JP2693460B2 - Distributed network system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Object of the Invention (Industrial field of application) The present invention relates to a distributed network suitable for sharing a series of tasks by interlocking between nodes connected to an online network. Regarding the system. (Prior Art) Conventionally, when performing distributed tasks by linking distributed computer nodes (distributed nodes) connected by online networks, an application program is placed in each node and The communication protocol was established between the programs, and the processing procedure for performing a series of online communication according to this protocol was incorporated into each program and processed. For example, the a program of the A node is used to access a resource (file, etc.) in the A node, and the b program of the B node receives the access result by the a program to access the resource (printer, etc.) in the B node. The functions are divided such that they are used for output processing. At this time, a series of processes for passing information from the a program to the b program are realized in the a program, and a series of processes for receiving information from the a program are carried out in the b program. It will be realized. In the above-described function sharing (distributed) method, for example, in the above example, a part of the resources of the A node (a part of the files to be accessed) is transferred to the B node for the purpose of load adjustment or the like. When it became necessary to relocate / investigate the resources to be accessed within the network, it was necessary to review and change the processing functions of the program itself. Also, node pairs with different resource allocations, such as files
A node A having x, v and a node B having a printer (first node pair), a node C having a file x and a file y
In addition, it is necessary to change a program to be operated for each pair of the node D having the printer (second node pair), and it is necessary to devise a great deal of commonality. (Problems to be Solved by the Invention) As described above, conventionally, in order to divide and execute a series of tasks linked with a plurality of nodes, an application program for each node is considered in consideration of resources possessed by each node. Since it has to be created, if the resource distribution relationship between each node changes, or if the resource distribution relationship is different for each network even if the business content is the same, a significant program change is required. was there. Therefore, according to the present invention, the nodes connected to the online network attempt to share a series of tasks, but if the tasks are the same, the resource distribution relationship between the nodes in the network changes. Even if the networks have different resource distribution relationships, it is an object to be solved to provide a distributed network system in which application programs arranged in each node can be shared. [Structure of the Invention] (Means for Solving Problems) The present invention relates to some distributed nodes that share a target task, and some common to each node that executes a target task for several small tasks. Application processing means, a control table in which execution designation information for specifying an execution node for each application processing means is set, and the application processing means to be activated next is operated by the own node by referring to this control table. The execution control means for deciding whether or not to perform it and the memory synchronization updating means for matching the contents of the designated memory area of the communication memory means used for information communication between the respective application processing means are respectively provided. Characterize. (Operation) According to the above configuration, even if all the application processing means necessary for performing the target job are prepared in common for each node, the execution instruction content of the control table for the division (allocation) of the job As long as it is defined according to the resource distribution relationship between the nodes, only the application processing means defined in the control table is selectively activated according to the task execution status in each node, and interlocked between the nodes. Then, the target work can be shared. Therefore, if the business contents are the same, all the application processing means necessary for the execution of the business are not affected by the resource distribution relationship among the nodes, and therefore, the functions of these application processing means and the execution control means, etc. It becomes easy to create an application program that realizes (Embodiment) FIG. 1 shows a block configuration of a distributed network system to which the present invention is applied. In the figure, A and B are arranged in a distributed manner, and nodes (distributed node, computer node) that perform a series of tasks in cooperation with each other, L
Is a transmission line interconnecting the nodes A and B. The internal configuration of nodes A and B is as follows. At nodes A and B, Ip1, I
p2 ... Ipn is prepared for each small task that constitutes the target task, and the CT is a control table that defines the execution instructions for the corresponding task, Ip1 to Ipn.
Cp is a control component that has a control table interpreter Cpi that interprets the specified contents of the control table CT and an application component execution controller Cpc that controls activation of the application processing components Ip1 to Ipn according to the interpretation result of the control table interpreter Cpi. is there. The control component Cp, the control table CT, and the application processing components Ip1 to Ipn are the nodes A and B.
It is a functional element realized by an application program installed in. PCM is a communication memory part that is defined and secured as a global area that is used for information communication between control component Cp and application processing components Ip1 to Ipn, and Sm is synchronization establishment for establishing synchronization between its own node and the partner node. The part u is a memory synchronous update mechanism. The memory synchronization update mechanism u detects a change in the contents of the same area in order to match the memory information of the designated memory area (area indicated by diagonal lines) on the communication memory unit PCM of the own node and other nodes, and Node (communication memory block PCM
Specified memory area). L1
Is a logical path (logical path) between the synchronization establishment units Sm of the nodes A and B, and L2 is a logical path (logical path) between the memory synchronization update mechanisms u of the nodes A and B. FIG. 2 shows an example of the control table CT. As shown in the figure, the content of each entry in the control table CT includes an identifier (application processing component name) of an application processing component and execution designation information necessary for performing a series of tasks. Here, three types of P (Primary), S (Secondary), and B (Both) are prepared as execution designation information. P is that the primary node executes the corresponding application process (small task), S is that the secondary node executes the corresponding application process, and B is that both the primary and secondary nodes execute the corresponding application process in parallel. It shows that each is executed. Note that the control table CT of this embodiment also includes the retribution necessary for managing the activation sequence of the application processing components Ip1 to Ipn necessary for performing a series of tasks,
The boot order is managed by another means, and the control table CT
It suffices to set only a pair of the application processing component name and the execution designation information described above. Next, the operation of the embodiment of the present invention will be described. First,
In the nodes A and B, the control component Cp sequentially refers to the control table CT and interprets it, and activates the application processing components Ipi (1 ≦ i ≦ n) with execution designation one after another. The activated application processing component Ipi communicates information via the communication memory unit PCM in its own node. Now, here, the application programs for realizing the respective functional elements of the control component Cp, the control table CT and the application processing components Ip1 to Ipn are equally arranged in the nodes A and B connected online by the transmission line L. Therefore, if the contents of the designated memory areas on the communication memory units PCM of the nodes A and B are adjusted to be the same by the memory synchronous updating mechanism u of the nodes A and B, for example, the application of the node A The processing component Ip1 can communicate with the application processing component Ip2 of the node A through the same communication memory unit PCM, just as it can communicate with the application processing component Ip2 of the node A through the communication memory unit PCM in the own node. Information can be exchanged between them. Therefore, the applied processing component Ip2
In order to execute corresponding application processing (small tasks) by
Which of the node A and the node B should be operated may be arbitrarily determined in consideration of the resource situation and the like, and the program part for realizing the application processing component Ip2 is the node A, B.
It can be created without considering which one of them is used. This is common to all the application processing parts Ip1, Ip2 ... Ipn placed in the nodes A and B, respectively. Further, the same applies to the control component Cp and the control table CT, and the same control component Cp and control table CT can be arranged in the nodes A and B, respectively, and no change is required. Now, as shown in the control table CT of FIG. 2, each of the application processing parts Ip1 to Ipn defines whether to operate in the primary node or the secondary node. Therefore, at the time of starting the processing, it is necessary to determine which of the nodes A and B will be the primary node and which will be the secondary node. This determination method is defined in advance as a part of the system generation information of the nodes A and B, or a sequence for obtaining a consensus between the nodes A and B determined by an instruction from the operator at the time of execution is executed. The method etc. can be applied. In any case, the above primary / secondary division is preset in the nodes A and B. With this setting, the following operations are possible. First, the control table interpretation unit Cpi in the control component Cp prepared for each of the nodes A and B interprets the control table CT and calls the application component execution control unit Cpc when the activation of the application processing component Ipi is required. The application component execution control unit Cpc is
The primary / secondary classification information of the own node (set in advance) is compared with the execution designation information of the application processing component Ipi requested (being activated) by calling the control table interpreter Cpi, and if it matches, the own node is executed , If they do not match, the other node is executed. In addition, applied processing parts Ipi
Is B (Both), the application component execution control unit Cpc determines parallel execution of the own node and the partner node. Then, the application component execution control unit Cpc activates the requested application processing component Ipi when the execution of the own node or the parallel execution is determined. On the other hand, when the partner node execution is determined, the application component execution control unit Cpc makes the control table interpretation unit Cpi appear as if the application processing component Ipi was activated, and does not actually perform activation control. The control is returned to the control table interpretation unit Cpi. This is called "empty control". By this "empty control", the control table interpretation unit Cpi can shift to the interpretation regarding the application processing component to be activated next. The application component execution control unit Cpc makes the above determination each time it is called from the control table interpretation unit Cpi. At this time, when the application component execution control unit Cpc detects that the continuous execution of the own node has changed to the execution of the other node, the application component execution control unit Cpc notifies the synchronization establishing unit Sm in the own node of the “other party activation request”. Synchronization establishment section
The Sm sends this request to the synchronization establishing unit Sm of the node via the logical path L1. In addition, the application component execution control unit Cpc
On the contrary to the above case, when it is detected that the execution of the other node has been changed to the execution of the own node, the WAIT (wait) of "partner activation request" is requested to the synchronization establishing unit Sm in the own node. In this case, the synchronization establishing unit Sm in the own node waits for the “other party activation request” from the synchronization establishing unit Sm in the other node, and when receiving this request, the application component execution control unit in the own node indicates that. Notify Cpc. The application component execution control unit Cpc waits until the synchronization establishment unit Sm in the own node receives the “other party activation request” from the partner node, and the activation of the application processing component requested by the control table interpretation unit Cpi is waited. . Then, when the synchronization establishing unit Sm in the own node receives the "partner activation request" from the partner node and the fact is notified to the application component execution controller Cpc, the controller Cpc is requested by the control table interpreter Cpi. Start the application processing component. At this time, in the partner node, if the "partner activation request" is sent from the synchronization establishing unit Sm, the application component execution control unit Cp
The control is returned to c, and the “empty control” of the partner node execution is repeated. Now, change from the other node execution to the own node execution,
What is important when activating the application processing component that is the target of execution of its own node is that the contents of the specified memory area on the communication memory unit PCM that changed as a result of execution of the other node are affected by the operation of the memory synchronous update mechanism u. It is reflected in the corresponding memory area of the communication memory section PCM of, and the contents of the designated memory areas of both nodes match. Therefore, when the memory synchronous communication mechanism u detects that the memory information has changed due to the writing to the specified memory area on the communication memory unit PCM in the node in which the mechanism u is placed, the communication memory unit PCM of another node is detected. In order to reflect the change in the corresponding designated memory area above, the write information is transmitted to the memory synchronous update mechanism u of another node along with the position information via the logical path L2. The memory synchronous update mechanism u that has received this transmission information
Using this transmission information, the write information is written at the corresponding position in the designated memory area on the communication memory unit PCM in the own node. By this operation, each communication memory block PCM in both nodes
The contents of the designated memory areas are matched. By the way, the application component execution control unit Cpc sets the synchronization flag at a predetermined position of the designated memory area of the communication memory unit PCM when notifying the synchronization establishing unit Sm in the own node of the “other party activation request”. There is. Now, the synchronization establishment unit Sm of the node A
It is assumed that the above-mentioned synchronization flag setting has been performed by. If the memory synchronous update mechanism u of the nodes A and B is processing so that the contents of the designated memory areas on the communication memory units PCM of the nodes A and B are matched, the predetermined value of the communication memory unit PCM of the node A is determined. The synchronization flag set at the position should be set at the corresponding position in the designated memory area on the communication memory unit PCM of the node B. Therefore, when the synchronization establishing unit Sm in the node B, which is the partner node of the node A, receives the "start request" from the synchronization establishing unit Sm in the node A, the synchronization establishing unit Sm of the designated memory area of the communication memory unit PCM of the node B is set to a predetermined position. The synchronization flag is checked, and the result of this check is used to check whether the matching of the contents of the designated memory area on the communication memory unit PCM (by the memory synchronization update mechanism u) is completed. When the synchronization establishing unit Sm of the node B confirms that the contents of the designated memory area match, the application component execution control unit Cp in the node B
Return control to c. With the above processing functions, for example, node A is defined as the primary node and node B is the secondary node.
When the application processing components Ip1, Ip2, Ip5 are designated as P and the application processing components Ip3, Ip3 are designated as S in the control table CT in the state defined as a node, as shown in FIG. In A, the application processing parts Ip1, Ip2, Ip5 operate, and in the node B, the application processing parts Ip3, Ip4 operate and a series of tasks are performed. If the resource (for example, a file) to be processed by the application processing component Ip4 is moved from the node B to the node A, the execution designation information corresponding to the component name indicating the application processing component Ip4 in the control table CT is displayed. By rewriting from S to P, the application processing parts Ip1, Ip2, Ip4, and Ip5 operate at the node A, and the application processing part Ip3 operates at the node B, as shown in FIG.
Also, for example, in both nodes A and B, the file x,
There is an application processing component Ipx that outputs the contents of the communication memory section PCM to this file x, and this Ipx is a node.
If you want to operate in parallel with A and B, control table CT
In, the application processing component Ipx may be designated as B (Both). By doing so, as shown in FIG. 5, both nodes A and B perform output processing to the file x in parallel. That is, the processing that does not change the contents of the communication memory unit PCM, such as reflecting the contents of the communication memory unit PCM in an external file, can be performed in parallel at the nodes A and B, so that the performance is significantly improved. When the application component execution control unit Cpc detects that the execution of its own node has been changed to parallel execution of both nodes or the parallel execution of both nodes has been changed to execution of the other node by the B designation, the application node execution control unit Cpc A "partner activation request" is notified to the synchronization establishing unit Sm therein. Also, the application component execution control unit Cpc
When it detects that the execution of the other node has been changed to parallel execution of both nodes, or that the execution of both nodes has been changed to parallel execution of its own node, it sends a “partner activation request” to the synchronization establishment unit Sm in its own node. WAIT of the other node to wait for the completion of the execution of the other node, and prevent the contents of the communication memory unit PCM from being changed during the execution of the other node. In addition, when it is detected from both node parallel execution to both node parallel execution, the execution can be continued as it is, notification of "other party start request" to the other node and
WAIT instruction is unnecessary. Although P (primary) and S (secondary) are used as the execution instruction information in the above embodiment, a node name or the like can also be used. Further, in the above embodiment, the case where two nodes perform one task has been described, but the present invention can also be applied to the case where three or more nodes perform one task. [Effects of the Invention] As described above in detail, according to the present invention, the following effects can be obtained. When attempting to perform work in cooperation with the nodes connected to the online network, create an application program assuming that each node has all the resources necessary for performing this work. In order to divide (divide) the work, the target work can be performed by simply defining only the execution instruction contents of the control table according to the actual resource distribution state.
Program creation is significantly easier. From the above, even if the resource distribution relationship between the nodes changes, it can be easily dealt with by simply rewriting the control cable.
When performing the same task in any of several distributed network systems having different resource distribution relationships, the same application program can be applied only by rewriting the control table. Since the adjustment according to the resource distribution relationship can be easily performed, it is possible to find the optimum isolation point of the interlocking performance between the nodes while verifying it while changing the resource allocation.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block configuration diagram of a distributed network system to which the present invention is applied, FIG. 2 is a diagram showing an example of a table structure of a control table shown in FIG. 1, and FIGS. FIG. 5 is a diagram showing an example of task sharing between the nodes A and B shown in FIG. A, B ... Node (distributed node), Cp ... Control component, CT ...
... Control table, Ip1 to Ipn ... Applied processing parts, PCM ...
Communication memory unit, Sm ... Synchronization establishment unit, u ... Memory synchronization update mechanism.

Claims (1)

(57) [Claims] In a distributed network system including a plurality of nodes interconnected by an online network, the above-mentioned nodes are prepared for each of a plurality of small jobs that constitute a target job and execute the corresponding small jobs. Application processing means common to each of the above nodes, a control table in which execution specification information for specifying an execution node for each application processing means is set, and this control table should be referenced to start next. Execution control means for deciding whether or not to operate the application processing means at its own node, and activating the corresponding application processing means in accordance with the result of the judgment, and communication memory means provided for information communication between the application processing means. And a change in the memory information in the designated memory area of this communication memory means is detected and the same information is transmitted to another node to specify the communication memory means of the own node and the other node. Distributed network system to the memory synchronous update means for matching the contents of the memory area, characterized by providing a synchronization establishing means between the own node and the other nodes to synchronize the operation of the application processing means.