JP5223909B2 - Allocation system, allocation device, and allocation program - Google Patents

Description

The present invention relates to a full-text search system of a character string matching method that searches a large amount of data in a short time using a plurality of processing devices.
Here, the character string matching method is a method in which the specified character string and the text data to be searched are sequentially matched from the beginning of the text data to be searched toward the back of the text data to be searched. It is a method to check whether it exists.

Conventionally, as a method of shortening search time using a plurality of search processing devices in Japanese Patent Laid-Open No. 8-137734 “Information Processing System and Database Distribution Method”, a small-scale information processing device such as a personal computer is used. Connected with a personal computer LAN, the data is distributed and arranged so that each information processing device is almost uniform, the data processing is performed in parallel in each information processing device, and the processing result is one information processing device. A method has been proposed which is integrated and displayed.

  Further, there has been considered a method in which search target data is divided into a plurality of areas, and the divided area units are assigned to different search processing apparatuses, and each search processing apparatus performs search processing in parallel.

  For example, Japanese Patent Laid-Open No. 9-13364 “Information Search System” includes a plurality of search servers that perform search processing and a search management server for managing the operations of the plurality of search servers. The text base to be searched and the related information about the text base are respectively divided, and a part or all of the plurality of search servers are formed by combining the divided part of the text base and the related information corresponding to the divided part. And a plurality of search servers, each configured to perform information search for the text-based divided portion allocated by the search management server in parallel and independently. Even when searching the text base at the same time, no special hardware is required and the text to be searched To perform the information retrieval at high speed without also being affected by the magnitude of Tobesu method of performing concurrently search process has been proposed.

However, in the first conventional technique, when the database is divided into files of each device physically independently, the number of search processing devices increases or decreases due to an increase in search processing devices or failure of the search processing devices. When this occurs, physical re-division of the search target data is required, and the system stop time for the processing becomes longer.

In the case of the second prior art, when the data is added or a part of the data is changed, the entire file is rewritten. When the file is huge, the rewriting process is not performed. It takes a lot of time.
Therefore, in a full-text search system of a character string collation method in which search processing and additional data update processing are performed in parallel, it will not be practical unless the problem of the rewrite processing time is solved.
Further, by using many search processing devices, it is considered that the possibility that a failure will occur in any of the search processing devices is increased accordingly.

  Therefore, the present invention can update and add search target data in parallel with the search, and when a failure occurs in any of the search processing devices, the search processing device in which the failure has occurred It is an object of the present invention to provide a means that makes it possible to quickly perform the degenerate operation with the removal of the.

FIG. 1 shows a configuration diagram of the present invention.
In the figure, 1 is search target character string data.
That is, it represents character string data that is a search target for the original search, and is logically connected character string data stored in a magnetic disk device or a semiconductor storage device.
These are divided into records that can be identified by scanning and reading character strings, and these character string records are collected and stored as a set of information.
The search is performed so that the character string information constituting the record is collated based on the search condition to find and output a satisfactory character string record.

Reference numeral 4 denotes a search processing apparatus which is composed of a plurality of units.
The full-text search system according to the present invention distributes the search target character string data 1 to the plurality of search processing devices 4 so as to perform search processing. Reference numeral 2 denotes the search target character string data 1 as a character string record group. Search integration means for dividing and assigning to a plurality of search processing devices 4, transmitting a given character string search condition to each search processing device 4 as a search instruction, and receiving and integrating search results from each search processing device 4 It is.

Normally, the content of character string data to be searched is not fixed, and new data needs to be added or updated.
In the present invention, when the record of the search target character string data 1 is additionally updated so that the addition and update of such data can be performed in parallel during the execution of the search process, an additional update unit (not shown) Is provided with additional update temporary storage means 3 for storing the record additionally updated.
Thus, the contents of the search target character string data 1 are not changed at the time of addition or update, and the addition / update data is stored in the additional update temporary storage unit 3.

  In order to make the additionally updated record a target of the search processing, an addition is made to instruct one of the predetermined search processing devices 4 as a search charge data the character string record group stored in the additional update temporary storage means 3 An update record search instruction means 8 is provided.

  The record before update corresponding to the updated record stored in the additional update temporary storage means 3 is deleted from the search target character string data 1 so that the additional updated record can be reflected in a timely manner. Update result reflecting means 5 for incorporating the additionally updated record into the search target character string data 1 is provided.

When a large number of search processing devices 4 share processing, and there is a faulty device among them, the search processing device is excluded and the search processing is shared and executed.
That is, after the search integration means 2 transmits a search instruction to the plurality of search processing devices 4, a search that stores the time at which the search result is received from the search processing device 4 assigned by dividing the search target character string data 1. Failure search processing for determining, as a failure device, the result processing time storage means 6 and the search processing device 4 that has not received the search result within a preset time from the search result reception time of the search processing device 4 received first. Device determining means 7.

  In addition, the search processing device 4 instructed as the search charge data using the record group stored in the additional update temporary storage means 3 is also a malfunctioning device when the stored record group increases and the search time becomes long. In this case, after the update result reflecting means 5 incorporates the additionally updated record into the search target character string data 1, the search process is executed again.

As is apparent from the above description, according to the present invention, in a system for retrieving a character string record of a condition given by a character string collation method from a large amount of data composed of a sequence of character strings with record delimiters. In parallel with the processing, a large amount of data to be searched can be added and updated. This facilitates the operation of a large-volume data search system and allows users to search for the data to be searched. Since the latest information that has been added or updated from time to time can be searched, the quality of search data can be improved and the reliability can be improved.
In addition, even in the mechanism of the present invention for increasing the search speed by simultaneously operating a large number of search processing devices for searching a large amount of data, the allocation of search target data shared by each search processing device can be changed in a short time. Therefore, even if some trouble occurs in the search processing device, it can be operated with almost no interruption in the configuration with the reduced number of search processing devices, improving the reliability of the search system as a whole and operating it. The rate can be improved.

Configuration diagram of the present invention Overall concept of full-text search system Search target data example Search request example Example of search request screen Search result screen example Overall configuration of full-text search system Configuration diagram of reception server Configuration diagram of search server Overall flow of processing Preprocessing block diagram Director program flow chart (1) Example of environment table (1) Response table example Example of search server a allocation table (1) Flow chart of search server initialization program Search data example Update processing block diagram Data update program flow chart Update file response table example Block diagram of search processing Director program flow chart (2) Search request buffer Search result table example (1) Search thread flow diagram Search result example (1) Search result table example (2) Flow chart of full text search program Flow of time monitoring program processing Director program flowchart (3) Example of environment table (2) Reassign search area Example of allocation table of search server a (2) Overall configuration diagram of the second embodiment General reception server configuration diagram Processing block diagram of the second embodiment Example of environment table G (1) General reception server director program flowchart (2) Example of search result table G (1) Flow chart of reception thread Search result example (2) Example of search result table G (2) Example of environment table G (2)

In the present embodiment, a form realized by a computer program executed on a computer used for general purposes such as a personal computer or a workstation is shown.

The full-text search system of the present invention is realized by executing a computer program on a computer including a processing device, a main storage device, an input / output device, and the like.
The computer program is provided by being stored in a floppy (registered trademark) portable medium such as a disk or a CD-ROM, or in a main storage device or an auxiliary storage device of another computer connected to the network.
The recording medium of the present invention corresponds to the portable medium, the main storage device, and the auxiliary storage device.

The provided computer program is loaded directly from the portable medium into the main storage device of the computer, or in a computer equipped with an auxiliary storage device, the program is temporarily copied from the portable medium to the auxiliary storage device and then installed in the main storage device. Load and execute.
Also, when stored and provided in another device connected to the network, after being received from another device via the network, it is loaded into the main storage device, or in a computer having an auxiliary storage device, the auxiliary storage device Are copied into the main storage device and executed.

FIG. 2 shows the overall concept of the full-text search system of the present invention (hereinafter referred to as the present system).
In response to a search request that occurs irregularly from a plurality of terminals (terminal A, terminal B, terminal C), the present system selects from search target data (corresponding to search target character string data 1 in FIG. 1). It is an apparatus that extracts information satisfying each search request and returns the information to each terminal as a search result.
Further, in parallel with the search process, this system can add new data to the search target data and change the search target data in response to a data addition / update request from the terminal (terminal D).

First, an example of the contents of search target data is shown in FIG.
The search target data is composed of a plurality of records.
A record is composed of a plurality of items.
The content of the item is a string.
Numerical values and dates are also expressed as character strings.
Records can be identified by means such as record delimiters.
Items in the record are delimited by an item delimiter code, and can be uniquely identified by an item identification code (hereinafter referred to as an item tag).
In addition, a record NO, which is a code that can uniquely identify the record, is assigned to the record.
Further, description will be made assuming that the search target data according to the embodiment of the present invention is sufficiently large.

An example of a search request is shown in FIG.
5 and 6 show examples of search request screens and search result screens when a search is performed from the terminal.
In this example, the search request has an enumeration of information in which item tags and search terms are paired, and a search condition expression.
These pieces of information mean that a search is made for a record in which the specified search word exists in the item specified by the item tag in the search target record and the search condition expression is satisfied.

Specifically, in this search for personnel information, a name and a school of origin are input as search conditions, and personnel information satisfying both is searched as related information.
In addition, when only the name, name and birthplace, or information known on the input screen is input on the search request screen, the information including the related information is returned when satisfactory information is found.
As described above, the conditional expression shown in the example of the search request shown in FIG. 4 is automatically created by the combination of the inputted information.

In this way, the system finds a record satisfying the search request from the file, and returns a part or all of the record to the search request source such as a terminal screen.
In the case of this method, search processing is performed in parallel using a plurality of search processing devices (corresponding to the search processing device 4 in FIG. 1), so that the search time is reduced in proportion to the number of search processing devices. .

In the case of this method, the divided area is not a physically independent file, but the search target data is configured as one physical file or a logically connected file, in units of records. Then, the area is logically divided into the number of processing apparatuses, and each area is assigned to another search processing apparatus to perform a search.
As a result, when the number of search processing devices increases or decreases, a new file is logically subdivided into the number of search processing devices, and each region is quickly reassigned to the search processing devices. You can continue processing.

The file format in this method is a text format file in which character information is simply arranged.
However, since it is premised that the search target data handled by the full-text search system targeted by the present invention is sufficiently large, when the search target data is collected into one file, the file becomes huge.

In the above text format file, when data is added or a part of data is changed, the entire file is rewritten. If the file is huge, the rewriting process requires a long time.
Therefore, in the full-text search system of the character string collation method in which the search process and the additional data update process targeted by the present invention are performed in parallel, it will not be practical unless the problem of the rewrite processing time is solved.
Moreover, since this system uses many search processing devices, the probability that any of the search processing devices will fail increases.
Therefore, when a failure occurs in any of the search processing devices, it is necessary to quickly perform the degenerate operation by removing the search processing device in which the failure has occurred.

FIG. 7 shows an overall configuration diagram of the embodiment of the present invention.
This system includes one reception server, a plurality of search servers, and one file server.
The reception server receives a search request from the terminal and passes it to the search servers a to e.
The search servers a to e find a record satisfying the search request from an area of the search target file that is handled by the search server, and return the result to the reception server.

The search server e finds a record satisfying the search request from the update file and returns the result to the reception server.
The reception server collects the responses from all the search servers and returns them to the terminal.

Further, the reception server performs addition registration of new data or registration of changed data in the update file in response to a request for addition and update of data to the search target file from the terminal.
The file server is a device that reads and writes data of the search target file and the update file in response to a request from the reception server or the search server. Since the file server is a general device, a description thereof is omitted in this embodiment.

Next, details of the reception server will be described.
The configuration of the reception server is as shown in FIG.
The reception server is composed of an information processing device, a magnetic file device, and a program storage device.
The components of the program storage device are a control program, a director program (corresponding to the search integration means 2 in FIG. 1), a data update program, a time monitoring program (search result reception time storage means 6 and failure search processing device determination in FIG. 1). Corresponding to the means 7), search thread, environment table, response table, update file response table, search request buffer, and search result table.

Here, the data update program is a program for updating a record of a file storing character string data to be searched.
Records are updated by the receiving server that manages the search target.

  The director program distributes search target character string data to multiple search servers, passes search requests received from the terminal to each search server, and integrates the search results returned from each search server and returns them to the requester. It is a program that plays a role of integrating the divided search.

  The time monitoring program requests a search request from a search server, and then monitors a response from the search server to detect an abnormality of the search server that takes much longer than the response of another search server.

  The search thread sends the search requests that arrive to the search server in parallel.

  The environment table, response table, and update file response table are tables that manage the distribution of search processing to the search server, as will be described in detail later.

The search request buffer is a buffer for accumulating search requests from the terminal in the reception server as a process waiting state, and is sequentially transmitted to each search server as a search instruction.
The search result table accumulates the results of searching the search target data in the divided areas of each search server. When the results of all the search servers are collected, the search results are integrated and sent to the requesting terminal. Will do.

Next, details of the search server will be described.
The configuration of the search server is shown in FIG.
The search server includes an information processing device, a magnetic file device, and a program storage device.
The components of the program storage device are a control program, an initial setting program, a full text search program, an allocation table, and search data.

The flow of processing will be broadly divided into four processes: pre-processing, update processing, search processing, and area reallocation processing.
FIG. 10 shows the relationship between the four processes.
The pre-process is a process from when the system is activated until a search request or a data update request is received from the terminal.
The update process is a process for adding new data and registering changed data in response to a request from the terminal.
The search process is a process from receiving a search request from a terminal until returning a search result to the terminal.
The area reassignment process is a process for reallocating the data area to the search server when the search process ends abnormally in the course of the search process.

First, pre-processing will be described.
FIG. 11 shows the relationship between the basic functions of the preprocessing and the table.
FIG. 11 shows the tables and processes related to the pre-processing operation in the configuration diagrams of the reception server and the search server shown in FIGS. 8 and 9, and further, the search area of each search server for the search target file An example of sharing and the relationship with an update file are shown.

The search target file shown in FIG. 11 illustrates a state in which the search server a, the search server b, and the search server d share an area.
Each area is assigned in units of records. For example, the search server a is assigned to record Nos. 1 to 784, and the search server b is assigned to 785 to 1568.
Usually, in order to equalize the load, the number of records, the length of the character string, or the sum of the sizes is divided so as to be almost equal.
An update file (corresponding to the additional update temporary storage means 3 in FIG. 1) stores an added or updated record as a file different from the search target file.

The processing flow of the reception server and the search server will be described in detail with reference to FIG. 11 having such contents.
In the operation of the reception server, first, after the control program of the reception server initializes the system, the director program is started.
FIG. 12 shows the flow of processing after the director program is started.

In step S1201, the director program first merges the data of the update file with the search target file (corresponding to the update result reflecting means 5 in FIG. 1).
The update file is a file that temporarily stores the updated or added records of the search target file, so there may be no contents, but usually the records accumulated during the previous operation are stored, so search When restarting for some reason, such as starting or interruption of search, the file is first merged with the search target file.

At this time, if data of the same record NO exists in the update file and the search target file, the data of the search target file is replaced with the data of the update file.
After the merge process, all the records in the update file are unified into the search target file, and therefore all the data in the update file is deleted.

Next, in step S1202, the director program creates an environment table.
As shown in FIG. 13, the environment table is a list of search servers, and includes a search server name, an address (network address), a status flag, and an update file charge flag.
The status flag indicates the status of the search server.
For example, “1” is set when the search server is not usable, and “0” is set when the search server is usable.
The update file charge flag indicates that the search server is in charge of searching for update files.
Information other than the update file charge flag is recorded in the magnetic file device in advance, and the director program obtains information from the information and sets the environment table.

In step S1203, the director program reads the search target file and creates a response table.
As shown in FIG. 14, the response table is a list of all records in the search target file, and includes a record NO, a head address of the record in the file, and a length of the record.

Next, in step S1204, the director program creates an allocation table for each search server based on the environment table and the response table.
The allocation table is a table in which the search coverage of each search server is allocated as shown in FIG. 15, and among the search servers in which the status flag of the environment table (illustrated in FIG. 13) is the usable state “0”, Create all but one.
Further, the update file charge flag (corresponding to the additional update record search instruction means 8 in FIG. 1) of the remaining search server in the environment table (FIG. 13) is set to ON “1”.
The allocation table is a list of records that each search server is responsible for searching, and is composed of a record NO, the start address of the record, and the length of the record.
The records handled by each search server are checked from the top of the response table so that the number of records in each allocation table or the sum of the record lengths is equalized by looking at the response table (FIG. 14). Assign to.

In step S1205, the director program notifies the allocation server (FIG. 15) together with the search target file name to the corresponding search server.
In addition, an empty allocation table is notified together with the update file name to the search server in which the update file charge flag of the environment table is ON “1”, and then the search request wait state is entered.

  The above is the description of the operation of the preprocessing of the receiving server along the flow of the director program of the receiving server in FIG.

Next, the pre-processing operation in the search server will be described.
First, after the initialization of the system, the search server control program starts the initial setting program and the full-text search program.

FIG. 16 shows the flow of processing of the initial setting program after being started.
In step S1601, the initial setting program receives the allocation table from the receiving server. If there is an allocation record in the allocation table in step S1602, in step S1603, all records of records NO are read from the search target file into the program storage device. Search data (FIG. 17) is created.

If there is no allocation record in the allocation table, the search server is in charge of the update file, and the process ends without doing anything.
Thereafter, after the full-text search program is started, it waits for a search request.

Next, the update process shown in FIG. 10 will be described following the description of the preliminary process.
FIG. 18 shows the relationship between the basic function of the update process and the table.
The processing flow of the reception server in the update process will be described in detail with reference to FIG.

The control program of the receiving server starts the data update program together with the director program after system initialization and the like.
FIG. 19 shows a flow of processing of the data update program after being activated.

After the data update program is started, it initializes the program and waits for a data update request.
After receiving the data update request, if the sent data is new additional registration data in step S1901, the record NO of the maximum record NO + 1 existing in the conventional search target file and the update file is assigned, and the update file Record at the end of.
If the sent data is changed data, it is recorded as it is at the end of the update file.

If the data of the same record NO as the sent change data exists in the update file, the existing update file data is deleted.
After registering data in the update file, an update file response table (FIG. 20) is created in step S1902, and the data update request is waited again.

Next, search processing in the overall flow of the processing shown in FIG. 10 will be described.
FIG. 21 shows the relationship between the basic function of the search process and the table.
With reference to FIG. 21, the processing flow of the receiving server and the search server in the search process will be described in detail.
As shown here, when a search request is received from the terminal, as the director program sets in the pre-processing, the records of the allocated areas of the search target file are transferred to the search servers a, b, and d, and the update file record The search request is transmitted to each search server such that the search server e is in charge of the search.

At the end of the processing flow of FIG. 12, the director program waits for a search request when the preprocessing is completed.
The subsequent processing flow is shown in FIG.

The search request from the terminal is stored in the search request buffer (FIG. 23).
In step S2201, when a search request is received from the terminal, the director program retrieves the search request from the search request buffer, and first creates a search result table (FIG. 24) for each search request.

  In the search result table, a list of search server names in a usable state of the environment table (FIG. 13) is created, the update file charge flag of the environment table is transcribed, and further, a process result flag, a process completion time, and a hit record NO. Provide a column.

Next, in step S2202, the director program activates search threads corresponding to all search servers in a usable state.
Simultaneously with the activation, the search thread is notified of the search request (FIG. 4), the address of the search server in charge in the environment table (FIG. 13), and the column where the answer is written in the search result table (FIG. 24).
Search threads are created for each search request by the number of search servers in a usable state.

  After starting the search thread, the director program waits for a search result notification in step S2203, and the subsequent processing will be described later.

The flow of processing of the search thread is shown in FIG.
After the search thread is activated, it receives a search request, and notifies it to the full-text search program of the search server in charge in step S2501.
In step S2502, the search result is awaited. In step S2503, the search result is confirmed. When a hit record NO (FIG. 26) is received from the search server as an answer to the search result, the answer is stored in the search result table (FIG. 26). 27) is written in the hit record NO column, and in step S2505, the current time is written in the processing completion time column. In step S2506, the processing result flag is set to the normal end state “1”.
If there is an abnormal end reply from the search server, the processing result flag column is set to an abnormal state “2” in step S2506.

Next, FIG. 28 shows the flow of processing of the full-text search program after receiving a search request from the search thread.
The full-text search program searches for a record satisfying the received search request from the search data (FIG. 17) on the program storage device in step S2802 if there is a record in the allocation table (FIG. 15) in step S2801, If the allocation table is empty in step S2801, search is performed from the update file in step S2803.
Next, the record NO of the record searched in step S2804 is set as a hit record NO (FIG. 26), and the requested search thread is answered.
If an abnormality is detected during the search process, an abnormal end message is returned to the requested search thread.

FIG. 29 shows the flow of processing of the time monitoring program.
The time monitoring program starts processing at regular time intervals.
After the processing is started, the following processing is performed for all the search result tables (FIG. 27).

  First, in step S2901, if there is a search server whose processing result flag is abnormal termination “2”, the director program is notified of abnormal termination as a processing result notification in step S2906.

If the all process result flag is “1” in step S2902, the normal termination is notified to the director program as a process result notification in step S2904.
If there is a search server for which the processing result flag has not been input, that is, the processing has not been completed, in step S2903, if a certain time has elapsed from the earliest time among the processing completion times of the search servers other than the update file manager, After the processing result flag is not input in S2905 and the processing result flag in the search result table (FIG. 27) of the search server other than the update file manager is abnormally ended “2”, the director program is notified as a processing result in step S2906. Notification of abnormal termination.

Returning to FIG.
After the search thread is activated, the director program receives a processing result notification from the time monitoring program in step S2204.

If the content of the processing result notification is normal termination, the next processing is performed in step S2205.
For the search server whose update file charge flag is OFF, that is, “0”, the head address and the record length of the record corresponding to the record NO described in the hit record NO are extracted from the response table (FIG. 14). Read the contents of the corresponding record from the search target file.
For the search server whose update file handling flag is ON, that is, “1”, the start address and the record length of the record corresponding to the record NO described in the hit record NO are extracted from the update file response table (FIG. 20). Then, the contents of the corresponding record are read from the update file.
However, if there is data of the same record NO in the response table (FIG. 14) and the update file response table (FIG. 20), the data of record NO in the response table (FIG. 14) is ignored.
Finally, the contents of the record read in step S2206 are edited into a predetermined format and transmitted to the terminal, and the search request data of the terminal in the search request buffer is deleted.

  If the content of the processing result notification is abnormal end in step S2204, an area reallocation process is performed.

FIG. 30 shows the flow of area reallocation processing of the director program.
In the area reallocation process, first, all search threads are stopped in step S3001.
Next, in step S3002, for the search server in which the processing result flag in the search result table (FIG. 27) indicates an abnormality “2”, all the state flags in the environment table (FIG. 31) are set to the unusable state “1”. Clear the search result table.

In step S3003, the update file data is merged with the search target file.
At this time, if data of the same record NO exists in the update file and the search target file, the data of the search target file is replaced with the data of the update file.
After the merge process, all data in the update file is deleted.

  Thereafter, in step S3004, the response table (FIG. 14) is re-created.

Next, in step S3005, using the response table (FIG. 14) and the environment table (FIG. 31), as shown in FIG. 32, the search responsibility range of each search server is reassigned.
For example, as shown in FIG. 31, when it is found that the search server d is unusable, as shown in FIG. 32, the search servers a, b, excluding the search server e in charge of the update file, The search target files assigned to the three of d are reassigned to the two search servers a and b.

  In step S3006, the allocation table of the re-allocation result (the format is shown in FIG. 33) is notified to all search servers.

  The search server initialization program receives the notification and performs the same processing (FIG. 16) as the pre-processing.

  After the above processing, the director program retrieves the search request in the search request buffer and restarts the search processing (FIG. 22).

  This is the end of the description of the first embodiment.

A second embodiment will be described.
The first embodiment is a case where there is one reception server and search target file.
In the second embodiment, a case will be described in which a plurality of receiving servers and search target files exist as a set.

  The second embodiment is effective when a search requester wants to make a cross-sectional search request for a plurality of systems according to the first embodiment that are distributed in regions.

FIG. 34 shows an overall configuration diagram of the second embodiment.
When there are a plurality of search target files, the same system as in the first embodiment is created corresponding to each search target file.
As a result, a plurality of reception servers are created.
One general reception server is provided between the reception server and the terminal.
Each terminal makes a search request to the general reception server and receives a search result from the general reception server.

  In such a case, the process of updating and maintaining the search target file is performed for each receiving server of each management unit. Therefore, a data update request from a terminal accessed via the general receiving server is not permitted. In general, it is assumed that the general reception server does not receive a data update request from the terminal.

Since everything except the general reception server is the same as the first embodiment, only the general reception server will be described in detail.
The configuration of the general reception server is as shown in FIG.
The general reception server includes an information processing device, a magnetic file device, and a program storage device.

  The components of the program storage device are a control program, a director program, an environment table G, a search result table G, and a reception thread.

FIG. 36 shows the relationship between the basic functions and the table of the second embodiment.
The processing flow of the general reception server will be described in detail with reference to FIG.
First, after the control program initializes the system, the director program is started.

After being started, the director program creates an environment table G (FIG. 37), and then waits for a search request.
The environment table G is a list of reception servers, and includes reception server names, addresses, and status flags.
The status flag indicates the status of the receiving server.
For example, if the reception server is not usable, “1” is set, and if it is usable, “0” is set.
These pieces of information are recorded in advance in the magnetic file device, and the director program obtains the information therefrom.

FIG. 38 shows the flow of processing of the director program after receiving the search request.
After receiving the search request, the director program creates a search result table G (FIG. 39) based on the environment table in step S3801.
The search result table G is a list of accepting servers in a usable state, and has a processing result flag column and a search result column.

Next, the director program activates reception threads corresponding to all reception servers that are ready for use in step S3802.
Simultaneously with the activation, the reception thread is notified of the search request, the address of the reception server in charge, and the column in which the answer is written in the search result table.
As many reception threads as the number of reception servers that can be used are created for each search request.

  After starting the reception thread, the director program waits for a search result notification in step S3803, and the subsequent processing will be described later.

  FIG. 40 shows the processing flow of the reception thread.

  After the reception thread is activated, it receives a search request in step S4001, and transmits it to the reception server in charge in step S4002.

  When the search result (FIG. 41) is received from the reception server in step S4003, in step S4004 it is written in the search result column of the search result table G (FIG. 42) and at the same time the processing result flag is in a normal end state (for example, “1”). ).

  In step S4003, if there is no response from the reception server within a certain time, it is determined that there is an abnormality in the reception server in step S4005, and the processing result flag column is set to an abnormal state (for example, “2”).

Returning to FIG.
The director program manages the waiting for the completion of processing of all the receiving threads for each search request after the receiving thread is activated, and checks the processing result flag of the search result table G (FIG. 42) when the processing is completed in step S3804. .
As a result, in step S3806 and step S3807, the search result of the reception server in which all the search results are normal is edited into a predetermined format and transmitted to the terminal.

  As a result of checking the processing result flag, if any one of the processing result flags indicates an abnormality, the status flag of the corresponding receiving server in the environment table G (FIG. 43) is disabled in step S3805.

Subsequent search requests received from the terminal are sent only to available reception servers.
This is the end of the description of the second embodiment.

Here, the handling of the record number will be supplementarily described.
In the above embodiment, the record NO is held as the record structure in the search target character string data. However, since the search condition is performed by comparing the character strings, there is no need to have a record number. This is an example for explaining the arrangement of records and their division.

  Actually, when the record NO is allocated to each search processing device as the search charge range of the search target character string data, when the reception server creates the response table, the character strings are read sequentially from the beginning of the search target character string data. Therefore, if the head of the record is known, the record number is sequentially assigned and recorded in the table, and the position on the file (search target character string data) and the record length can be recorded.

  Further, as described above, since the reception server holds the response table, it is possible to easily perform the allocation and redistribution of the assigned area to each search processing device using the continuous record NO. ing.

  Further, in the description of the above embodiment, addition and update are taken up as maintenance work of search target character string data being searched, but an update file response table is not shown in FIG. It is possible to provide a deletion flag so that it can be used, and not only the additional update but also the search target character string data can be deleted during the search process.

  In addition, the update data and the data updated by the determination of the record of the same record NO found in each area of the search target character string data to be shared and searched by each search processing device are prioritized. Alternatively, the update flag may be recorded in a table such as a response table.

  This may also be applied to a method of reflecting the search target character string data such as addition, update, and deletion by the update result reflecting means 5.

The search servers in charge of each search server have conventionally been practically searched on a file in practice. However, in the embodiment of the present invention, the data in the assigned area is used. Is expanded on the memory of each search server to search for a character string.
However, with respect to the update file, when the number of records stored in the update file is small, the search on the file may be performed as usual, and in the above-described embodiment, development on the memory is not performed.
If the update frequency is high, the search data may be developed in the program storage device and the character string may be searched even for the search server in charge of the update file, like the other search servers.

In the embodiment of the present invention, each search processing device is a server and is connected to a reception server via a network. However, a device that accepts a search and a device that performs a search are connected to other high-speed links or the like. The same effect can be expected even if the method is used.
(Supplementary note 1) The location of search data in charge and the character string search condition are instructed.
It has a plurality of search processing devices that respond to search results, divides search target character string data into character string record groups, assigns them to one or more search processing devices, and uses each given character string search condition as a search instruction. A full-text search system provided with search integration means for transmitting to the search processing device and receiving and integrating search results from each search processing device,
Additional update temporary storage means for temporarily storing only the record to be added or updated for the search target character string data;
An additional update record search instruction means for instructing any one of the predetermined search processing devices as a search charge data character string record group stored in the additional update temporary storage means;
A full-text search system comprising:
(Appendix 2) Update result reflecting means for deleting a pre-update record corresponding to the additional update record stored in the additional update temporary storage means from the search target character string data and incorporating the additional update record in the search target character string data The full-text search system according to supplementary note 1, comprising:
(Supplementary note 3) After the search integration means transmits a search instruction to the plurality of search processing devices,
Search result reception time storage means for storing the time when the search result is received from the search processing device;
The search result can be received within a preset time from the first received search result reception time stored by the search result reception time storage means in the search processing device assigned by dividing the search target character string data. Failure search processing device determination means for determining a search processing device that has not been determined as a failure device;
The full-text search system according to either Supplementary Note 1 or Supplementary Note 2, wherein
(Supplementary Note 4) When the failure search processing device determination unit determines a failure of the search processing device, the search integration unit discards all the search results transmitted from the plurality of search processing devices and uses the update result reflection unit. After the character string record group stored in the additional update temporary storage means is incorporated into the search target character string data, the search processing device determined to be the failure and the search device instructed to search by the additional update record search instruction means The full-text search system according to any one of Supplementary Note 2 and Supplementary Note 3, wherein the search target character string data is divided and assigned to the search processing devices that can be used except for and a search instruction is issued.
(Supplementary note 5) When the location of search data in charge and the character string search condition are instructed,
It has a plurality of search processing devices that respond to search results, divides search target character string data into character string record groups, assigns them to one or more search processing devices, and uses each given character string search condition as a search instruction. In a full-text search system provided with search integration means for transmitting to the search processing device and receiving and integrating search results from the search processing devices,
Additional update temporary storage means for temporarily storing only the record to be added or updated for the search target character string data,
Additional update record search instruction means for instructing any one of the predetermined search processing devices as a search charge data character string record group stored in the additional update temporary storage means;
A full-text search program characterized by functioning as
(Appendix 6) When the location of the search charge data in charge and the character string search condition are instructed,
It has a plurality of search processing devices that respond to search results, divides search target character string data into character string record groups, assigns them to one or more search processing devices, and uses each given character string search condition as a search instruction. In a full-text search system provided with search integration means for transmitting to the search processing device and receiving and integrating search results from the search processing devices,
Additional update temporary storage means for temporarily storing only the record to be added or updated for the search target character string data,
Additional update record search instruction means for instructing any one of the predetermined search processing devices as a search charge data character string record group stored in the additional update temporary storage means;
A computer-readable recording medium storing a program that functions as a computer.
(Supplementary note 7) The location of the data in charge of search and the character string search condition are instructed.
Connected to multiple search processing servers that respond to search results via a network, search target character string data is divided into character string record groups, assigned to one or more search processing servers, and given character string search conditions are instructed A search integration server that transmits to each of the search processing servers and receives and integrates the search results from each of the search processing servers,
Additional update temporary storage means for temporarily storing only the record to be added or updated for the search target character string data;
Additional update record search instruction means for instructing one of the predetermined search processing servers as a character string record group stored in the additional update temporary storage means as search charge data;
A search integration server comprising:
(Supplementary note 8) When the location of search data in charge and the character string search condition are instructed,
It has a plurality of search processing devices that respond to search results, divides search target character string data into character string record groups, assigns them to one or more search processing devices, and performs each search using given character string search conditions as search instructions In a full-text search system provided with a search integration means for transmitting to a processing device and receiving and integrating search results from each search processing device, when adding or updating a record of the search target character string data,
Temporarily storing only the records to be updated,
A step of instructing the stored character string record group to be additionally updated as the search charge data in any one of the predetermined search processing devices;
A method for adding and updating records during a full-text search.

DESCRIPTION OF SYMBOLS 1 Search object character string data 2 Search integration means 3 Additional update temporary storage means 4 Search processing apparatus 5 Update result reflection means 6 Search result reception time storage means 7 Fault search processing apparatus determination means 8 Additional update record search instruction means

Claims (4)

An allocation system including a storage device that stores a plurality of records, a plurality of search devices, and an allocation device,
The assigning device is
An assigning unit that divides the plurality of records according to the number of the plurality of search devices, and assigns each record group obtained by the division to each of the plurality of search devices;
Each of the plurality of search devices
The search request is received, the record group assigned by the assigning unit is read from the storage device, and it is determined whether or not there is a search target related to the search request for each record included in the read record group. Search means for transmitting the result of
The assigning device further comprises:
When the determination result from at least one of the plurality of search devices cannot be received, the plurality of records are obtained by dividing the plurality of records according to the number of search devices that have received the determination result. Reallocating means for allocating each recorded group as a target record group to be read from the storage device to each of the search devices that have received the determination result,
An allocation system characterized by that. The assigning means is
Transmitting information indicating an address storing a record to be assigned to each search processing device to each of the plurality of search processing devices, and assigning each record group to each of the plurality of storage devices;
The allocation system according to claim 1, wherein: An allocation device that allocates a plurality of records stored in a storage device to a plurality of search devices,
An assigning unit that divides the plurality of records according to the number of the plurality of search devices, and assigns each record group obtained by the division to each of the plurality of search devices;
For each record included in the record group assigned by the assigning unit, from at least one of the plurality of search devices that determines whether or not there is a search target related to the received search request, from the search device Retrieval device capable of obtaining the determination result for each record group obtained by dividing the plurality of records according to the number of retrieval devices that have obtained the determination result when the determination result cannot be obtained Reassignment means for assigning each as a record group to be read from the storage device ,
An allocation apparatus characterized by the above. In a computer that can communicate with multiple search processing devices,
Dividing a plurality of records stored in a storage device according to the number of the plurality of search devices, assigning each record group obtained by the division to each of the plurality of search devices,
For each record included in the record group assigned by the assigning unit, the determination is made from any one of the plurality of search devices that determines whether there is a search target related to the received search request. Each of the record groups obtained by dividing the plurality of records according to the number of search devices that have acquired the determination result, the search device that has acquired the determination result. Assign each as a record group to be read from the storage device ,
An allocation program characterized by causing

Images