TW201828117A - Database operation method and apparatus improving the transaction execution efficiency and increasing the throughput - Google Patents

Abstract

This invention provides a database operation method and apparatus. The method comprises the following steps: sequentially obtaining database operation instructions in a target transaction executed by an application server in a process of executing the target transaction by the application server; performing speculative execution on the database operation instruction, and returning a speculative execution result to the application server for the application server to determine the next database operation instruction to be executed, and recording the database operation instruction locally and recording speculative execution data generated by the speculative execution when the database operation instruction belongs to one of a modified database class instruction and a locked non-modified database class instruction; and when a transaction commit instruction in the target transaction is obtained, controlling a database corresponding to the application server according to the locally recorded database operation instruction and the speculative execution data to actually execute the target transaction. This invention can improve the transaction execution efficiency and increase the throughput.

Description

 Database operation method and device  

The present invention relates to the field of database technology, and in particular, to a database operation method and apparatus.

Database Transaction refers to a series of operations performed as a single logical unit of work, either completely or completely. A transaction consists of a series of database manipulation instructions, such as statements.

In the database operation, in order to ensure the correctness of concurrent reading of data, the concept of transaction isolation level is proposed. In the Standard Structured Query Language (SQL) specification, four transaction isolation levels are defined. Among them, Read Committed is the isolation level that most applications prefer to avoid dirty reads. And it has better concurrency performance, but still encounters the problem of non-repeatable reading.

For the non-repeatable read problem faced by reading the committed isolation level transaction, you can temporarily read the isolation level of a certain data in the submitted isolation level transaction through the for update statement to ensure that the data can be read repeatedly, but this will be seriously reduced. The concurrency of the system results in lower execution efficiency and lower transaction throughput.

Aspects of the present invention provide a database operation method and apparatus for ensuring concurrent performance of a database system, improving transaction execution efficiency, and increasing transaction throughput.

An aspect of the present invention provides a database operation method, including: sequentially acquiring, in an execution process of an application server, a database operation instruction in the target transaction executed by the application server; The database operation instruction performs predictive execution, and returns a predicted execution result to the application server for the application server to determine a next database operation instruction to be executed, and locally records the database operation And predicting, when the database operation instruction belongs to one of a modified database class instruction and a locked non-modified database class instruction, predicting execution data generated by the prediction execution; when acquiring the transaction in the target transaction When the instruction is submitted, the database corresponding to the application server is controlled to actually execute the target transaction according to the database operation instruction and the predicted execution data recorded locally.

In another aspect of the present invention, a method for operating a database is provided, comprising: sequentially acquiring a database operation instruction in the target transaction executed by the application server during a process of executing a target transaction by an application server; Performing predictive execution on the database operation instruction, returning the predicted execution result to the application server for the application server to determine the next database operation instruction to be executed, and locally recording the database The operation instruction and the predicted execution data generated by the prediction execution when the database operation instruction belongs to the locked non-modified database class instruction; when the transaction commit instruction in the target transaction is obtained, according to the local record The database operation instruction and the predicted execution data, and the database corresponding to the application server is controlled to actually execute the target transaction.

According to still another aspect of the present invention, a database operating device is provided, including: an obtaining module, configured to sequentially acquire the target transaction executed by the application server during a process of executing a target transaction by an application server a database operation instruction; a prediction execution module, configured to perform prediction execution on the database operation instruction, and return the prediction execution result to the application server for the application server to determine the next execution required a database operation instruction, and locally recording the database operation instruction and recording the prediction execution when the database operation instruction belongs to one of a modified database class instruction and a locked non-modified database class instruction And a control execution module, configured to: when the transaction commit instruction in the target transaction is obtained, control the application server according to the database operation instruction and the predicted execution data recorded locally The database actually executes the target transaction.

According to still another aspect of the present invention, a database operating device is provided, including: an obtaining module, configured to sequentially acquire the target transaction executed by the application server during a process of executing a target transaction by an application server a database operation instruction; a prediction execution module, configured to perform prediction execution on the database operation instruction, and return the prediction execution result to the application server for the application server to determine the next execution required a database operation instruction, and locally recording the database operation instruction and recording the predicted execution data generated by the prediction execution when the database operation instruction belongs to the locked non-modified database type instruction; controlling the execution module Controlling, when the transaction commit instruction in the target transaction is obtained, controlling the database corresponding to the application server to actually execute the target according to the database operation instruction recorded locally and the predicted execution data. Transaction.

In the present invention, the database operating device cooperates with the application server to join the predictive execution process in the process of executing the target transaction by the application server, and obtains and records all the database operation instructions that need to be executed in the target transaction in advance, and only targets A database operation instruction that is a modified database class instruction and a locked non-modified database class instruction, or a prediction execution data generated only for a database operation instruction record belonging to a locked non-modified database class instruction record In order to adapt to the locking scheme under the submitted isolation level, provide conditions for the real execution of the transaction, and then control the database corresponding to the application server to actually execute the target transaction according to the recorded database operation instruction and the predicted execution data. It is beneficial to improve execution efficiency and thus increase transaction throughput.

10‧‧‧Application Server

20‧‧‧Database

30‧‧‧Database operating device

51‧‧‧Getting module

52‧‧‧Predictive Execution Module

53‧‧‧Control Execution Module

61‧‧‧Get the module

62‧‧‧Predictive execution module

63‧‧‧Control Execution Module

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, a brief description of the drawings used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are some of the present invention. For the embodiments, other drawings may be obtained according to these drawings without any creative labor.

1 is a deployment diagram of a database application system in the prior art; FIG. 2 is a deployment diagram of a database application system according to an embodiment of the present invention; FIG. 3 is a schematic flowchart of a database operation method according to another embodiment of the present invention; FIG. 4 is a schematic flowchart diagram of a database operation method according to another embodiment of the present invention; FIG. 5 is a schematic structural diagram of a database operation device according to another embodiment of the present invention; Schematic diagram of the structure of the database operating device.

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the drawings in the embodiments of the present invention. It is a partial embodiment of the invention, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

FIG. 1 is a deployment diagram of a database application system in the prior art. As shown in FIG. 1, the system includes an application server 10 and a database 20. The application server 10 may be one or multiple. The database 20 may be one or plural. Only one application server 10 and one database 20 are shown by way of example in FIG. When the database 20 needs to be accessed in a transactional manner, the application server 10 executes the transaction logic to sequentially apply the database operation instructions in the transaction to the database 20. Because the database operation instructions that need to be executed in the transaction cannot be known in advance, the transaction execution efficiency is low, especially in the scheme of locking the data through the for update statement under the configuration read commit isolation level, the execution efficiency is lower. , transaction throughput is low.

In view of the deficiencies of the prior art, the present invention provides a new database application system, as shown in FIG. The system adds a database operating device 30 between the application server 10 and the database 20. The database operation device 30 is configured to execute the database operation method provided by the present invention, and implement a new transaction execution logic in the locking scheme under the read isolation level, that is, the prediction execution is performed on the transaction, and the transaction is obtained in advance. The database operation instruction that needs to be executed, realizes the prediction of the execution path, and combines the read and commit this isolation level and locks, only for the non-modified database class instruction belonging to the modified database class instruction and the locked database class instruction. The database operation instruction, or only the database execution instruction record belonging to the locked non-modified database class instruction record, the prediction execution data generated by the prediction execution, and the database operation instruction and the prediction execution based on the prediction execution The actual execution of the transaction on the database 20 is beneficial to improve transaction execution efficiency and increase transaction throughput. Transaction throughput is the number of transactions processed per unit of time.

It should be noted that the database operating device 30 is actually a logical processing device, which can be deployed separately, between the application server 10 and the database 20, or can be deployed on one end of the application server 10, or Can be deployed on the end of the database 20 to achieve.

The following embodiment will specifically describe the method flow of the technical solution of the present invention.

FIG. 3 is a schematic flowchart diagram of a method for operating a database according to another embodiment of the present invention. As shown in FIG. 3, the method includes:

301. In the process of executing the target transaction by the application server, sequentially acquiring the database operation instruction in the target transaction executed by the application server.

302. Perform predictive execution on the database operation instruction, return the predicted execution result to the application server to supply the application server to determine the next database operation instruction to be executed, and record the database operation instruction locally and in the database. The operation instruction is a prediction execution data generated by the prediction execution when the operation instruction belongs to one of the modification database class instruction and the locked non-modified database class instruction.

303. When the transaction commit instruction in the target transaction is obtained, according to the locally recorded database operation instruction and the predicted execution data, the database corresponding to the application server is controlled to actually execute the target transaction.

First, for convenience of description, the present embodiment refers to a transaction that the application server needs to execute as a target transaction. The target transaction mainly includes operation instructions for operating the database, and in addition to the database operation instruction, some control instructions for controlling the execution state of the target transaction, such as a transaction start instruction, a transaction commit instruction, and a transaction return (rollback) ) instructions, etc. These instructions are actually statements written in the database language. Depending on the language of the database, these instructions can be statements written in different languages. For example, if SQL is used, the above database operation instructions and control instructions are actually a series of SQL statements.

In addition, in some application scenarios, the transaction processing needs to select some records first, and then process the records, so it is necessary to add an exclusive lock to the selected records. This means that the database operation instruction in this embodiment may be a locked database operation instruction. For example, suppose you transfer 100 yuan from Bob's account. This operation is completed in three steps. Query Bob's account, determine if the account is 100 yuan, and reduce the amount of money in Bob's account by 100. If there are two transfer operations in parallel, when the first transfer operation finds the balance of Bob's account, it finds that there are 100 pieces. If there is another transfer operation at this time, Bob's account is also checked, and Bob's account is also found. There are 100, so the two transfer operations will be transferred, so the first operation subtracts 100 bucks from Bob's account, and the other operation also subtracts 100 bucks from Bob's account. At this time, Bob's account The balance may be negative. In order to avoid this situation, you need to use for update to lock Bob's account, to ensure that during the first transfer operation, the second transfer operation can not obtain the lock, and therefore can not be queried.

The method provided in this embodiment is applicable to operating a database whose isolation level is read and submitted. Among them, some databases are determined to read this isolation level. For example, the Oracle database defaults to read the isolation level that has been submitted. Some databases do not specify the isolation level that has been submitted. For example, the MySQL database defaults to This isolation level can be read repeatedly. Based on this, before performing the method provided in this embodiment, it can be determined whether the isolation level of the database is a read commit, and if not, the database needs to be pre-configured to read the isolation level. For example, you can configure the repository to read this isolation level before the application server executes the target transaction.

Under the premise that the isolation level of the database is read, the method provided in this embodiment may be executed to execute the target transaction in the database. The specific process is as follows: The application server controls the execution logic of the entire target transaction, and the target transaction can be executed in the existing manner. The database operating device sequentially acquires the database operation instruction in the target transaction executed by the application server during the execution of the target transaction by the application server.

In an alternative embodiment, the application server sends a database operation instruction to the database in an existing manner to enable access to the database. The database operating device can monitor the communication between the application server and the database to intercept the database operation instructions sent by the application server to the database.

In another alternative embodiment, the processing logic of the application server is slightly modified, and the processing logic that originally sent the database operation instruction to the database is changed to be sent to the database operating device. Based on this, the database operating device can receive the database operation instruction actively issued by the application server.

Here, if the target transaction is an explicit transaction, the first instruction in the target transaction is a transaction start instruction, such as a begin transaction, based on which the database operation device acquires the data executed by the application server. Before the library operation instruction, the transaction start instruction in the target transaction sent by the application server to the database can be intercepted, and according to the transaction start instruction, the target transaction needs to be executed. Alternatively, the database operating device may receive a transaction start instruction in the target transaction actively issued by the application server before acquiring the database operation instruction executed by the application server, and according to the transaction start instruction, learn that the target transaction needs to be executed. . If the target transaction is an implicit transaction, the first instruction of the target transaction is the database operation instruction, and does not include the step of obtaining the transaction start instruction.

After obtaining the database operation instruction executed by the application server, the database operation device performs predictive execution on the acquired database operation instruction, returns the predicted execution result to the application server, and determines the supply server program. A database operation instruction that needs to be executed. The predicted execution result determines the execution path of the target transaction, where the execution path refers to the jump logic between the database operation instructions. By returning the predicted execution result to the application server, the purpose of controlling the execution logic of the entire target transaction by the application server is achieved.

In addition, the database operating device also needs to locally record the acquired database operation instruction, and when the database operation instruction belongs to one of the modified database class instruction and the locked non-modified database class instruction, the record pair The database operation instruction performs prediction execution data generated by the prediction execution. The modified database class instruction refers to a database operation instruction that has a modified database, for example, an instruction that starts with update, insert, etc.; generally, an instruction that starts with a select does not belong to a modified database class instruction (referred to as a non-modified database). Class instruction). Although in essence, reading the submitted isolation level only guarantees that writing multiple data is atomic, and does not lock when reading, but in order to ensure transaction consistency, reduce the probability of return, some non-modified database class instructions It may need to be locked by the for update statement. For example, in the above transfer example, the select instruction needs to be locked to form a database operation instruction in the form of select...forupdate, which can be called a locked non- Modify the database class directive. The predicted execution data mainly refers to some data in the process of predictive execution of the database operation instruction, such as the data operated by the database operation instruction, the primary key ID of the data, and the version number information corresponding to the primary key ID.

It is worth noting that for the same database operation instruction, it is either a modified database class instruction, a locked non-modified database class instruction, or an unlocked non-modified database class instruction. However, from the perspective of the entire target transaction, the predicted execution data generated by all the modified database class instructions and all the locked non-modified database class instructions in the target transaction will be recorded locally.

For example, the database operating device may locally create a memory library, and store the acquired database operation instructions and predicted execution data in a memory library. Further, if the predicted execution data does not include the predicted execution result, the predicted execution result may also be stored in the memory bank.

In order to implement the simulation of the locking scheme under the isolation level that has been submitted for reading on the basis of the core idea of predictive execution, this embodiment is directed to the instruction of the modified database class and the locked non-modified database class instruction. One of the database operation instructions records the predicted execution data generated by the prediction execution, but does not record the predicted execution data corresponding to the database operation instruction of the unmodified non-modified database type instruction. It can be seen that in the case of realizing that the lock scheme under the isolation level has been submitted, the amount of recorded data is relatively small, so that the subsequent database-based operation instructions and prediction execution data based on the local record are in the database. In the process of real execution of the target transaction, the amount of data to be processed is relatively small, which is beneficial to further improve execution efficiency and performance, and increase transaction throughput.

The operation of the database operation instruction on the database mainly refers to accessing the data in the database, and the data environment of the database operation instruction can be simulated, and the database operation instruction is predicted and executed based on the simulated data environment. Further, considering that the database operation instruction belonging to the non-modified database class instruction does not cause the change of the database, the database operation instruction can be directly predicted and executed in the database, thereby saving the operation of the simulation data environment. The implementation is relatively simple, which is conducive to saving resources and improving execution efficiency. However, for the database operation instruction belonging to the modified database class instruction, since it will cause the database to change (mainly the data in the database is changed), the simulation data environment is required to be performed in the simulated data environment. Forecast execution.

Based on the above analysis, a method for predictive execution of the database operation instruction may be: identifying the category of the obtained database operation instruction; if the database operation instruction belongs to the modification database type instruction, the locally created memory The data environment of the simulation database operation instruction in the library performs prediction execution on the database operation instruction based on the simulated data environment; if the database operation instruction belongs to the non-modified database type instruction, the database operation instruction is executed in the database, Predictive execution of the database operation instructions.

Further, a data environment for simulating a database operation instruction in a locally created memory library, and an implementation manner of predicting execution of the database operation instruction based on the simulated data environment includes: splitting the obtained database operation instruction into A read instruction and a write instruction, the read instruction is run on a real database, that is, the read instruction is executed in the database to obtain a read data set (ReadSet), and the read data set is stored in the local memory library, The data environment required to simulate database operations instructions. Then, the write command is applied to the memory bank to implement predictive execution of the database operation instruction, that is, executing a write instruction in the memory bank to modify the read data set, for example, updating or querying related data in the read data set. Processing and so on. Wherein, executing a write command to modify the read data set may generate a result data set (affectRowInMemdb), the result data set including the predicted execution result.

In the above prediction execution process, if the database operation instruction belongs to the modification database class instruction, the database operation device can locally record the read data set read by the read instruction and the result data set generated by executing the write instruction. It can be seen that the predicted execution data may include a read data set and a result data set, or may also include some data in the read data set and the result data set, for example, some of the beneficial effects on the actual execution of the target transaction in the database. Data, such as various indexes of some values, such as primary index, secondary index, primary key ID, version number, and so on.

Preferably, in order to further reduce the amount of recorded data, when the database operation instruction belongs to the locked non-modified database class instruction, only the primary key ID and version number of the data operated by the database operation instruction may be recorded as a prediction. Execution data. In this way, in the subsequent process of realizing the target transaction in the database based on the local record-based database operation instruction and the prediction execution data, the amount of data to be processed is further reduced, which is beneficial to further improve the execution efficiency and performance, and increase the transaction. Throughput. In addition, sometimes the data may be modified several times during the transaction, but eventually return to the initial value, but the primary key ID and version number of the data will not change. Therefore, if the method of recording the data itself is used, the data may be different when the forecast execution is compared with the real execution, and the return will occur, and the cost of the return is greater. However, in this embodiment, the method of recording only the primary key ID and the version number of the data is adopted, and the probability of occurrence of inconsistency when the prediction execution is compared with the actual execution is low, and the probability of the return is low.

When the database operation device acquires the transaction commit instruction in the target transaction (that is, the application server executes to the transaction commit instruction), it can control the corresponding application server according to the locally recorded database operation instruction and the predicted execution data. The database actually performs the target transaction.

In the above, according to the locally recorded database operation instruction and the predicted execution data, an optional implementation manner of controlling the database corresponding to the application server to actually execute the target transaction includes: the database operation device delivering the locally recorded data to the database a library operation instruction to instruct the database to execute the database operation instruction, and receive the actual execution result of the database operation instruction returned by the database; and then compare the actual execution result of the database operation instruction with the predicted execution result, if the actual execution The result is the same as the predicted execution result, and the transaction commit instruction is sent to the database for the database to submit the target transaction; if the actual execution result is different from the predicted execution result, the transaction return instruction is sent to the database for the database. Transfer back to the target transaction.

In the above embodiment, considering the atomicity of the transaction, by comparing the actual execution result with the predicted execution result, the transaction execution process error can be avoided, and the success probability of the transaction execution is improved.

Further, in the prior art, since all the database operation instructions that the transaction needs to execute cannot be known in advance, the database operation instruction can be executed sequentially according to the execution logic of the transaction, which causes frequent interaction between the application server and the database. This will consume a lot of network resources in the remote application scenario. However, in this embodiment, all the database operation instructions that need to be executed by the target transaction, that is, the local record database operation instruction, have been obtained in advance through the prediction execution process, so that the local record database operation instruction is sent to the database to In the process of instructing the database to execute the database operation instruction, the local record database operation instruction can be simultaneously (or collectively) sent to the database, which is conducive to saving network resources. In addition, for some parallel database operation instructions, the database can also operate in parallel, which is beneficial to further improve transaction execution efficiency and increase transaction throughput.

Further, all the database operation instructions recorded locally can be sent to the database together with the locally recorded predicted execution data.

Further, in an optional implementation manner of controlling a database corresponding to the actual execution of the target transaction according to the locally recorded database operation instruction and the predicted execution data, the database operation device may operate the instruction according to the locally recorded database. , determining whether the target transaction is a stand-alone transaction; if the determination result is yes, according to the locally recorded database operation instruction and the predicted execution data, the single-machine transaction processing logic controls the database to actually execute the target transaction; if the judgment result is no, then according to The locally recorded database operation instructions and the predicted execution data, and the distributed transaction logic control the database to actually execute the target transaction.

The database operation instruction obtained during the prediction execution process can identify whether the target transaction is a stand-alone transaction or a distributed transaction. For example, it may be determined whether the operation object of the locally recorded database operation instruction acts on the same entity device. If the determination result is yes, the target transaction may be determined to be a stand-alone transaction; if the determination result is negative, the target transaction may be determined. For distributed transactions.

Because the processing logic of distributed transactions is different from the processing logic of stand-alone transactions, the processing logic of single-machine transactions is relatively simple, for example, it does not involve problems such as read-write and lock, so the resources consumed are relatively small. Based on this, the present embodiment uses the processing logic of the single-machine transaction to identify the target transaction as a single-machine transaction, instead of using the distributed transaction processing logic to process the same as the prior art, which is beneficial to improve processing. Efficiency, saving processing costs.

In an optional implementation manner, if the database operation device acquires a transaction return instruction in the target transaction (ie, the application server executes to the transaction return instruction), the local record database operation instruction and the prediction execution may be deleted. For example, the local memory bank can be directly emptied to implement the transfer operation. In this case, since the target transaction is not actually executed in the database, the database does not need to perform the transfer operation. It can be seen that in the case of the transaction return, the transaction efficiency can be improved by using the method of the embodiment.

It can be seen from the above that in the method provided by the embodiment, the database operation device cooperates with the application server to join the prediction execution process in the process of executing the target transaction by the application server, and the target can be obtained in advance during the prediction execution process. The execution path of the transaction, that is, the database operation instruction that really needs to be executed, and only for the database operation instruction belonging to the modified database class instruction, or only for the database operation instruction record belonging to the locked non-modified database class instruction. Predicting the execution data generated by the execution to adapt to the locking scheme under the submitted isolation level, providing conditions for the real execution of the transaction, and then actually executing the target transaction based on the recorded database operation instructions and the predicted execution data. For example, based on the predicted execution data, the related information information can be obtained in advance and the jump between instructions can be reduced, thereby improving execution efficiency and increasing transaction throughput.

In the following, taking the inventory reduction transaction as an example, the workflow of the technical solution of the present invention will be described in detail. The isolation level of the database is the isolation level that the read has been submitted.

The SQL code for the inventory reduction transaction is as follows, the text in parentheses is the comment: begin transaction (select transaction) select * from inventory where itemId=? For update (use commodity id to view current inventory, note that for update external display lock) if (item.inventory>0) item.inventory--;update inventory set item.inventory=$item.inventory where itemId=? Commit; else rollback; (if the current inventory is greater than 0, reduce, update the merchandise inventory and submit, otherwise return)

In the actual execution process, the application server is responsible for the whole logic of the inventory reduction transaction. The database operation device (or also called the execution server) only gets the start transaction executed by the application server, and the SQL to be executed. Statements, as well as commit/return instructions.

First, after starting the transaction, the database operating device creates a memory library locally; when the database operating device obtains select * from inventory where itemId=? For update this SQL statement, it is recognized that this SQL statement will not modify the database, so this SQL statement directly acts on the actual database, that is, execute the SQL statement in the database to obtain the query result, ie Read the data set, return all the query results to the application server, and supply the application server to determine the next SQL statement to be executed, and record the SQL statement in the local memory library. Although the read level has been read, since the statement is locked, it is necessary to record the data related to executing the SQL statement. Preferably, only the primary key ID and the version number of the data queried by the SQL statement can be recorded.

After receiving the query result returned by the database operation device, the application server determines whether the current inventory quantity in the query result is greater than 0. If the determination result is yes, the current inventory quantity is reduced, and the update statement is executed, that is, the update inventory set item .inventory=$item.inventory where itemId=? If the result of the judgment is no, the rollback statement is executed.

In the first case, if the application server executes the rollback statement, the database operating device will get the rollback statement. Once the database storage device obtains the rollback statement, the local memory library is emptied, mainly to clear the previously selected select * from inventory where itemId=? For update this SQL statement. In this case, because the transaction is not executed in the repository, there is no need for the repository to perform the rollback operation.

In the second case, assume that the application server has executed update inventory set item.inventory=$item.inventory where itemId=? In this statement, the database operating device will get update inventory set item.inventory=$item.inventory where itemId=? This statement. Then the database operating device obtains update inventory set item.inventory=? Where itemId=? This SQL statement recognizes that the SQL statement will modify the database, so the SQL statement is split into a read command, ie select * from inventory where itemId=? And a write command, ie update inventory set item.inventory=? Where itemId=? Then, select * from inventory where itemId=? The statement is sent to the database, and the read data set is obtained, and the data in the read data set is written into the local memory library; wherein the read data set includes a primary key list (itemId) and a version number corresponding to the primary key. Then, according to update inventory set item.inventory=? Where itemId=? This statement is updated in the local memory library to obtain the result data set and return the result data set to the application server. The result data set may be modified record information similar to the modification of several records. For example, affectRow=1 indicates that a record is modified; or the result data set may also be modified details of the record modification detailed information. Among them, the result data set and the read data set constitute prediction execution data.

The application server executes the commit command at this time.

After the database operation device obtains the commit command, the inventory reduction transaction is executed in the database based on the SQL statement in the local memory library and the predicted execution data, and the transaction execution process is actually performed. For example, the locally recorded SQL statement and the predicted execution data are as follows: {select * from inventory where itemId=? Update inventory set item.inventory=? Where itemId=? [readVersion:itemId=xxx and version=xxx /affectRow=1]commit()}

Preferably, the database operating device can submit all the SQL statements in the local memory library to the database at one time, and the database is executed based on the data, thereby saving the network resources consumed by the transmission of the SQL statement.

This real execution will also return the execution result. You can compare the predicted execution result with the actual execution result. For example, you can compare the primary key ID and version number of the two results, and then compare the affectRow of the two results; if the comparison result is The same, you can submit this transaction in the database, and if the comparison results are not the same, then the submission is considered to be a failure, because in this case, the transaction itself has not actually executed the submission in the database, so Simply return the entire request and let the application server resubmit.

Among them, after the above prediction execution, all the SQL statements that need to be executed, and the corresponding predicted execution data, such as all the segmentation conditions, the index of the value, the version number, etc., have been obtained, and the data is sufficient for us to advance whether the transaction is The single machine transaction makes a judgment. Based on this, the corresponding transaction logic can be used for processing, which is conducive to saving resources.

FIG. 4 is a schematic flowchart diagram of a method for operating a database according to another embodiment of the present invention. As shown in FIG. 4, the method includes:

401. In the process of executing the target transaction by the application server, sequentially acquiring the database operation instruction in the target transaction executed by the application server.

402. Perform predictive execution on the database operation instruction, return the predicted execution result to the application server to supply the application server to determine the next database operation instruction to be executed, and record the database operation instruction locally and in the database. The operation execution instruction records the prediction execution data generated by the prediction execution when it belongs to the locked non-modified database class instruction.

403. When the transaction commit instruction in the target transaction is obtained, according to the locally recorded database operation instruction and the predicted execution data, the database corresponding to the application server is controlled to actually execute the target transaction.

This embodiment is similar to the embodiment shown in FIG. 3, except that only the predicted execution data generated by the prediction execution of the locked non-modified database class instruction in the target transaction is recorded, and the prediction execution of the modified database class instruction is generated. The forecast execution data and the forecast execution data generated by the predictive execution of the unblocked non-modified database class instructions are not recorded. In addition, the embodiment is applicable to the case where the target transaction includes the unlocked non-modified database class instruction, and the operated database can be not only the database level with the isolation level being read, but also other database scenarios. .

If the method provided in this embodiment is applied to the database whose isolation level is read, it is considered that some databases are defined as the isolation level that the read has been submitted. For example, the Oracle database defaults to read the submitted isolation level. Some databases do not specify that the isolation level has been read. For example, the MySQL database defaults to the isolation level. Therefore, before executing the method provided in this embodiment, it can be determined whether the isolation level of the database is read. Submitted, if not, you need to pre-configure the repository to read this isolation level. For example, you can configure the repository to read this isolation level before the application server executes the target transaction.

In an alternative embodiment, the application server sends a database operation instruction to the database in an existing manner to enable access to the database. The database operating device can monitor the communication between the application server and the database to intercept the database operation instructions sent by the application server to the database.

In another alternative embodiment, the processing logic of the application server is slightly modified, and the processing logic that originally sent the database operation instruction to the database is changed to be sent to the database operating device. Based on this, the database operating device can receive the database operation instruction actively issued by the application server.

Here, if the target transaction is an explicit transaction, the first instruction in the target transaction is a transaction start instruction, such as a begin transaction, based on which the database operation device acquires the data executed by the application server. Before the library operation instruction, the transaction start instruction in the target transaction sent by the application server to the database can be intercepted, and according to the transaction start instruction, the target transaction needs to be executed. Alternatively, the database operating device may receive a transaction start instruction in the target transaction actively issued by the application server before acquiring the database operation instruction executed by the application server, and according to the transaction start instruction, learn that the target transaction needs to be executed. . If the target transaction is an implicit transaction, the first instruction of the target transaction is the database operation instruction, and does not include the step of obtaining the transaction start instruction.

After obtaining the database operation instruction executed by the application server, the database operation device performs predictive execution on the acquired database operation instruction, returns the predicted execution result to the application server, and determines the supply server program. A database operation instruction that needs to be executed. The predicted execution result determines the execution path of the target transaction, where the execution path refers to the jump logic between the database operation instructions. By returning the predicted execution result to the application server, the purpose of controlling the execution logic of the entire target transaction by the application server is achieved.

In addition, the database operating device also needs to locally record the acquired database operation instruction, and when the database operation instruction belongs to the locked non-modified database type instruction, the record performs prediction execution on the database operation instruction. Forecasted execution data generated. The modified database class instruction refers to a database operation instruction that has a modified database, for example, an instruction that starts with update, insert, etc.; generally, an instruction that starts with a select does not belong to a modified database class instruction (referred to as a non-modified database). Class instruction). Sometimes, in order to ensure the consistency of the transaction and reduce the probability of the return, some non-modified database class instructions may need to be locked by the for update statement. For example, in the above transfer example, the select instruction needs to be locked to form a select... ...for update form database operation instructions, such instructions can be referred to as locked non-modified database class instructions. The predicted execution data mainly refers to some materials in the process of predicting execution of the database operation instruction, such as the data operated by the database operation instruction and the primary key ID (itemId) of the data and the version number information corresponding to the primary key ID (version). .

It is worth noting that for the same database operation instruction, it is either a modified database class instruction, a locked non-modified database class instruction, or an unlocked non-modified database class instruction. In this embodiment, the predicted execution data generated for the prediction execution of all locked non-modified repository class instructions in the target transaction is recorded locally.

For example, the database operating device may locally create a memory library, and store the acquired database operation instructions and predicted execution data in a memory library. Further, if the predicted execution data does not include the predicted execution result, the predicted execution result may also be stored in the memory bank.

In order to implement the simulation of the locking scheme on the basis of the core idea of predictive execution, the present embodiment records the predicted execution data generated by the prediction execution for the database operation instruction of the unlocked non-modified database class instruction. The predicted execution data corresponding to other database operation instructions is not recorded. It can be seen that in the case of implementing the locking scheme, the amount of recorded data is relatively small, so that in the subsequent process of executing the target transaction in the database in the database based on the local record-based database operation instruction and the prediction execution data, The amount of data to be processed is relatively small, which is beneficial to further improve execution efficiency and performance, and increase transaction throughput.

Preferably, when the database operation instruction belongs to the locked non-modified database class instruction, only the primary key ID and the version number of the data operated by the database operation instruction may be recorded as the predicted execution data. In this way, in the subsequent process of realizing the target transaction in the database based on the local record-based database operation instruction and the prediction execution data, the amount of data to be processed is further reduced, which is beneficial to further improve the execution efficiency and performance, and increase the transaction. Throughput. In addition, sometimes the data may be modified several times during the transaction, but eventually return to the initial value, but the primary key ID and version number of the data will not change. Therefore, if the method of recording the data itself is used, the data may be different when the forecast execution is compared with the real execution, and the return will occur, and the cost of the return is greater. However, in this embodiment, the method of recording only the primary key ID and the version number is adopted, and the probability of occurrence of inconsistency when the prediction execution is compared with the real execution is low, and the probability of the return is low.

The operation of the database operation instruction on the database mainly refers to accessing the data in the database, and the data environment of the database operation instruction can be simulated, and the database operation instruction is predicted and executed based on the simulated data environment. Further, considering that the database operation instruction belonging to the non-modified database class instruction does not cause the change of the database, the database operation instruction can be directly predicted and executed in the database, thereby saving the operation of the simulation data environment. The implementation is relatively simple, which is conducive to saving resources and improving execution efficiency. However, for the database operation instruction belonging to the modified database class instruction, since it will cause the database to change (mainly the data in the database is changed), the simulation data environment is required to be performed in the simulated data environment. Forecast execution.

Based on the above analysis, a method for predictive execution of the database operation instruction may be: identifying the category of the obtained database operation instruction; if the database operation instruction belongs to the modification database type instruction, the locally created memory The data environment of the simulation database operation instruction in the library predicts execution of the database operation instruction based on the simulated data environment; if the database operation instruction belongs to the non-modified database type instruction, the database operation instruction is executed in the database, Predictive execution of the database operation instructions.

Further, a data environment for simulating a database operation instruction in a locally created memory library, and an implementation manner of predicting execution of the database operation instruction based on the simulated data environment includes: splitting the obtained database operation instruction into A read instruction and a write instruction, the read instruction is run on a real database, that is, the read instruction is executed in the database to obtain a read data set (ReadSet), and the read data set is stored in the local memory library, The data environment required to simulate database operations instructions. Then, the write command is applied to the memory bank to implement predictive execution of the database operation instruction, that is, executing a write instruction in the memory bank to modify the read data set, for example, updating or querying related data in the read data set. Processing and so on. Wherein, executing a write command to modify the read data set may generate a result data set (affectRowInMemdb), the result data set including the predicted execution result.

When the database operation device acquires the transaction commit instruction in the target transaction (that is, the application server executes to the transaction commit instruction), it can control the corresponding application server according to the locally recorded database operation instruction and the predicted execution data. The database actually performs the target transaction.

The above-mentioned optional implementation manner of controlling the database corresponding to the application server according to the locally recorded database operation instruction and the prediction execution data includes: the database operation device sends the local record database operation to the database An instruction to instruct the database to execute the database operation instruction, and receive the actual execution result of the database operation instruction returned by the database; and then compare the actual execution result of the database operation instruction with the predicted execution result, if the actual execution result is The forecast execution result is the same, and the transaction commit instruction is sent to the database for the database to submit the target transaction; if the actual execution result is different from the predicted execution result, the transaction return instruction is sent to the database for the database to be returned. The target transaction.

In the above embodiment, considering the atomicity of the transaction, by comparing the actual execution result with the predicted execution result, the transaction execution process error can be avoided, and the success probability of the transaction execution is improved.

Further, in the prior art, since all the database operation instructions that the transaction needs to execute cannot be known in advance, the database operation instruction can be executed sequentially according to the execution logic of the transaction, which causes frequent interaction between the application server and the database. This will consume a lot of network resources in the remote application scenario. However, in this embodiment, all the database operation instructions that need to be executed by the target transaction, that is, the local record database operation instruction, have been obtained in advance through the prediction execution process, so that the local record database operation instruction is sent to the database to In the process of instructing the database to execute the database operation instruction, the local record database operation instruction can be simultaneously (or collectively) sent to the database, which is conducive to saving network resources. In addition, for some parallel database operation instructions, the database can also operate in parallel, which is beneficial to further improve transaction execution efficiency and increase transaction throughput.

Further, all the database operation instructions recorded locally can be sent to the database together with the locally recorded predicted execution data.

Further, in an optional implementation manner of controlling a database corresponding to the actual execution of the target transaction according to the locally recorded database operation instruction and the predicted execution data, the database operation device may operate the instruction according to the locally recorded database. , determining whether the target transaction is a stand-alone transaction; if the determination result is yes, according to the locally recorded database operation instruction and the predicted execution data, the single-machine transaction processing logic controls the database to actually execute the target transaction; if the judgment result is no, then according to The locally recorded database operation instructions and the predicted execution data, and the distributed transaction logic control the database to actually execute the target transaction.

The database operation instruction obtained during the prediction execution process can identify whether the target transaction is a stand-alone transaction or a distributed transaction. For example, it may be determined whether the operation object of the locally recorded database operation instruction acts on the same entity device. If the determination result is yes, the target transaction may be determined to be a stand-alone transaction; if the determination result is negative, the target transaction may be determined. For distributed transactions.

Because the processing logic of distributed transactions is different from the processing logic of stand-alone transactions, the processing logic of single-machine transactions is relatively simple, for example, it does not involve problems such as read-write and lock, so the resources consumed are relatively small. Based on this, the present embodiment uses the processing logic of the single-machine transaction to identify the target transaction as a single-machine transaction, instead of using the distributed transaction processing logic to process the same as the prior art, which is beneficial to improve processing. Efficiency, saving processing costs.

In an optional implementation manner, if the database operation device acquires a transaction return instruction in the target transaction (ie, the application server executes to the transaction return instruction), the local record database operation instruction and the prediction execution may be deleted. For example, the local memory bank can be directly emptied to implement the transfer operation. In this case, since the target transaction is not actually executed in the database, the database does not need to perform the transfer operation. It can be seen that in the case of the transaction return, the transaction efficiency can be improved by using the method of the embodiment.

It can be seen from the above that in the method provided by the embodiment, the database operation device cooperates with the application server to join the prediction execution process in the process of executing the target transaction by the application server, and the target can be obtained in advance during the prediction execution process. The execution path of the transaction, that is, the database operation instruction that really needs to be executed, and only records the prediction execution data generated by the prediction execution for the database operation instruction belonging to the locked non-modified database class instruction to adapt to the locking scheme. Providing conditions for the real execution of the transaction, and then when the target transaction is actually executed according to the recorded database operation instruction and the prediction execution data, for example, based on the prediction execution data, the related information information can be acquired in advance and the jump between the instructions can be reduced, etc., Can improve execution efficiency and increase transaction throughput.

It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should understand that the present invention is not limited by the described action sequence. Because certain steps may be performed in other sequences or concurrently in accordance with the present invention. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

In the above embodiments, the descriptions of the various embodiments are different, and the parts that are not detailed in an embodiment can be referred to the related descriptions of other embodiments.

FIG. 5 is a schematic structural diagram of a database operating device according to another embodiment of the present invention. As shown in FIG. 5, the device includes an acquisition module 51, a prediction execution module 52, and a control execution module 53.

The obtaining module 51 is configured to sequentially acquire a database operation instruction in the target transaction executed by the application server during the execution of the target transaction by the application server.

The predictive execution module 52 is configured to perform predictive execution on the database operation instruction, return the predicted execution result to the application server to supply the application server to determine the next database operation instruction to be executed, and record the database locally. The operation instruction and the prediction execution data generated by the prediction execution are recorded when the database operation instruction belongs to one of the modified database class instruction and the locked non-modified database class instruction.

The control execution module 53 is configured to control the database corresponding to the application server to actually execute the target transaction according to the locally recorded database operation instruction and the predicted execution data when the transaction commit instruction in the target transaction is acquired.

In an optional implementation manner, the predictive execution module 52 records the predicted execution data generated by the predictive execution when the database operation instruction belongs to one of the modified database class instruction and the locked non-modified database class instruction. Specifically, when the database operation instruction belongs to the locked non-modified database class instruction, only the primary key ID and the data version number of the data operated by the database operation instruction are recorded as the predicted execution data.

In an optional implementation, the apparatus further includes: a configuration module, configured to configure the database to read the submitted isolation level before the application server performs the target transaction.

In an optional implementation manner, the prediction execution module 52 is specifically configured to: if the database operation instruction belongs to the modified database class instruction, simulate the data environment of the database operation instruction in the locally created memory library, and the simulation based data The environment performs predictive execution on the database operation instruction; if the database operation instruction belongs to the non-modified database type instruction, the database operation instruction is executed in the database to perform prediction execution on the database operation instruction.

Further, the predictive execution module 52 simulates the data environment of the database operation instruction in the locally created memory library when the database operation instruction belongs to the modified database type instruction, and performs the database operation instruction based on the simulated data environment. When the prediction is executed, it is specifically used to: split the database operation instruction into a read instruction and a write instruction; execute a read instruction in the database to acquire the read data set, and store the read data set in the memory library to simulate the database operation. The data environment of the instruction; a write instruction is executed in the memory bank to modify the read data set.

In an optional implementation, the control execution module 53 is specifically configured to: deliver a locally recorded database operation instruction to the database, instruct the database to execute the database operation instruction, and receive the database operation instruction returned by the database The actual execution result; if the actual execution result is the same as the predicted execution result, the transaction commit instruction is sent to the database for the database to submit the target transaction; if the actual execution result is different from the predicted execution result, the transaction is sent to the database. Return instructions for the database to return to the target transaction.

The database operation device provided in this embodiment cooperates with the application server to join the prediction execution process in the process of executing the target transaction by the application server, and obtains and records all the database operation instructions that need to be executed in the target transaction in advance, and only For the database operation instruction belonging to the modified database class instruction and the locked non-modified database class instruction, or only for the database operation instruction belonging to the locked non-modified database class instruction, the prediction execution generated by the prediction execution is recorded. The data is adapted to read the locking scheme under the submitted isolation level, and provides conditions for the real execution of the transaction, and then controls the database corresponding to the application server to actually execute the target according to the recorded database operation instruction and the predicted execution data. Transactions help improve execution efficiency and increase transaction throughput.

FIG. 6 is a schematic structural diagram of a database operating device according to another embodiment of the present invention. As shown in FIG. 6, the apparatus includes an acquisition module 61, a prediction execution module 62, and a control execution module 63.

The obtaining module 61 is configured to sequentially acquire the database operation instruction in the target transaction executed by the application server during the execution of the target transaction by the application server.

The predictive execution module 62 is configured to perform predictive execution on the database operation instruction, return the predicted execution result to the application server to supply the application server to determine the next database operation instruction to be executed, and record the database locally. The operation instruction and the prediction execution data generated by the prediction execution are recorded when the database operation instruction belongs to the locked non-modified database class instruction.

The control execution module 63 is configured to control the database corresponding to the application server to actually execute the target transaction according to the locally recorded database operation instruction and the predicted execution data when the transaction commit instruction in the target transaction is acquired.

In an optional implementation, the predictive execution module 62 is specifically configured to: when the database operation instruction belongs to the locked non-modified database class instruction, record only the primary key ID of the data operated by the database operation instruction and The version number is used as the forecast execution data.

In an optional implementation, the apparatus further includes: a configuration module, configured to configure the database to read the submitted isolation level before the application server performs the target transaction.

In an optional implementation manner, the prediction execution module 62 is specifically configured to: if the database operation instruction belongs to the modified database class instruction, simulate the data environment of the database operation instruction in the locally created memory library, and simulate the data The environment performs predictive execution on the database operation instruction; if the database operation instruction belongs to the non-modified database type instruction, the database operation instruction is executed in the database to perform prediction execution on the database operation instruction.

Further, the prediction execution module 62 simulates the data environment of the database operation instruction in the locally created memory library when the database operation instruction belongs to the modified database class instruction, and performs the database operation instruction based on the simulated data environment. When the prediction is executed, it is specifically used to: split the database operation instruction into a read instruction and a write instruction; execute a read instruction in the database to acquire the read data set, and store the read data set in the memory library to simulate the database operation. The data environment of the instruction; a write instruction is executed in the memory bank to modify the read data set.

In an optional implementation, the control execution module 63 is specifically configured to: deliver a locally recorded database operation instruction to the database, instruct the database to execute the database operation instruction, and receive the database operation instruction returned by the database The actual execution result; if the actual execution result is the same as the predicted execution result, the transaction commit instruction is sent to the database for the database to submit the target transaction; if the actual execution result is different from the predicted execution result, the transaction is sent to the database. Return instructions for the database to return to the target transaction.

The database operation device provided in this embodiment cooperates with the application server to join the prediction execution process in the process of executing the target transaction by the application server, and obtains and records all the database operation instructions that need to be executed in the target transaction in advance, and only Recording the predicted execution data generated by the prediction execution for the database operation instruction belonging to the locked non-modified database class instruction to adapt to the locking scheme, providing conditions for the real execution transaction, and then operating the instruction according to the recorded database and Predicting the execution data and controlling the database corresponding to the application server to actually execute the target transaction is beneficial to improve the execution efficiency and thereby increase the transaction throughput.

A person skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above can refer to the corresponding process in the foregoing method embodiment, and details are not described herein again.

In the several embodiments provided by the present invention, it should be understood that the disclosed system, apparatus, and method may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or may be Consolidate to another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be electrical, mechanical or otherwise.

The units described as separate components may or may not be physically separated, and the components displayed as the unit may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. . Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist as a separate entity, or two or more units may be integrated into one unit. The above integrated unit can be implemented in the form of a hardware or a hardware plus a soft functional unit.

The integrated unit described above in the form of a software functional unit can be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to perform the embodiments of the present invention. Part of the steps of the method. The foregoing storage media include: USB flash drive, mobile hard disk, Read-Only Memory (ROM), Random Access Memory (RAM), disk or CD-ROM, and other storage programs. Code media.

It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not limited thereto; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that The technical solutions described in the foregoing embodiments are modified, or the equivalents of the technical features are replaced. The modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (14)

 A method for operating a database, comprising: acquiring, in a process of executing a target transaction by an application server, a database operation instruction in the target transaction executed by the application server; and operating the database operation instruction Performing predictive execution, returning the predicted execution result to the application server for the application server to determine the next database operation instruction to be executed, and locally recording the database operation instruction and the operation instruction in the database belongs to The predicted execution data generated by the prediction execution is recorded when one of the database class instruction and the locked non-modified database class instruction is modified; when the transaction commit instruction in the target transaction is obtained, the database is operated according to the local record The instruction and the predicted execution data control the database corresponding to the application server to actually execute the target transaction.    The method of claim 1, wherein the predictive execution data generated by the prediction execution is recorded when the database operation instruction belongs to one of a modified database class instruction and a locked non-modified database class instruction, including When the database operation instruction belongs to the locked non-modified database class instruction, only the primary key ID and the version number of the data operated by the database operation instruction are recorded as the predicted execution data.    The method of claim 1, wherein the method further comprises: configuring the repository to read the submitted isolation level before the application server executes the target transaction.    According to the method of claim 1, wherein the predictive execution of the database operation instruction comprises: if the database operation instruction belongs to a modification database class instruction, simulating the database in a locally created memory library The data environment of the operation instruction is predicted to execute the database operation instruction based on the simulated data environment; if the database operation instruction belongs to the non-modified database type instruction, the database operation instruction is executed in the database, so as to The database operation instruction performs prediction execution.    According to the method of claim 4, wherein the data environment required to simulate the database operation instruction is generated in the locally created memory library, and the database operation instruction is predicted and executed based on the simulated data environment, including : splitting the database operation instruction into a read instruction and a write instruction; executing the read instruction in the database to obtain a read data set, and storing the read data set in the memory bank to simulate the database operation instruction Data environment; executing the write command in the memory bank to modify the read data set.    The method of any one of claims 1 to 5, wherein the database corresponding to the locally recorded database operation instruction and the predicted execution data, the database corresponding to the application server is controlled to actually execute the target transaction, The method includes: sending the local database operation instruction to the database to instruct the database to execute the database operation instruction, and receiving an actual execution result of the database operation instruction returned by the database; if the actual execution The result is the same as the prediction execution result, and the transaction commit instruction is sent to the database for the database to submit the target transaction; if the actual execution result is different from the predicted execution result, the transaction is sent to the database. A rollback instruction for the repository to return to the target transaction.    A method for operating a database, comprising: acquiring, in a process of executing a target transaction by an application server, a database operation instruction in the target transaction executed by the application server; and operating the database operation instruction Performing predictive execution, returning the predicted execution result to the application server for the application server to determine the next database operation instruction to be executed, and locally recording the database operation instruction and the operation instruction in the database belongs to The unlocked non-modified database class instruction records the predicted execution data generated by the prediction execution; when the transaction commit instruction in the target transaction is obtained, the database operation instruction and the predicted execution data are locally controlled, and the control is performed. The database corresponding to the application server actually executes the target transaction.    A database operating device, comprising: an obtaining module, configured to sequentially acquire a database operation instruction in the target transaction executed by the application server during a process of executing a target transaction by the application server; The predictive execution module is configured to perform predictive execution on the database operation instruction, and return the predicted execution result to the application server for the application server to determine the next database operation instruction to be executed, and record locally The database operation instruction and the prediction execution data generated by the prediction execution when the database operation instruction belongs to one of the modified database class instruction and the locked non-modified database class instruction; the control execution module is used when When the transaction commit instruction in the target transaction is obtained, according to the database operation instruction recorded locally and the predicted execution data, the database corresponding to the application server is controlled to actually execute the target transaction.    The device of claim 8 , wherein the predictive execution module is specifically configured to: when the database operation instruction belongs to a locked non-modified database class instruction, record only the operation of the database operation instruction The primary key ID and version number of the data are used as the predicted execution data.    The device according to claim 8 , further comprising: a configuration module, configured to configure the database to read the submitted isolation level before the application server executes the target transaction.    The device according to claim 8 , wherein the predictive execution module is specifically configured to: if the database operation instruction belongs to a modified database class instruction, simulate the data of the database operation instruction in a locally created memory library Environment, based on the simulated data environment, predictive execution of the database operation instruction; if the database operation instruction belongs to a non-modified database class instruction, execute the database operation instruction in the database to operate the database The instruction performs predictive execution.    The apparatus of claim 11, wherein the predictive execution module is specifically configured to: split the database operation instruction into a read instruction and a write instruction; execute the read instruction in the database to obtain a read data set And storing the read data set in the memory library to simulate a data environment of the database operation instruction; executing the write instruction in the memory library to modify the read data set.    The device according to any one of the preceding claims, wherein the control execution module is configured to: deliver the locally recorded database operation instruction to the database to instruct the database to execute the data The library operation instruction receives the actual execution result of the database operation instruction returned by the database; if the actual execution result is the same as the prediction execution result, the transaction commit instruction is sent to the database for submission by the database The target transaction; if the actual execution result is different from the predicted execution result, a transaction return instruction is sent to the database for the database to return to the target transaction.    A database operating device, comprising: an obtaining module, configured to sequentially acquire a database operation instruction in the target transaction executed by the application server during a process of executing a target transaction by the application server; The predictive execution module is configured to perform predictive execution on the database operation instruction, and return the predicted execution result to the application server for the application server to determine the next database operation instruction to be executed, and record locally The database operation instruction and recording the predicted execution data generated by the prediction execution when the database operation instruction belongs to the locked non-modified database class instruction; and the control execution module is configured to acquire the transaction in the target transaction When the instruction is submitted, according to the database operation instruction recorded locally and the predicted execution data, the database corresponding to the application server is controlled to actually execute the target transaction.