JPH0823861B2 - Distributed processing method - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates to a distributed processing method, and more particularly to performing online processing and software eware development processing by the same method, and performing a large amount of input information at high speed by data flow type processing. The present invention relates to a distributed processing method capable of performing.

[Background of the Invention]

Along with the development of personal computers and system networking, there are changes in the way software is developed. That is, in the past, it was developed collectively by specialized software designers, but recently, software has been individually developed on-site by the end user's hands and transferred to other processing devices via the network. Therefore, there is a demand for a function to be executed there, or to create one piece of software by using resources of software and hardware distributed in the system.
The inventors of the present invention have already conducted a so-called data flow type distributed processing method (see Japanese Patent Application No. 56-31303) for executing a program in each processing device in a data flow manner, and a transmission control method (Japanese Patent Application No. 55-13725). As a result, the reliability, expandability, and maintainability of the distributed processing system should be greatly improved. However, no consideration has yet been given to how and in what system a program or data array in the processing device is created and stored in the processing device.

Furthermore, in information systems such as banks and securities systems,
It is required to process a large amount of data (transaction) in a short time. In this case as well, it is considered that high speed processing can be performed by applying the data flow type distributed processing method. However, in the past, that point has not yet been considered.

[Object of the Invention]

An object of the present invention is to solve these problems and to develop software for a distributed processing system with high reliability, high maintainability,
(EN) It is possible to provide a distributed processing method that can be realized with high expandability and that can process a large amount of input information at high speed by a data flow type distributed processing.

[Summary of the Invention] In order to achieve the above object, the dispersion treatment method of the present invention is
(A) A plurality of processing devices connected to the common transmission path communicate messages with each other via the common transmission path, the message including data to be communicated and a content code indicating the content of the data, and When each processing device detects that the data necessary for executing the stored program is input from the common transmission line by the content code, the distributed processing method for processing the data includes: i) arbitrary processing When the device creates a program, the created program with the content code added is sent as a message to the common transmission line, and ii) the other processing device that takes in the message including the created program is included in the message. The program for storing the program can be started by the content code to be input, and the input creation program can be executed on the processing device. It is characterized by storing a program. (B) When the plurality of processing devices connected to the common transmission line detect that all the information necessary for executing the stored programs is input from the common transmission line, the information is input. In a distributed processing method for executing processing by providing a reception buffer for storing input information, when the information is loaded into a program in the processing device,
When there are a plurality of pieces of information of the same type in the reception buffer as viewed from the content code, a plurality of pieces of information having the same content code are fetched in a batch, and input / output processing for that plurality of pieces is performed for all processing Command code to 1 or
Another feature is that one instruction code is shared by all processes so that they can be executed collectively.

Example of Invention

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a block diagram of a distributed processing system showing an embodiment of the present invention.

In FIG. 2, 11, 12, ... 1i, ... 1n are distributed processing units (hereinafter,
1i), 21, 22, ... 2i, ... 2n are transmission control devices (hereinafter, 2i is representative) for exchanging data with other devices in the system. Is a shared transmission line using a public communication network, a dedicated communication network, or the like. Each processing device 1i in the system has a transmission control device 2i connected to it.
Can be used to communicate with each other via the shared transmission line 3.

In the present embodiment, in a system in which the information flowing through the shared transmission line 3 is input to each processing device 1i and the processing device is driven by the input information, the program itself of this processing device The information is treated as the same as other information and is sent to the processing device in the same manner as the other information, and is stored in the processing device in a data flow manner so that it can be executed by each processing device. Further, in a data flow type processing device which inputs information flowing through the shared transmission path 3 to each processing device 1i, judges that all necessary information is gathered and executes a program in the processing device, it should be processed. When there is a plurality of input information, the speed is increased by collectively fetching and processing the information. For these two functions,
It will be described separately.

FIG. 1 is a block diagram of a distributed processing device showing an embodiment of the present invention.

As shown in FIG. 1, each processing device 1i in the processing system shown in FIG. 2 comprises a storage unit 80, interface units 30, 50, a terminal 60, a timer 70, and a calculation unit 40. The terminal 60 is externally arranged as a keyboard / display device attached to the processing device 1i, and is a user (operator).
With the interface. User can use this terminal
Using 60, Edit (a subsystem for text processing, an editing program for creating sentences) and Compile (a program that translates to machine language by an automatic translation performed by a compiler) built in the terminal 60 A program can be created by the function. The program created at the terminal 60 is a data flow type program that realizes the distributed processing method shown in the above-mentioned prior application (see Japanese Patent Application No. 56-31303).

FIG. 3 is a diagram showing a message format when transmitting and receiving through the shared transmission line of FIG.

In FIG. 1, the information transferred from the terminal 60 to the processing device 1i has the format shown in FIG. 3 like the information taken in from the shared transmission line 3 to the processing device 1i. That is, a content code indicating the content of the information is added, and for example, the information transferred from the terminal 60 to the processing device 1i is composed of the content code 100 and the data section 120 as shown in FIG. The information input from the transmission line 3 comprises a content code 100, an input information code 110, and a data section, as shown in FIG. 3 (b). In the processing device 1i, the information from the shared transmission line 3 and the information from the terminal are handled in the same manner. That is, it is determined whether or not information necessary for execution of the program stored in the processing device 1i is prepared, and only when the information is prepared, the program is executed (see the above-mentioned prior application specification. ).

In the present embodiment, the program created on each distributed processing device 1i using the terminal 60 is also sent on the shared transmission line 3 like other information, and is flowed on each distributed processing device 1i. A program is fetched as needed, and the program is checked by a data flow type processing method so that it can be executed by the fetched distributed processing device 1i.

Data of the format shown in FIGS. 3 (a) and 3 (b) is sent to the transmission line and is connected to each distributed processing device 1i.
Captured by 2i. The transmission control device 2i reads the content code of the fetched information and transfers the information to the processing device 1i if necessary. In the processing device 1i, in the program stored in the storage unit 80, it is processed as the input data of the program executed when the information having the content code is prepared. In this case, as shown in the specification of the prior application, the necessary content code is registered in the transmission control device 2i in advance.

A program created on at least one processing device 1i of the distributed processing system and transmitted to the transmission path 3 is written in the data section 120 of FIGS. 3 (a) and 3 (b). Looking at the content code, the program loaded into the processing device 1i is treated as one data, as in the case of processing other general information, and is driven by that data. The program runs.

This program to be executed is a process for registering and storing as a program in the processing device 1i after the flowing program is converted into a form executable by the data flow type execution processing method in the processing device 1i. It is something to do.

 Further description will be made using a specific example.

The following assumptions are made for the program executed in the processing device 1i. Execution of the processing method according to the present embodiment,
It is started when all the information necessary for executing the program is prepared from the information taken in from the transmission path 3. Therefore, it is assumed that the plurality of programs stored in the storage unit 80 can be independently executed one by one without interfering with each other. Here, it is assumed that the creation program flowing on the transmission path 3 is in machine language (object) format and is relocatable. A statement is a statement written using a computer language, and a source module is a collection of statements in a program that makes a logical meaning. A source module is translated into a machine. The term used is the machine language module (object module). In this case, the source module creates the machine code module of the relative address,
If the relative address is not changed to the absolute address, it will not become executable. This conversion work is relocation.

In the present embodiment, the creation program flowing through the shared transmission line 3 is called a development program, and the program in the storage unit 80 that performs a series of processes using the development program as input information is called a constituent program. Regarding the processing of the development program by the constituent programs, a fixed block method and a variable block method can be considered, depending on the difference in the loading (loading) of the development program in the storage unit 80 and the type (addition or modification) of the development program.

(A) Fixed block method: The maximum length of each program is limited and the start address to be stored in the storage unit 80 is determined in advance. The configuration program has the value of this address as internal data. When the processing method of the present embodiment is executed, the development program flowing through the transmission path 3 has the content code 100 and the content of the input information of the development program, as shown in FIG. 3 (b). A code 110 is added, and a data section 120 is embedded after that.

FIG. 4 is an explanatory diagram of a development process of a distributed processing program showing an embodiment of the present invention, FIG. 5 is a storage state diagram of a storage unit in the processing device of FIG. 4, and FIG. FIG. 7 is a correspondence diagram of the content code and the start address of the constituent program in FIG. 4.

Now, as shown in FIG. 4, consider a case where the program A created in the processing device 1i is flowed along the dotted arrow through the transmission control device 2i and the shared transmission line 3. In the transmission information, the content code CCK of the information, the development program A, and the content code CCA of the input information when the program A is executed are embedded. The transmission control device 2j looks at the content code CCX of the information that has flowed in and fetches it as necessary information in its own processing device. It is assumed that the programs X and Y are now stored in the storage unit 80 in the processing device 1j connected to the transmission control device 2j. From the content code CCX of the information taken in the processing device 1j, the constituent program X having A as input information is read by the arithmetic unit 40, stored in the arithmetic register 90, and executed. The processing content of A is performed in the order of to be described below.

Change of address of program A ... Since the program A is described by a logical address (generally, address 0 is the start address), it cannot be executed on the arithmetic unit 40 as it is. Therefore, the address of the constituent program X is changed. Now, as shown in FIG. 5, when the program A is stored from the address aaa of the storage unit 80, it is necessary to add only the relative address between the head address of the logical address and aaa.

Storing of program: The program A is stored in the storage unit 80 from the head address aaa. The method of defining aaa in the configuration program is performed by the following method. The configuration program X has, as internal data, the content code and start address of the input information of the program stored in the processing device 1j and the start address of the newly added program as internal data. Fig. 6 shows an example in which X, Y, Z are existing programs. CCX, CCY, CCZ are content codes, xxx, yyy, z
zz is the start address, and the new program (NL) is
It indicates that the data is stored from aaa. NL indicates that it has not yet been stored. When the development program is fetched, the constituent program looks at the input information content code 110 (CCA in FIG. 4) of the development program, and from this content code, it is judged whether it is an existing program in the processor 1j. Search from the table in the figure. In the case of an existing program, the start address of the program is read from the table and stored from that start address. If it is not an existing program, the storage start address (aaa in FIG. 6) of the new program is read and stored from this address aaa. In this case, the table CCA NL
To update. Next, when storing a new program, it is stored from aaa + k.

Input of development program Content code registration …… CC
Register A in the transmission control device 1j.

By the above operation, the program A is stored in the storage unit 80 of the processing device 1j, and when the information having the content code of CCA flows into the shared transmission line 3, it is fetched by the transmission control device 2j and is processed by the processing device 1j. This is done above.

(B) Variable block system ... In the fixed block system described above, the actual program length is shorter than the block length, so that there is a space in the storage device 80. The variable block system eliminates this space.

When this method is implemented, the size of the development program is added as data to the input information content code 110 of the data portion of the information flowing on the transmission path 3. The constituent program reads the data and sets the start address to be stored in the development program flowing next to the development program. For example, in the case of FIG. 4, FIG. 5 and FIG. 6, assuming that the size of the program A is k _a , the configuration program X stores A from the address aaa and then stores the next storage start address. It can be set to aaa + k _a (see FIG. 6). The other processing contents of the configuration program are the same as those of the fixed block system described above.

FIG. 7 is an overall block diagram of a distributed processing system showing another embodiment of the present invention.

The system configuration of FIG. 7 is almost the same as the system configuration shown in FIG. 2 except that a plurality of external devices 60 are installed in the processing device 1i. External device 60
Is a device such as a secondary storage device, a display device, a printer, etc., and information is read / written by an input / output command of a program executed in the arithmetic unit 40. Seventh
Also in the processing device 1i shown in the figure, the data flow type process for executing the program in the processing device 1i is performed by inputting the information flowing through the shared transmission line 3 and determining that the necessary information is prepared. . When there are a plurality of input information items to be processed, these information items are collectively fetched and processed to speed up the process.

FIG. 8 is a diagram showing a transmission format of information flowing on the shared transmission line of FIG. 7.

In the figure, 100 is a content code and 120 is a data part. The difference from the data format of FIG. 3 is that the data section 120 of FIG. 3 is the program itself, whereas the content of the data section 120 of FIG. 8 is intended for general information such as numerical data. That is.

In this distributed processing system, the information flowing on the transmission path 3 is taken into the transmission control device 2i connected to each distributed processing device 1i, the content code 100 of the information is read, and compared with the content code registered in advance. , When it is necessary,
This information is transferred to the connected processing device 1i. The processing device 1i constantly monitors whether or not the input information of the programs stored in the storage unit 80 is gathered, and when the information is gathered, the program that receives the information is executed (for detailed operation, see , See the aforementioned Japanese Patent Application No. 56-31303).
As a result, when performing communication via the transmission path 3, it is not necessary to specify an address or execute an execution request, and all the processing is executed by seeing the content code only in each processing device 1i. Make a decision. Therefore, there is no parent-child relationship in the system, and the reliability and maintainability of the system can be improved.

2. Description of the Related Art In recent years, information systems including banks and securities systems mainly focus on processing a large amount of data (transaction) online, and particularly this processing includes access to almost all secondary storage devices. It has the characteristic. In the present embodiment, it is possible to sufficiently withstand a large amount of high-speed information processing without changing the data flow processing.

FIG. 9 is a data storage state diagram in the storage unit in FIG.

In the processing device 1i, the execution of the program of which the input information is complete among the programs A _{1 to} A _N stored in the storage unit 80 is started. When the input information of (1) further arrives, it is combined into a queue and waits for processing until the execution of the program ends. Therefore, when a large amount of information to be processed is received as compared with the processing time of one program, the information to be processed is accumulated in a queue.

Here, the queue is on the transmission control device 2i or the processing device.
It is assumed that it exists within the buffer of 1i and can be monitored by the processing device 1i.

In FIG. 9, 80 is a storage unit, 200 is an execution management table, 300 is a queue register, 100 is a content code, 120 is a data section, and A _{1 to} A _N are programs that start execution when input information is complete.

In this embodiment, Assortment ivy information capture pair while, instead of running the program A ₁ to A _N, the input information having the same content code 100 captures a plurality of sets collectively, attempts to process at the same time To do. Therefore, it is assumed that the programs A _{1 to} A _N stored in each distributed processing device 1i have a structure for processing these plural sets of information.

In FIG. 9, when the input information is sequentially input, these are first combined in the queue register 300, and the program A
_{By j} , all the information linked to the queue register 300 is sequentially registered in the execution management table 200. Program A
_{When j} is registered in the execution table 200, it is collectively registered for each type according to the content code 100. Then, as indicated by the dotted arrows, the programs A _{1 to} A _N in the storage unit 80 are sequentially used for each content code 100 using the information.

FIG. 10 is a detailed diagram of the execution management table in FIG.

Now, as shown in FIG. 9, it is assumed that the storage unit 80 stores _N programs A ₁ , A ₂ , ... A _N , and each of these programs has a content code of each input information. CC1,
If there is information that has CC2 ... CCN, it is possible to start processing by using that information as input. Then, the information of the content code CC1 is processed by the program A _1, information code CC2 is processed by the program A _2, information of the code CC3 is processed by the program A _3, it is processed sequentially program corresponding to the content code below Shall be. Program A _j
As shown in FIG. 10, the information linked to the queue register 300 is arranged in order by content code and registered in the execution management table 200. As a result, the information of code CC1 is n ₁
Number, _two information of the code CC2 is n, information of ...... code CCN is n _N
Assuming that there are two pieces of information, the program A ₁ processes n ₁ pieces of information at the same time, the program A ₂ processes n ₂ pieces of information at the same time, and similarly, other programs simultaneously output corresponding pieces of information. To process.

Various processing structures are conceivable for a plurality of sets of input information. This will be described below.

11 (a), (b), and (c) are processing flow charts of the programs shown in FIG.

As shown in FIG. 11 (a), in the case of executing processing of n sets of input information by a program consisting of processing a, b, c, ... The structure shown can be used. Particularly, in the structure of (b), the I / O processing part can be collectively performed. That is, in FIG. 11 (b), initial setting is performed for each program a, b, c ,.
Repeat execution from = 1 to i = n (Do). On the other hand, in the case of a method that can be set commonly to programs, as shown in FIG. 11 (c),
First, after repeatedly inputting “Do” (i) from i = 1 to i = n only once, the processes are continuously executed from the programs a to z.

As shown in FIG. 9, the programs A _{1 to} A _N are stored in the processor 1i, and each of the programs A _{1 to} A _N has a content code CC1 or CC2 of each input information as shown in FIG. , ...... information I _1, I ₂ with CCN, ... when I _N exists, performs processing such information as input, each output _{O 1, O 2, ...... O} N
Shall be output. Now, the above input / output process
I will write as follows.

O ₁ ＝ f _A1 (I ₁ ) O ₂ ＝ f _A2 (I ₂ ) O ₃ ＝ f _A3 (I ₃ ) ・ ・ ・ O _N ＝ f _AN (I _N ) …… (1) CCi, I _i , O _i may be a scalar or a vector. If it is a scalar, the program has one input / output, and if it is a vector, it has multiple input / output, and it can be said that they are related to each other.

In the present embodiment, for example, if there are n _i pieces of information having CCi content codes in the queue 300, the program A _i
Process. Maximum number n of information to be fetched in a batch
_i is set in advance for each program and held as data in the program A _j in FIG.

In this way, in this embodiment, the information that has reached the processing device 1i via the shared transmission line 3 is arranged in the queue 300 in the order of arrival. Then, the arrival information is taken into one program A _j in the storage unit 80 that detects and executes the management of the arrival of information, and the content code of each information is referred to
Arranged as shown in Figure 10. Incorporation of information from Kiyu by the program A _j is performed until either the number of information with one content code is maximum uptake number n _i, or the information in the Kiyu eliminated. Further, when writing to the table 200, the number of fetched information corresponding to the content code is used as data and the data section 1 of the execution management table 200 of FIG.
Add to 20.

The information taken in by the program A _j becomes the input information of the corresponding program by the content code, and the processing is performed. In that case, in the method of the present invention, when each program is executed, information is collectively fetched by the number of information data portions 120, and the processing is collectively performed.

As described above, in this embodiment, a series of processes in the distributed processing system can be performed in a data flow manner without forming a parent-child relationship in the system, and the system reliability, maintainability, and expandability are improved. Can be made
Since a large number of transactions can be collectively processed according to the content code, it is possible to speed up the processing.

〔The invention's effect〕

As described above, according to the present invention, for a distributed processing device that inputs and drives information that flows via a shared transmission line, the program itself also flows as one piece of information on the transmission line, and another processing device. Since this is executed in, the expandability becomes great and the software development becomes easy. Further, according to the present invention, even if a large amount of normal information flowing through the above-mentioned transmission line is input to each processing device, it is possible to process them all at once, enabling high-speed processing and scalability. Also improves.

[Brief description of drawings]

FIG. 1 is a block diagram of a distributed processing apparatus showing an embodiment of the present invention, FIG. 2 is an overall configuration diagram of a distributed processing system of the present invention, FIG. 3 is a format diagram of a message used in FIG. Fig. 4 is a diagram showing the software development process in Fig. 1, Fig. 5 is a data storage state diagram of the storage unit in Fig. 4, Fig. 6 is a correspondence diagram between the content code and the start address in Fig. 4, FIG. 7 is a block diagram of a distributed processing apparatus showing another embodiment of the present invention, FIG. 8 is a diagram showing a message format used in FIG. 7, and FIG. 9 is an explanatory diagram showing a processing state of FIG. FIG. 10 is a detailed diagram of the execution management table in FIG. 9, and FIG. 11 is a diagram showing the processing method of each program in FIG. 11-1n, 1i: Distributed processing device, 21-2n, 2i: Transmission control device, 3:
Shared transmission line, 30, 50: Interface, 40: Arithmetic unit, 60: External device, 80: Storage unit, 70: Timer, 100: Content code, 110:
Input information content code, 120: data part, 200: execution management table, 300: queue register.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Masayuki Oshige 1099 Ozenji, Aso-ku, Kawasaki-shi, Kanagawa, Ltd. Inside the Hitachi, Ltd. System Development Laboratory (72) Inventor Minoru Koizumi 1099, Ozen-ji, Aso-ku, Kawasaki, Kanagawa Hitachi, Ltd. System Development Laboratory (72) Inventor Kozo Nakai 5-2-1 Omika-cho, Hitachi City, Ibaraki Prefecture Hitachi Ltd. Omika Factory (72) Inventor Hirokazu Kasashima 5-chome, Omika-cho, Hitachi City, Ibaraki Prefecture No. 1 Incorporated company Hitachi Ltd. Omika factory (56) Reference JP-A-57-146361 (JP, A)

Claims (4)

[Claims] 1. A plurality of processing devices connected to a common transmission line communicate messages with each other via the common transmission line,
The message includes data to be communicated and a content code indicating the content of the data. Each of the processing devices uses the content code to indicate that the data necessary for executing the stored program is input from the common transmission path. When detected,
In a distributed processing method for executing the processing of the data, i) any processing device creates a program, and sends the created program with a content code added as a message to the common transmission path, and ii) the creation Another processing device that has received the message including the program may activate the program for storing the program by the content code included in the message, and store the input creation program as a program executable on the processing device. Characterized distributed processing method. 2. A processing device which takes in a message including the creation program stores the creation program in a predetermined storage area and sets a condition for activating the creation program in a management table. The distributed processing method according to claim 1. 3. When a plurality of processing devices connected to the common transmission line detect that all the information necessary for executing the stored programs has been input from the common transmission line.
In the distributed processing method of inputting the information and executing the processing, a reception buffer for storing the input information is provided, and when the information is taken into a program in the processing device, the content code is stored in the reception buffer as seen from the content code. When a plurality of pieces of information of the same type are present, a plurality of pieces of information having the same content code are collectively fetched, and the input / output processing corresponding to the plurality of pieces is added with an instruction code for every processing or A distributed processing method characterized by executing instructions in a batch by sharing the instruction code with all processes. 4. The input information is in the form of a message.
The input information having the content code and the data portion is collectively fetched for each input information having the same content code among a plurality of information, and the fetched input information is processed by the same program stored in the processing device. The distributed processing method according to claim 3, wherein: