KR20000018923A - Memory managing method and a device for information searching system for plurality of users - Google Patents

Abstract

PURPOSE: A memory managing method and the device is provided to enhance the speed of applied program and memory usage by using the memory in the operative usage while operating as client/server and to enhance the speedy information search. CONSTITUTION: The device comprises a search client(100) which provides a user searching interface and which requests search at the server; a search engine(220) which searches the index data stored at the hard disk(400) and carries out the search result; and a memory manager(210) which manages and stores usable memory using the memory pool index (311) and the data type(312) according to the request of memory search engine. The search engine(220) detects the necessary data by searching the index data existing in the hard disk(400).

Description

Memory management method and apparatus for multi-user information retrieval system

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a memory management method and apparatus for excluding dynamic memory allocation from an operating system (OS) in a multi-user information retrieval system in the form of a client / server.

The conventional information retrieval system uses a method of dynamically allocating memory for each intermediate search result in order to maintain the intermediate search result.

On the other hand, "Retrieval system KRISTAL-Ⅱ [author Cho Hyun-yang et al. 6, pp. 47-69, 1997.10.16]" published in a technical paper conference as a preceding paper, the storage and storage of unstructured text documents of variable length Kernel that manages object using blob for development of retrieval system through providing approach, storage engine that manages document and index data, file catalog and database document that store large database in several volumes A catalog manager, consisting of a section catalog, which combines section catalogs to maintain information of sections, a database manager that creates database schema information created by the system administrator and writes it to the system catalog, a user created by the Boolean model using the catalog manager and storage engine Documents that satisfy the query It is possible to save / index / search for large amounts of data by using a database under the search system. However, since there is no mention of memory management, it is possible to improve the search speed through the memory manager. There is a problem in the efficient management of memory.

Also, as a prior patent, the US patent "Multiprocessor memory management method [right holder Bishop; Thomas P. et al. 4, registration number 941703, 1989.3.7]" has its own memory that is not logically divided into two parts and shared. In a multiprocessor system with multiple processors, each processor has its own memory for a memory allocation method that effectively allocates memory to the processors that make each processor memory request. Requiring memory from a processor for execution can effectively handle the memory management requirements of processes in a multiprocessor environment by satisfying the memory requirements of two logically unshared processes. Menu in the system Lee managed to look forward to effective follow this difficult problem.

In general, the method of allocating memory uses a system call, which requires a lot of time.

Therefore, this method reduces the search speed in proportion to the number of memory allocations, which is useful for small search cells, but not suitable for large search cells.

In order to solve the above problems, an object of the present invention is to improve the search speed when searching a large data rule using a memory manager in a multi-user information retrieval system, and to efficiently use the memory in the retrieval process. .

In the present invention, when a server is initially performed in a search system in the form of a client / server, a memory pool is created by allocating memory for processing a search result.

In order to perform a search, a user generates a query to find information using a search client, and the generated query is transmitted to a server.

The server retrieves a set of appropriate documents from the index information for each keyword included in the query.

At this time, if a search result for each keyword exists, it is necessary to store the search result. In order to secure the storage space, the searcher requests a memory required for the memory manager.

The memory manager responds to the searcher's memory request through its algorithm.

Since memory requests are issued by the client simultaneously by multiple users, the memory manager uses a locking mechanism to prevent more than one user from using the same memory.

The used memory is returned to the memory manager so that it can be used again when another user's search request is made.

The memory pool managed by the memory manager is returned to the system in batches when the server shuts down.

1 is a block diagram of a memory management apparatus to which the present invention is applied;

2 is a block diagram illustrating a memory manager according to the present invention;

3 is a flowchart illustrating a memory management method according to the present invention;

* Explanation of symbols for main parts of the drawings

100: search client 200: search server

210: memory manager 211: concurrency control module

212: memory manager control module 213: overflow processing module

214: variable memory pool management module 215: memory deallocation module

216: data type module 217: memory management pool index management module

220: search engine 300: main memory

310: memory pool 311: memory pool index

312: Data Type 400: Hard Disk

410 index data

In order to achieve the above object, the present invention provides a plurality of search clients that perform index combination / weighting / ranking of extracted data by searching index data stored on a hard disk according to a search model required for a search request. Search engine, a search server comprising: a memory manager for finding available memory in main memory having a memory pool having a data type requested by the search engine and a memory pool index, and delivering the available memory to the search engine It is done.

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

1 is a block diagram illustrating a memory management apparatus to which the present invention is applied. The memory management apparatus proposed by the present invention provides a user search interface for providing a fast search speed in a multi-user information retrieval system composed of a search client / server. A search client 100 requesting a search to a server, a search engine 220 for searching the index data 410 stored in the hard disk 400 according to a search model, and performing combination / weighting / ranking of the results and a main memory A memory manager 210 that provides and manages available memory using a data type 312 for memory requirements of a search engine and a memory pool index 311 in a memory pool 310 located at It consists of a search server 200 that performs a search for the request.

When a search request of the search client 100 occurs in the search server 200, the search server executes one search engine.

The search engine 220 searches the index data 410 present in the hard disk 400 for each keyword included in the search query to extract the required data in order to perform the requested search.

In order to store the extracted data, the search engine 220 requests a memory required by the memory manager 210, and the memory manager 210 searches for available memory existing in the memory pool 310 and delivers it to the search engine 220. do.

The search engine 220 continues to request memory from the memory manager 210 to process the search results of each keyword.

When the search is performed for all the keywords, the search engine 220 combines and weights the searched results and delivers the ranking results to the search client 100 requesting the search.

In this case, since a plurality of search clients 100 generate a search request to one search server 200, the search server 200 executes N search engines 220, and each search engine 220 has one memory manager. Through 210, a required memory is obtained to perform a search.

2 is a block diagram of a memory manager according to the present invention, in which the memory manager 210 includes a concurrency control module 211, a memory manager control module 212, an overflow processing module 213, and a variable memory. The pool management module 214, the memory allocation / release module 215, the data type management module 216, and the memory pool index management module 217 are configured.

The concurrency control module 211 performs a concurrency control function to ensure the integrity of the memory existing in the memory pool 310. When a request for a plurality of memories occurs simultaneously, the concurrency control module 211 If the seriality of the operation that accesses the existing memory is not guaranteed, a defect occurs in the contents of the memory pool 310.

In this case, the operation of approaching the memory pool 310 to guarantee the directivity means an operation that guarantees that no interruption or interruption occurs between the start and end of each operation.

Therefore, concurrency control is performed by using semaphores provided by the operating system of UNIX.

The memory manager control module 212 is a module for controlling a request generated from the search engine 220 to the memory manager 210.

There are four major things happening in the memory manager 210: first, the memory request in the memory pool, second, the return of the used memory to the memory pool, third, the initialization of the memory manager, and fourth, the execution of the memory manager.

The overflow processing module 213 is a module that processes when there is no more memory available in the memory pool.

When a memory request comes in when there is no more memory available, the overflow processing module 213 checks whether a new memory can be allocated.

If the memory manager 210 is configured to use only one block initially set, the memory manager 210 sends a message to the search engine 220 requesting the memory that there is no more memory available.

However, if the memory manager 210 is configured to use multiple memory blocks, the variable memory pool management module 214 is called to allocate new M memory blocks from the system.

The variable memory pool management module 214 functions to cope with new memory requests in situations where no more available memory is in the pool.

That is, the memory manager 210 operates normally by dynamically allocating memory from the system to perform flexible memory management.

The use of the variable memory pool management module 214 enables efficient memory management by eliminating the need to initialize and use much memory when the memory manager first starts.

The memory allocation / release module 215 generates a memory pool 310 by allocating memory from the system and performs a function of returning the allocated memory from the memory pool 310 to the system at the end of the memory manager.

Memory allocation occurs at the start of the memory manager 210 and at the memory request when no more memory is available in the memory pool 310 and memory release occurs at the end of the memory manager 210.

The data type management module 216 is a module that manages data types stored in a memory handled by the memory manager 210, and allows the memory manager 210 to respond to memory requests for any type of data. Do this.

The memory pool index management module 217 manages a memory pool index 311 indicating a state of use of memories existing in the memory pool 310.

Since the memory pool index 311 must be referred to at the time of a memory request or return, the memory pool index 311 should be searchable at a high speed.

Accordingly, the memory pool index 311 and additionally, a pointer indicating the memory available in the memory pool 310 may be maintained to provide a memory that is readily available upon request.

In addition to the request for a single data type, it also provides the function of providing multiple data of the same type at once.

3 is a flowchart illustrating a memory management method according to the present invention, in which the memory manager 210 requires a large number of requests for memory in the memory pool 311, returning used memory to the memory pool 311, and a memory manager ( There are four phases of initialization of 210 and termination of execution of memory manager 210.

First, it is determined whether a memory request occurs (S1), and if so, first, concurrency processing for the memory pool is started (S2).

Then, it checks what type of memory is requested (S3) and determines whether there is an available memory for that type in the memory pool (S4).

After the determination, if there is available memory in the memory pool, one memory is taken out (S5) and the memory pool index that maintains the state of the memory pool is changed (S6) to match the state of the index and the memory pool, and then a new request is made. The concurrency process ends so as to enable the pool (S7).

Therefore, the search engine requesting the memory obtains the required memory (S8) and continues to the next task.

If the available memory does not exist in the memory pool, it is determined whether a condition for allocating a new memory is satisfied (S9).

The new memory allocation condition is whether the number of currently allocated memory blocks exceeds the maximum number of allocated blocks.

At this time, depending on the application, it is possible to allocate a block of memory only once.

However, variable memory pool management is required when allocating multiple memory blocks.

If the new memory allocation condition is not satisfied, an error message is output (S10) and the memory manager is terminated. If satisfied, new M memory blocks are allocated from the system (S11).

In this case, available memory exists in the memory pool and the required memory is obtained (S12).

After that, the memory pool index that maintains the state of the memory pool is changed (S6) to match the state of the index and the memory pool, and then the concurrency processing is terminated so that a new request is possible for the memory pool (S7). Pass the memory required for (S8).

Secondly, if it does not occur after the memory request occurrence determination (S1), it is determined whether a memory return request occurs (S13). If so, the concurrency process is started (S14), and the memory to be returned is returned to the memory pool (S15). After changing the memory pool index maintaining the state of the pool (S16) to match the state of the index and the memory pool, the concurrency process is terminated so that a new request is possible for the memory pool (S17).

The returned memory can then be reused at the request of a search engine.

Third, if the memory return request does not occur after the determination (S13), it is determined whether a memory manager initialization request occurs (S18). When a memory manager initialization request occurs, N memory blocks are allocated from the system to generate a memory pool (S18). In operation S19, a memory pool index indicating a state of the memory pool is also generated (S20), and then the memory pool index is initialized as in the state of the memory pool (S21).

Fourth, when the memory manager termination request occurs because the initialization request does not occur, all memory in the memory pool is returned to the system (S22), the memory index is deleted (S23), and the memory manager is terminated.

As described above, the memory management apparatus and method for a multi-user information retrieval system of the present invention can be expected in a fast retrieval speed in the information retrieval system, and is used for applications that dynamically use a lot of memory while operating as a client / server. Doing so has the effect of improving memory usage and application speed.

Claims (8)

A memory management apparatus of a multi-user information retrieval system configured as a client / server for managing a memory for storing a search result by executing a plurality of search engines in response to a search request from a plurality of clients, At the start of the search server, the memory is allocated from the system to create a memory pool including memory pool indexes for the data type memory and the usage status of the memory data type in the main memory. By using the memory and memory pool index for each data type to find available memory according to the memory requirements of the search engine to provide to the search engine, and a memory manager for managing the return after use is configured to include in the search server The memory management device of the user information retrieval system. The memory manager of claim 1, wherein the memory manager comprises: Concurrency control function that guarantees the integrity by using the semaphore provided by the Unix operating system for the operation of guaranteeing the seriality without any interruption or termination between the start and the end of the operation accessing the main memory pool of the main memory. A concurrency control module for performing the; A memory manager control module configured to control a request for a memory pool generated from a check engine of the search server to a memory manager under the control of the concurrency control module, to return the used memory to the memory pool, and to initialize and terminate execution of the memory manager; When there is no more memory available in the memory pool and a memory request comes in, if only one block initially set is used, a message is sent to the search engine indicating that there is no more memory available. An overflow processing module that calls a variable memory pool management module and allocates a new memory block from the system when it is set to use the memory; A variable memory pool management module that dynamically allocates memory from the system and allows the memory manager to operate normally to perform flexible memory management; A memory allocating / releasing module configured to receive memory from the system, create a memory pool, and return the allocated memory from the memory pool to the system when the memory manager terminates; A data type module for managing a data type such that the memory manager copes with a memory request for data of a type stored in a memory in the memory manager; It maintains a pointer to the memory pool when the memory request or return and additionally the memory available to the memory pool to provide free memory when the memory request, and to provide multiple indexes of the same type as one data type request at a time A memory management apparatus of a multi-user information retrieval system, characterized by comprising a memory pool index management module that performs a function. In the memory management method of a multi-user information retrieval system composed of a client / server, A memory manager that manages memory for processing search results in the server, the memory manager comprising: Creating a memory pool in main memory by allocating memory at server startup; Making the search engine use memory existing in the memory pool when the memory is requested; The used memory is returned and controlled to be used for storing other search results, and the memory pool is deleted and returned to the system when the server is terminated. Memory management method. The method of claim 3, wherein the memory management method comprises: In response to the memory requests of a plurality of search engines included in the server, one memory manager provides each search engine with necessary memory from the memory pool created in the main memory, and manages return after use. Memory management method for user information retrieval system. The method of claim 3, wherein the memory management method comprises: If requests for multiple memories occur concurrently, In order to guarantee the seriality of operations accessing the memory in the memory pool, the concurrency control can be performed by using semaphores provided by the UNIX operating system so that multiple users can use the memory manager at the same time. Memory management method for a multi-user information retrieval system, characterized in that. The method of claim 3, wherein the memory management method comprises: Allocate a plurality of data type memory to the memory pool, manage the use state of the plurality of data type memory as a memory pool index and maintain a pointer to the available memory, A memory management method for a multi-user information retrieval system, characterized by providing memory through fast memory access using a memory pool indexing method for memory needs for different types of data. The method of claim 3, wherein the memory management method comprises: When no more free memory is in the memory pool, If the number of currently allocated memory blocks does not exceed the maximum number of allocated blocks, the condition for allocating a new memory is checked.If the new memory allocation condition is not satisfied, an error message is output and the memory manager is terminated. If satisfied, new M memory blocks are allocated from the system to expand the memory blocks of the memory pool to perform variable memory pool management that provides the necessary memory to the search engines that require the memory. Memory management for search systems. In a memory management method for a multi-user information retrieval system, When a memory manager initialization request occurs at server startup (117), N memory blocks are allocated from the system to create a memory pool (118), a memory pool index indicating the state of the memory pool is also generated (119), and then a memory pool. Initializing the state and memory pool indexes of When a memory request occurs from the search engine (101), it starts concurrency processing (102) for the memory pool (60), checks (103) what type of memory is requested, and makes available memory suitable for that type of memory. Verifying (104) the pool; If there is available memory in the memory pool, one memory is obtained (105), the memory pool index that maintains the state of the memory pool is changed (106) to match the state of the index and the memory pool, and a new request is made. Terminating (107) the concurrency process to enable for; If there is no available memory in the memory pool, determining (109) a condition for allocating new memory by determining whether the number of currently allocated memory blocks does not exceed the maximum number of allocated blocks; If the new memory allocation condition is not satisfied, an error message is output 110 and the memory manager is terminated. If the new memory allocation condition is satisfied, new memory blocks are allocated from the system (111) and the memory pool is expanded. Then, after changing the memory pool index that maintains the state of the memory pool (106) to match the state of the index and the memory pool, the concurrency process is terminated (108) so that a new request is possible for the memory pool and the search engine that requested the memory. (108) handing over the memory required for When a memory return request occurs 112, the concurrency process starts 113 to return 113 the memory to be returned to the memory pool, and to change 115 the memory pool index that maintains the state of the memory pool to the index and the memory pool. Terminating (116) the concurrency process so that a new request is possible for the memory pool after matching the state of; When a memory manager termination request occurs, all memory in the memory pool is returned to the system (121), the memory index is deleted (122), and memory management is terminated. Memory management for search systems.