CN100428226C - A method for implementing memory-like database access and retrieval - Google Patents

Abstract

The method for realizing memory-like database access and retrieval described in the present invention is applied to an embedded system, and the method establishes a storage table of corresponding data nodes according to the requirement of data retrieval. A corresponding index table is established according to the above storage table, and the index table is a dynamically sortable data table structure established based on a balanced binary tree model. The sorting field of the established index table is stored in priority, scans and analyzes the specified previous queries, sorts the priority of attribute queries, and determines the sorting rules of the index, so that the query of data is more efficient. Compared with the existing in-memory database retrieval method, the access and retrieval method can instantly update the statistical query hit rate, and has relatively high database management and security performance, so it can improve efficiency and reduce the occupation of internal and external storage resources.

Description

A method for implementing memory-like database access and retrieval

technical field

The invention relates to a method for accessing and retrieving, in particular to realizing fast data accessing and retrieving under the premise of low resource occupation for memory-like databases.

Background technique

With the rapid development of computer and network information technology, the use of computers and networks has become more and more inseparable from existing office and life. Various large amounts of information and data are run through the network or stand-alone, realizing data sharing and remote transmission. While managing the data in the database, methods and software for effectively classifying and accessing data and implementing advanced retrieval are also produced and applied accordingly.

At present, the data management methods that are widely used include hierarchical, network and relational databases. Although the problems of data access and retrieval can be solved to a certain extent, there are still certain defects in the establishment and application of such commercial databases. . For example, the cost of database establishment and use is relatively high, which is too expensive for small data management units; in addition, existing databases such as MS SQL, DBII, and Oracle are all based on hard disk or tape operations. It cannot meet the requirements of real-time response, fast access and retrieval. Now there is also a mode of data access and retrieval using an in-memory database. Although the retrieval speed is fast, it also takes up a large amount of external storage and internal memory resources, especially for embedded systems. A serious waste of resources.

As mentioned above, the various existing database access and retrieval methods have obvious shortcomings and deficiencies for small data application units, and there is no corresponding solution in the disclosed technology.

Contents of the invention

The method for realizing memory-like database access and retrieval of the present invention aims to solve the above-mentioned problems and deficiencies and is designed with a data access and retrieval method applied to embedded system databases.

In order to solve the above-mentioned technical problems, the present invention adopts the following technical solutions to achieve:

A method for realizing memory-like database access and retrieval, comprising the following steps: a memory-like database initialization step and a memory-like database retrieval and access step, wherein the memory-like database initialization step further includes the following steps:

Read and configure the latest m hit rate table. m is the total number of queries that need to be executed in a query hit rate calculation cycle. The following m is explained in the same way;

Establish a data node storage table based on the Hash table type structure for data content and type;

Establish an index table based on a balanced binary tree model, and a priority storage table for sorting fields of the index table;

Calling the storage function in the memory-like database retrieval and access steps, establishing all tables in the database, and simultaneously sorting the index tables according to the priority storage table of the index table sorting field;

The retrieval and access steps of the class memory database further include the following steps:

It is necessary to analyze m query records. From the m hit rate table, select the one with the highest hit rate as the next m value. If the hit rates are the same, select randomly;

Each time after receiving a query command, first add 1 to the total number of queries, then determine the total number of fields involved, and then assign the weight of this time to the index table sorting field priority according to the existing priority data from large to small The storage range of the storage table; when i<=m, i is the total number of queries after this query is included in the query hit rate calculation cycle, and the value of i is within the range of natural numbers. Simply combine the previous priority data and The priority data of this time can be obtained by summing the weights this time; when i>m, clear the sorting field of the index table to store the table data first and set i to 0;

Sort the index table according to the optimal field, skip this step if it is the same as last time, and add 1 to the value in the priority storage table of the index table sorting field; then decide how to get it according to whether there is an optimal field in the query condition The required data node key, if any, obtain the required data node key from the index table based on the balanced binary tree model; if not, obtain the required data node key from the data node storage table based on the Hash table type structure words, and finally read single records one by one. There is no keyword list and only all the data in the data node storage table can be traversed. The exact position can be obtained after the function conversion, without checking one by one;

When the total number of queries is greater than or equal to m, the hit rate is calculated, then fill in the corresponding position and set the total query times to 0, and then select the one with the highest hit rate as the next m value, if there are many in the m hit rate table is the maximum at one position at the same time, then randomly select one as the next m value;

Re-establish the index table based on the balanced binary tree model;

Clear the index table sorting field to store the table first;

When the service is stopped, the m hit rate table is stored as a configuration file in the external memory as the initial value for the next service operation.

The data node storage table can locate the storage location of data at one time through the primary key.

As mentioned above, in the method for realizing memory-like database access and retrieval described in the present invention, data storage is carried out according to the third normal form of relational database, which can effectively prevent redundancy and abnormality. Moreover, compared with the existing memory database retrieval method, the statistical query hit rate can be updated in real time, and has relatively high database management and security performance, so it can improve efficiency and reduce internal and external memory resource occupation.

Description of drawings

Fig. 1 is described data node storage table;

Fig. 2 is the balanced binary tree model figure of described building index table;

Fig. 3 is described sorting field priority storage table;

Figure 4 is the latest hit rate table after retrieving several records;

Fig. 5 is the initialization flowchart of class memory database;

Fig. 6 is a schematic diagram of the end stage of the memory-like database;

Fig. 7 is described configuration file format table;

Fig. 8 is a flow chart of querying and accessing the memory-like database.

Detailed ways

As shown in Figures 1 to 8, the method for realizing memory-like database access and retrieval according to the present invention establishes a storage table of corresponding data nodes for data content and type, so as to directly access target data according to the primary key. According to the storage table, the corresponding index table based on the balanced binary tree model is established. Create an index table sorting field priority storage table based on the above index table, scan and analyze the specified prior queries, sort the priority of attribute queries, and determine the sorting rules of the index.

As shown in Figure 1, the described storage table is based on the Hash table type structure, and the storage location of the data can be located once by the primary key (bold), so that fast data access can be realized by using the primary key.

According to the data content and type in the storage table, in order to effectively avoid redundancy and abnormality in the entire database, a corresponding query table for fast query and data location is usually established according to the third normal form mode in the relational database.

As shown in Figure 2, a balanced binary tree model of the index table is established based on the above storage table, so that each node is only stored corresponding to the primary key of the index table, and the optimal field sorting has been performed in advance. Therefore, it can ensure that the query complexity is minimized when the index is hit.

As shown in Figure 3, the index table sorting field priority storage table includes the priority data of all attributes of the storage table, and the last row of data in the table is the sum of m rows of data above its column.

As shown in FIG. 4 , after analyzing the m query records in the priority storage table in the sorting field of the above index table, the latest hit rate is stored in the hit rate table shown.

As shown in Figure 5, the initialization process of the memory-like database is:

Read and configure the latest m hit rate table (as shown in Figure 4);

Set up described index table and balanced binary tree model (as Fig. 2), and index table sorting field priority storage table (as Fig. 3);

Call the storage function in the query and access flow of the class memory database (as shown in Figure 8), set up all tables in the database (as shown in Figure 1), and further improve the index table (as shown in Figure 2) simultaneously.

As shown in FIG. 6 , it is a schematic diagram of the end stage of the memory-like database.

As shown in Figure 7, it is the configuration file format table.

As shown in Figure 8, the query and access process of the class memory database is:

Need to analyze m query records, select the highest hit rate from the above Figure 4 as the next m value, if the hit rates are the same, randomly select;

Each time after receiving a query instruction, first add 1 to the total number of queries, and then determine that the number of fields involved is N0. Value, then according to the existing priority data from large to small, the weight of this time (N0...1) Assigned to the current storage range; when i<=m, simply sum the previous priority data and the current weight to get the current priority data; when i>m, clear the data in this table and place i is 0.

Sort the index table according to the optimal field, skip this step if it is the same as last time, and add 1 to the value in the sorting field priority storage table as shown in Figure 3; then according to the optimal field (if any) ) to perform the first screening, so as to finally obtain the required data node keywords, and finally read single records one by one (there is no keyword list, only all the data in the table can be traversed). When performing single-record access, you only need to directly check the storage table shown in Figure 1, which can ensure that the exact location can be obtained after a limited number of function conversions, without checking one by one;

When the total number of queries is greater than or equal to m, the hit rate is calculated, then fill in the corresponding position and set the total query times to 0, and then select the one with the highest hit rate for the next m value (if there are more than If a position is the largest at the same time, then randomly select one as the next m value);

Re-establish the storage table index table as shown in Figure 2;

Clear the index table sorting field priority storage table as shown in Figure 3;

When the service is stopped, the hit rate table is stored as a configuration file (as shown in Figure 4) in the external memory as the initial value for the next service operation.

Claims (2)

1. A method for realizing class memory database access and retrieval, comprising the steps of: class memory database initialization step and class memory database retrieval and access step, wherein, the class memory database initialization step further comprises the following steps: Read and configure the latest m hit rate table. m is the total number of queries that need to be executed in a query hit rate calculation cycle. The following m is explained in the same way; Establish a data node storage table based on the Hash table type structure for data content and type; Establish an index table based on a balanced binary tree model, and a priority storage table for sorting fields of the index table; Calling the storage function in the memory-like database retrieval and access steps, establishing all tables in the database, and simultaneously sorting the index tables according to the priority storage table of the index table sorting field; The retrieval and access steps of the class memory database further include the following steps: It is necessary to analyze m query records. From the m hit rate table, select the one with the highest hit rate as the next m value. If the hit rates are the same, select randomly; Each time after receiving a query command, first add 1 to the total number of queries, then determine the total number of fields involved, and then assign the weight of this time to the index table sorting field priority according to the existing priority data from large to small The storage range of the storage table; when i<=m, i is the total number of queries after this query is included in the query hit rate calculation cycle, and the value of i is within the range of natural numbers. Simply combine the previous priority data and The priority data of this time can be obtained by summing the weights this time; when i>m, clear the sorting field of the index table to store the table data first and set i to 0; Sort the index table according to the optimal field, skip this step if it is the same as last time, and add 1 to the value in the priority storage table of the index table sorting field; then decide how to get it according to whether there is an optimal field in the query condition The required data node key, if there is, obtain the required data node key from the index table based on the balanced binary tree model; if not, obtain the required data node key from the data node storage table based on the Hash table type structure words, and finally read single records one by one. There is no keyword list and only all the data in the data node storage table can be traversed. The exact position can be obtained after the function conversion, without checking one by one; When the total number of queries is greater than or equal to m, the hit rate is calculated, then fill in the corresponding position in the m hit rate table and set the total number of queries to 0, and then select the highest hit rate from the m hit rate table as the next m value, if more than one position in the m hit rate table is the maximum at the same time, randomly select one as the next m value; Re-establish the index table based on the balanced binary tree model; Clear the index table sorting field to store the table first; When the service is stopped, the m hit rate table is stored as a configuration file in the external memory as the initial value for the next service operation. 2. The method for realizing memory-like database access and retrieval according to claim 1, characterized in that: the data node storage table can locate the storage location of data at one time through the primary key.

Images