As a result, access to the copied tuple space is always controlled so that the same conditions are met. And
(3) When the request analysis / transfer / processing unit 500 issues a processing request to the local tuple space management unit 700, if a processing result cannot be obtained due to some trouble occurrence, then on another node where a copy exists. The tuple space management unit 700 is requested to perform processing via the network control unit 600, and the processing result is received.

Example 7. A seventh embodiment of the present invention will be described based on FIGS. 24 to 27. Figure 24
FIG. 8 is a configuration diagram of a process and tuple space communication mechanism 300 according to a seventh embodiment of the present invention. Here, the message processing unit 260 is a unit that controls communication between the process 200 and the tuple space communication mechanism 300, as in the first embodiment. In the flowchart of the request analysis / transfer processing unit of FIG. 25, the request analysis / transfer / processing unit 500 analyzes the processing request for the tuple space (S801), and if it is a multiplexing processing request that requests input of a plurality of tuples at the same time. , Register it in the multiplexing request table 510 (S8
03). Otherwise, the normal tuple space access processing (S802) is performed as in the first embodiment. In the multiplexing request processing unit 530, the multiplexing request management table 51
0 is monitored, and if there is a multiplexing request that has not been processed, one of the requests is executed. When all the processing related to one multiplexing request has been executed, the result is returned to the message processing unit 260 as a response. In addition, each multiplexed processing is processed in the same manner as in the first embodiment.

FIG. 26 is a structural example of the multiplexing request table 510 in the seventh embodiment of the invention. The item resulting from the processing request that has not been processed is described as unprocessed.
For example, a multiplexing request with a message ID of 45 indicates that three out processes are multiplexed, and two of them are completed.

FIG. 27 shows a message structure for making a multiplexing request in the seventh embodiment of the invention.
This message contains n input tuples.

Example 8. The ninth embodiment of the present invention will be described with reference to FIG. FIG. 28 is a configuration example of a tuple extraction message and its response message according to the eighth embodiment of the present invention. Here, the tuple space communication mechanism has the same configuration as that of the seventh embodiment.

The request side message shown in FIG. 28 (a) includes the tuple type / structure information and the number of tuples requested to be extracted. These are the multiplexing request processing unit 53 of FIG.
When 0, the request is expanded as a multiplexing request, and each fetch request is executed. Then, when all the extractions are completed, the result is returned by the response message of FIG. 28 (b). The response message has a structure in which n results are returned. The assembly of the response message is performed by the multiplexing request processing unit in FIG.

Example 9. A ninth embodiment of the present invention will be described based on FIGS. 29 to 31. FIG. 29
Is a tuple space management unit 70 according to the ninth embodiment of the invention.
0, and other elements of the tuple space communication mechanism have the configuration of the first embodiment, for example. Here, the tuple space management unit 700 has a tuple space request analysis / processing unit 701, a tuple space management table 710, a tuple input / output unit 702, a tuple management table 703, a tuple space area 720, a tuple processing request waiting list 704,
The processing time limit management unit 706 and the time limit table 705 are included. Tuple space requirement analysis / processing unit 7
At 01, a tuple space is searched from the tuple space management table, and the tuple input / output unit is requested to perform actual processing to perform tuple input / output processing. However, if there is no corresponding tuple in the tuple extraction or read processing, the request is stored in the tuple request processing waiting list 704. FIG. 30 shows an example of the tuple processing request waiting list 704. Then, when the time limit is defined in the request, the tuple request process ID and the time limit are registered in the process time limit management unit 706. When a new tuple is input, the tuple input / output unit 702 executes the executable process existing in the tuple process request waiting list 704.

The processing time limit management unit 706 manages the registered processing request ID and time limit in the time limit table 705. An example of the time limit table 705 is shown in FIG. When the processing time limit management unit 706 finds a processing request that exceeds the time limit, the processing time limit management unit 706 notifies the tuple input / output unit of the processing request. Then, the tuple input / output unit 702 removes the request from the tuple processing request waiting list 704, and returns a value notifying the timeout to the tuple space request analysis / processing unit 701 as a response to the request.

[0062]

Since the present invention is constructed as described above, it has the following effects.

Since the tuple space is designated by a name having a tree structure, a flexible designation method in which a regular expression is mixed with the designation of the tuple space can be realized. Then, when the tuple space is operated, the process ID is used to check whether or not the process is permitted to operate the tuple space, so that the security can be improved and the reliability as a system can be improved. is there. Also, since it is possible to have a copy of the tuple space in another node, even if one tuple space fails, the other node can back it up and maintain service continuity and system consistency. The effect is that it can be done. Furthermore, by providing an expiration date for the tuples to be input and tuple space operations, it is possible to deal with wasted storage space and real-time processing.
In addition, multiple tuples can be batch-populated at the same time, and the tuple type and data structure can be sent separately from the actual data, enabling efficient communication between nodes. There is an effect that can be done.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an interprocess communication system according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram showing a tuple space communication mechanism according to the first embodiment of the present invention.

FIG. 3 shows a message structure according to the first embodiment of the present invention.

FIG. 4 shows a syntax of a tuple space name definition according to the first embodiment of this invention.

FIG. 5 shows a processing flow of a request analysis / transfer / processing unit according to the first embodiment of the present invention.

FIG. 6 shows a processing flow of a network control unit according to the first embodiment of the present invention.

FIG. 7 shows a processing flow of a tuple space management unit according to the first embodiment of the present invention.

FIG. 8 illustrates a process ID management mechanism and a tuple space communication mechanism according to the second embodiment.

FIG. 9 shows a definition example of a process ID table according to the second embodiment.

FIG. 10 shows a definition example of an access management list according to the second embodiment.

FIG. 11 illustrates a configuration example of a tuple space management unit according to the third embodiment.

FIG. 12 shows an operation flow of a tuple space request analysis / processing unit, a tuple input / output unit, and a notification tuple type / structure information management unit according to the third embodiment.

FIG. 13 illustrates a configuration example of a tuple space management table according to the third embodiment.

FIG. 14 illustrates a configuration example of a tuple management table according to the third embodiment.

FIG. 15 shows a configuration example of a notification tuple type / structure information list according to the third embodiment.

FIG. 16 illustrates a configuration example of a tuple space management unit according to the fourth embodiment.

FIG. 17 shows a definition example of an extinction time management table according to the fourth embodiment.

FIG. 18 illustrates a configuration example of a tuple space management unit according to the fifth embodiment.

FIG. 19 shows a message configuration example of the fifth embodiment.

FIG. 20 shows a definition example of a tuple space management table according to the fifth embodiment.

FIG. 21 is a tuple type / structure unarrival data list according to the fifth embodiment.

FIG. 22 shows a configuration example of a tuple space communication mechanism according to the sixth embodiment.

FIG. 23 shows an example of a copy management table according to the sixth embodiment.

FIG. 24 shows a configuration example of a tuple space communication mechanism according to the seventh embodiment.

FIG. 25 illustrates a processing flow of a request analysis transfer processing unit according to the seventh embodiment.

FIG. 26 shows a definition example of a multiplexing request management table according to the seventh embodiment.

FIG. 27 shows an example of the request message structure of the seventh embodiment.

FIG. 28 shows an example of the structure of a request message and a response message of the eighth embodiment.

FIG. 29 is a configuration example of a tuple space management unit according to the ninth embodiment.

FIG. 30 is a definition example of a tuple processing request waiting list according to the ninth embodiment.

FIG. 31 is a definition example of a time limit table according to the ninth embodiment.

FIG. 32 is an operation diagram showing a basic concept of tuple space type inter-process communication.

FIG. 33 is a block diagram showing a conventional tuple space type inter-process communication system.

[Explanation of symbols]

 100 computer 200 process 205 tuple space name 210 tuple 220 tuple operation identifier 230 tuple operation additional information 240 message 241 message header 242 process information 243 operation 244 operation option 245 tuple space name 246 type and structure information 247 tuple data and processing result 250 processing result 260 message processing unit 270 process ID 400 network 500 request analysis / transfer / processing unit 510 multiplexing request management table 520 copy management table 530 multiplexing request processing unit 600 network control unit 700 tuple space management unit 701 tuple space request analysis / processing unit 702 tuple input / output unit 703 tuple management table 704 tuple processing request waiting list 705 time limit table 706 Time limit management unit 707 Notification tuple type / structure information list 708 Notification tuple type / structure information management unit 710 Tuple space management table 711 Tuple disappearance time management unit 712 Disappearance time table 713 Tuple type / structure non-arrival data list 720 tuples Space area 730 Tuple space 740 Access management list 800 Process ID management mechanism 810 Process ID table

Claims (9)

[Claims] 1. A tuple space communication type inter-process communication method for accessing a plurality of tuple spaces from processes on nodes distributed over a plurality of networks, wherein the nodes have a meaning equivalent to a hierarchical relationship of tuple spaces. A name of a tree structure with a name, a tuple space name analysis means that has the name of the rule inside and interprets the name and location of the tuple space, and the location information of the tuple space obtained as a result of the analysis. An inter-process communication method comprising: a distributed tuple space access means for accessing a tuple space distributed on a network based on the above. 2. A tuple space communication type inter-process communication method for accessing a plurality of tuple spaces from a process on a node distributed on a plurality of networks, wherein the node assigns an identifier to a specific process group. An identifier assigning means to be given, a tuple space access right generating means for adding a creator's identifier and access mode information to the tuple space when generating the tuple space, and an identifier and a tuple of a process to be accessed in the process of accessing the tuple space. An inter-process communication system characterized by increasing the security of tuple space communication by including a tuple space access management means for restricting the tuple space access based on space access right information. 3. A tuple space communication type inter-process communication method in which a plurality of tuple spaces are accessed by processes on nodes distributed over a plurality of networks, wherein the node tuples tuple spaces from the processes. Means for pre-designating the type and structure information (hereinafter simply referred to as structure information) of the tuple space, and when the tuple space requested in the tuple space becomes available for reading or read, the tuple space notifies the process to that effect. An interprocess communication method characterized by comprising means for performing. 4. A tuple space communication type inter-process communication method for accessing a plurality of tuple spaces from a process on a node distributed on a plurality of networks, wherein the node sets a tuple disappearance time. A tuple annihilation time interpretation means that interprets the data structure and the annihilation time set in the tuple from the data structure, and the tuple management means that notifies the tuple lapse time when the tuple annihilation time is measured by a timer and the annihilation time elapses. An inter-process communication system comprising: a tuple disappearance time management means for sending; and a tuple management means for deleting a tuple designated by the disappearance time elapsed notification information from the tuple disappearance time management means. 5. A tuple space communication type inter-process communication method for accessing a plurality of tuple spaces from processes on nodes distributed over a plurality of networks, wherein the nodes are tuple structure information and actual data information. A tuple format that has been divided is also targeted, and a tuple division transmission means that interprets the tuple format and transmits structure information and data information separately, and a tuple search that recognizes the divided tuple and retrieves tuples by the structure information. And a tuple search means that is executed prior to the completion of reception of the interprocess communication method. 6. A tuple space communication type inter-process communication method for accessing a plurality of tuple spaces from processes on nodes distributed over a plurality of networks, wherein the node is a tuple existing in a plurality of nodes. The tuple space management means that manages the management information about the copy of the space and the change information of the tuple space management information that is managed by the tuple space management means that exists in one node are consistent in the tuple space management means that exists in another node. And a management information transfer means for notifying while maintaining the property, and when communication is abnormal due to a failure of the tuple space management means of the own node when the process communicates with the tuple space management means on the own node. Communication control means for communicating with tuple space management means on each node Interprocess communication method, characterized in that. 7. A tuple space communication type inter-process communication method for accessing a plurality of tuple spaces from processes on nodes distributed on a plurality of networks, wherein the node collects a plurality of tuples from each other. Tuple concatenation means for converting into one message, and message analysis means for judging from the message sent to the tuple space whether it is a normal single tuple or concatenated tuple,
A message dividing unit for decomposing a message when a plurality of tuples are linked from the analysis result of the message analyzing unit; and a tuple space managing unit for actually registering the tuple in the management table. Interprocess communication method. 8. A tuple space communication type inter-process communication method for accessing a plurality of tuple spaces from a process on a node distributed on a plurality of networks, wherein the node is a process for responding to a tuple retrieval request. A request format for specifying the number of tuples, a tuple operation interpretation means that interprets the format, a tuple space management means that actually extracts multiple tuples from the tuple space, and a tuple that combines multiple tuples into one tuple. An inter-process communication system comprising a connecting means and a tuple dividing means for analyzing and dividing the tuples connected on the process side. 9. A tuple space communication type inter-process communication method in which a plurality of tuple spaces are accessed from processes on nodes distributed over a plurality of networks, wherein the node extracts and reads tuples from the tuple space. The request format for setting the time limit for the request, the request interpreting means that interprets the format to obtain the time limit set for the tuple space operation, and the time limit elapses by measuring the time limit of the request with a timer. In that case, the timer managing means that sends the request to the request managing means, and when the notification from the timer managing means is received, the processing of the corresponding unprocessed request is terminated and the process requesting the processing of exceeding the time limit is returned. An interprocess communication system characterized by comprising a request management means.