Parameter configuration adjustment method, device, electronic device and storage medium

Abstract

The present application discloses a parameter configuration adjustment method, apparatus, electronic device, and storage medium. The method includes: determining a weight coefficient corresponding to each of at least one performance indicator in a target database; discretizing the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; determining, from the at least one set of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that is a neighbor of the current weight coefficient combination and has undergone parameter configuration adjustment; determining first parameter configuration information corresponding to the first weight coefficient combination, and determining a target training sample set based on the first parameter configuration information and an initial training sample set corresponding to the current weight coefficient combination; adjusting the parameter configuration of the current weight coefficient combination based on the target training sample set to determine the target parameter configuration information corresponding to the current weight coefficient combination. This method not only saves computation time but also enables parameter configuration recommendations for multiple performance indicators.

Description

Technical Field

[0001] The present application relates to the technical field of database management, and in particular to a parameter configuration adjustment method, device, electronic device, and storage medium. Background Art

[0002] With the rapid development of computer technology, computer data processing capabilities have been greatly enhanced, and the amount of data handled by databases has also increased. A database management system (DBMS) is a large-scale software that operates and manages databases. It is used to create, use, and maintain databases. During operation, it requires the configuration of several parameters to control the behavior of the DBMS, resulting in different performance during data access. Therefore, the setting of certain parameters in the DBMS needs to be adjusted according to the business scenario to achieve optimal performance.

[0003] In the related art, there are two main approaches to the multi-objective problem of intelligent parameter adjustment (parameter configuration adjustment) of database management systems. One approach is to allow users to select the most important single performance indicator as the optimization target to avoid ignoring other performance indicators. However, this method does not support multi-objective optimization of intelligent parameter adjustment of database management systems in principle. The other approach is to allow users to manually determine the weight coefficients corresponding to each performance indicator, and then construct an addition formula based on the values ​​of each performance indicator and the weight coefficient determined by the user as a single target for parameter adjustment. However, this method only supports one weight coefficient combination for each of multiple performance indicators and cannot consider multiple weight coefficient combinations in the solution. Summary of the Invention

[0004] The present application provides a parameter configuration adjustment method, device, electronic device and storage medium, which can recommend parameter configuration information for multiple performance indicators under different weight coefficient combinations to users, and can accelerate the solution process and improve the solution efficiency.

[0005] The technical solution of this application is achieved as follows:

[0006] In a first aspect, an embodiment of the present application provides a parameter configuration adjustment method, the method comprising:

[0007] Determine a weight coefficient corresponding to at least one performance indicator in the target database;

[0008] Discretizing the at least one weight coefficient determined, respectively, to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of each of the at least one weight coefficients;

[0009] Determine, from the at least one set of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment;

[0010] Determining first parameter configuration information corresponding to the first weight coefficient combination, and determining a target training sample set based on the first parameter configuration information and an initial training sample set corresponding to the current weight coefficient combination;

[0011] Parameter configuration adjustment is performed on the current weight coefficient combination according to the target training sample set to determine target parameter configuration information corresponding to the current weight coefficient combination.

[0012] In a second aspect, an embodiment of the present application provides a parameter configuration adjustment device, the parameter configuration adjustment device comprising a determination unit, a discretization unit, and an adjustment unit; wherein:

[0013] The determining unit is configured to determine a weight coefficient corresponding to each of at least one performance indicators in the target database;

[0014] The discretization unit is configured to discretize the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of each of the at least one weight coefficients;

[0015] The determining unit is further configured to determine, from the at least one set of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment; and determine first parameter configuration information corresponding to the first weight coefficient combination, and determine a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination;

[0016] The adjustment unit is configured to perform parameter configuration adjustment on the current weight coefficient combination according to the target training sample set, and determine target parameter configuration information corresponding to the current weight coefficient combination.

[0017] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory and a processor, wherein:

[0018] The memory is used to store a computer program that can be run on the processor;

[0019] The processor is configured to execute the method according to the first aspect when running the computer program.

[0020] In a fourth aspect, an embodiment of the present application provides a computer storage medium, wherein the computer storage medium stores a computer program, and when the computer program is executed by at least one processor, the method described in the first aspect is implemented.

[0021] A parameter configuration adjustment method, device, electronic device and storage medium provided in the embodiments of the present application determine the weight coefficient corresponding to each of at least one performance indicator in a target database; discretize the at least one determined weight coefficient to determine at least one group of weight coefficient combinations; wherein each group of weight coefficient combinations is composed of a discrete weight value of at least one weight coefficient; from the at least one group of weight coefficient combinations, determine the current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has performed parameter configuration adjustment; determine first parameter configuration information corresponding to the first weight coefficient combination, and determine a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination; perform parameter configuration adjustment on the current weight coefficient combination based on the target training sample set, and determine the target parameter configuration information corresponding to the current weight coefficient combination. In this way, by discretizing the weight coefficients, it is possible to flexibly support weight coefficient combinations of different granularities; then, the optimal parameter configuration information corresponding to the weight coefficient combination that has a neighbor relationship and has been solved is selected as the target training sample set of the regression problem corresponding to the current weight coefficient combination to guide the solution process of the current weight coefficient combination, thereby not only saving the amount of calculation, accelerating the solution process, and improving the solution efficiency, but also recommending to users parameter configuration information under different weight coefficient combinations for multiple performance indicators, thereby improving practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flow chart of a parameter configuration adjustment method provided in an embodiment of the present application;

[0023] Figure 2 A flow chart of another parameter configuration adjustment method provided in an embodiment of the present application;

[0024] Figure 3 A flow chart of another parameter configuration adjustment method provided in an embodiment of the present application;

[0025] Figure 4 A flow chart of another parameter configuration adjustment method provided in an embodiment of the present application;

[0026] Figure 5 A schematic diagram of the structure of a parameter configuration adjustment device provided in an embodiment of the present application;

[0027] Figure 6A schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0028] Figure 7 A schematic diagram of the structure of another electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the accompanying drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. It should be understood that the specific embodiments described herein are only used to explain the related applications and are not intended to limit the applications. It should also be noted that for ease of description, only the portions relevant to the related applications are shown in the drawings.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0031] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0032] It should be pointed out that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0033] Database management systems, such as MySQL (Structured Query Language), PostgreSQL, and Oracle, require several parameters to be set during runtime to control their behavior, resulting in different performance during data access. For example, the MySQL configuration parameter innodb_buffer_pool_size specifies the amount of memory space MySQL uses to temporarily store table and index data in external memory. If the value of innodb_buffer_pool_size is set too small, the database management system will generally frequently access external memory such as disks during service data access, resulting in increased service latency. If the value of innodb_buffer_pool_size is set too large, in the case of multiple client connections, the memory capacity of each client connection will be limited, resulting in increased latency for some clients. Therefore, the setting of a certain parameter in the database management system needs to be adjusted according to the business scenario to achieve higher performance.

[0034] Intelligent parameter tuning for database management systems generally involves automatically selecting configuration values ​​for several database management system parameters based on the data access characteristics of business scenarios, thereby optimizing several performance indicators of the database management system. Intelligent parameter tuning for database management systems is generally a multi-objective problem, requiring parameter adjustments tailored to more than one objective, such as considering both latency and throughput targets.

[0035] The problem of intelligent parameter tuning in a database management system can be modeled as a regression problem. The independent variables in this problem consist of two parts: the value of each parameter in the database management system, and the characteristics of the data service. The dependent variable is the desired performance indicator, such as latency. Let the parameter list be K, the characteristics of the data service be I, and the desired performance indicator be y. The problem of intelligent parameter tuning in a database management system corresponds to finding a function f with the form y = f(K, I), which can predict the corresponding value of y for different values ​​of K and I. In a regression problem, the dependent variable y is a scalar (numeric value), not a vector. If a single objective is considered, y can be defined as the performance indicator value of that objective. If multiple objectives are considered, y can be defined as the weighted sum of the performance indicators for each objective. However, the weight coefficients are fixed before the regression problem is defined and do not change during the regression fitting process. It can be seen that using the regression fitting framework to solve the intelligent parameter adjustment problem of the database management system can only optimize one weight coefficient combination for each performance indicator among multiple performance indicators, and cannot solve the parameter adjustment optimization problem of multiple performance indicators under different weight coefficient combinations.

[0036] In the related art, there are two main approaches to the multi-objective problem of intelligent parameter tuning for database management systems. One approach is to allow users to select the most important single performance indicator as the optimization target, avoiding ignoring other performance indicators. For example, the intelligent parameter tuning system OtterTune. However, this approach does not support multi-objective optimization for intelligent parameter tuning of database management systems in principle. Although users can use this system to tune parameters for objective A first and then for objective B, if the user wants to comprehensively consider and balance objectives A and B, the system's solution cannot provide recommendations. Another approach is to allow users to manually determine the weight coefficients corresponding to each performance indicator, and then construct an addition formula based on the performance indicator values ​​and the user-determined weight coefficients as the single parameter tuning target. For example, the intelligent parameter tuning system CDBTune. However, this approach only supports one weight coefficient combination for each of the multiple performance indicators and cannot consider multiple weight coefficient combinations in the solution. Although users can set different weight coefficient combinations by running the system multiple times to obtain different parameter configuration results, the solution process for one weight coefficient combination is considered independent of the solution process for another weight coefficient combination.

[0037] Based on this, an embodiment of the present application provides a parameter configuration adjustment method, the basic idea of ​​which is: determining the weight coefficient corresponding to each of at least one performance indicator in the target database; discretizing the at least one determined weight coefficient to determine at least one group of weight coefficient combinations; wherein each group of weight coefficient combinations is composed of a discrete weight value of at least one weight coefficient; from at least one group of weight coefficient combinations, determining the current weight coefficient combination and the first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has performed parameter configuration adjustment; determining the first parameter configuration information corresponding to the first weight coefficient combination, and determining the target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination; performing parameter configuration adjustment on the current weight coefficient combination based on the target training sample set, and determining the target parameter configuration information corresponding to the current weight coefficient combination. In this way, by discretizing the weight coefficients, it is possible to flexibly support weight coefficient combinations of different granularities; then, the optimal parameter configuration information corresponding to the weight coefficient combination that has a neighbor relationship and has been solved is selected as the target training sample set of the regression problem corresponding to the current weight coefficient combination to guide the solution process of the current weight coefficient combination, thereby not only saving the amount of calculation, accelerating the solution process, and improving the solution efficiency, but also recommending to users parameter configuration information under different weight coefficient combinations for multiple performance indicators, thereby improving practicality.

[0038] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0039] In one embodiment of the present application, see Figure 1 , which shows a flow chart of a parameter configuration adjustment method provided by an embodiment of the present application. Figure 1 As shown, the method may include:

[0040] S101: Determine a weight coefficient corresponding to at least one performance indicator in a target database.

[0041] It should be noted that the embodiments of the present application provide a parameter configuration adjustment method, specifically a multi-objective method for intelligent parameter adjustment of a database management system. The method can be applied to a parameter configuration adjustment device, or an electronic device incorporating the device. The electronic device may be, for example, a computer, a smartphone, a tablet computer, a laptop computer, a PDA, a navigation device, a wearable device, a server, and the like, and the embodiments of the present application do not specifically limit this.

[0042] It should also be noted that in the embodiments of the present application, the application scenario may be a public cloud service built for China Mobile, namely Mobile Cloud, in which the "cloud database" service of Mobile Cloud is a high-availability database service hosted by Mobile Cloud and available to tenants out of the box, including dozens of database products, such as MySQL, PostgreSQL, etc.; it may also be other database products, which are not specifically limited in the embodiments of the present application.

[0043] It should also be noted that, in the embodiment of the present application, the target database may be MySQL, PostgreSQL, Oracle, etc., or other databases, and the embodiment of the present application does not make specific limitations.

[0044] It should also be noted that in the embodiments of the present application, the target database may include information such as performance indicator values, parameter configuration information, and state variable values. Performance indicator values ​​are indicator data used to represent the database data processing performance. Performance indicators may include throughput (Transactions Per Second, TPS), average latency, etc., where throughput refers to the number of data successfully transmitted per unit time for a network, device, port, virtual circuit, or other facility, and average latency refers to the average delay between the start of a database query and the database's response. Parameter configuration information is used to set specific values ​​for each parameter of the target database. State variable values ​​are data used to represent the operating status of the target database.

[0045] S102 . Discretize the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of the at least one weight coefficient.

[0046] It should be noted that in an embodiment of the present application, after determining the weight coefficient corresponding to each of at least one performance indicators in the target database, it is necessary to discretize the at least one determined weight coefficient separately to obtain a preset number of discrete weight values ​​corresponding to each weight coefficient in at least one weight coefficient, and then obtain at least one set of weight coefficient combinations based on the legality conditions.

[0047] It should also be noted that in the embodiment of the present application, the target database may include k (k is greater than or equal to 1) performance indicators, which are respectively denoted as m1, ..., m k ; The performance index values ​​of each performance index are recorded as r1,...,r k ; The weight coefficients corresponding to each performance indicator are denoted as w1,...,w k .

[0048] S103: Determine, from at least one group of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment.

[0049] It should be noted that in an embodiment of the present application, after determining at least one set of weight coefficient combinations, it is necessary to determine a second weight coefficient combination that has undergone parameter configuration adjustment from the at least one set of weight coefficient combinations; then if there is only one discrete weight value that is different between the second weight coefficient combination and the current weight coefficient combination, it is determined that the second weight coefficient combination and the current weight coefficient combination have a neighbor relationship, and the second weight coefficient combination is used as the first weight coefficient combination.

[0050] S104: Determine first parameter configuration information corresponding to the first weight coefficient combination, and determine a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination.

[0051] It should be noted that in an embodiment of the present application, after determining the first parameter configuration information corresponding to the first weight coefficient combination, because the target parameter configuration information corresponding to the current weight coefficient combination is located around the first parameter configuration information corresponding to the first weight coefficient combination, the first parameter configuration information is added to the initial training sample set corresponding to the current weight coefficient combination to obtain the target training sample set.

[0052] In some embodiments, determining the first parameter configuration information corresponding to the first weight coefficient combination may include:

[0053] Performing performance return calculations on the multiple sets of parameter configuration information of the first weight coefficient combination to determine return values ​​corresponding to the multiple sets of parameter configuration information;

[0054] A maximum reward value is selected from the multiple reward values, and parameter configuration information corresponding to the maximum reward value is used as the first parameter configuration information.

[0055] It should be noted that, in the embodiment of the present application, the target database may include n (n is greater than or equal to 1) parameters that need to be adjusted, which are respectively denoted as b1, ..., b n ; Parameter configuration information is recorded as C, which means setting a specific value for each parameter, where C(b i ) indicates that the parameter configuration information C is parameter b i Specified concrete value (i∈{1,...,n}).

[0056] It should also be noted that, in the embodiment of the present application, the reward value (Reward) corresponding to the parameter configuration information C is recorded as a function R, and the calculation method is as follows:

[0057]

[0058] Among them, R(C) represents a function with C as a variable, that is, the weighted sum of the performance index values. The larger the value of R(C), the better C is. i Represents any performance indicator m i The corresponding weight coefficient; r i Represents any performance indicator m i The performance index value of (i∈{1,...,k}).

[0059] It should also be noted that in the embodiment of the present application, after determining the reward values ​​corresponding to multiple sets of parameter configuration information, the maximum reward value can be determined by comparing them two by two, or by sorting them to determine the maximum reward value. This embodiment of the present application does not make specific limitations.

[0060] S105: Adjust the parameter configuration of the current weight coefficient combination according to the target training sample set, and determine the target parameter configuration information corresponding to the current weight coefficient combination.

[0061] It should be noted that in the embodiment of the present application, each training sample in the determined target training sample set includes state variable values, parameter configuration information and performance indicator values, and the target parameter configuration information is obtained by inputting the current weight coefficient combination into the parameter adjustment model.

[0062] In some embodiments, the target training sample set includes at least one training sample, and each training sample includes a state variable value, parameter configuration information, and a performance indicator value.

[0063] It should be noted that in the embodiment of the present application, the state variable value is data used to represent the operating status of the target database; the parameter configuration information is used to set a specific value for each parameter of the target database; and the performance indicator value is indicator data used to represent the database data processing performance.

[0064] It should also be noted that, in the embodiment of the present application, the state variable value of the target database can be recorded as a vector I, which serves as a dependent variable for analyzing the performance indicator value after configuration C. For example, one of the state variable values ​​of MySQL InnoDB is buffer_data_reads, which records the total number of bytes read from the buffer of the InnoDB storage system.

[0065] It should also be noted that in the embodiment of the present application, the training samples can be recorded as (I, C, r1, ..., r k ), further, the multi-objective problem of intelligent parameter adjustment of database management system is based on the training samples (I, C, r1, ..., r k ) constitutes the target training sample set, and outputs performance indicators m1,...,mk The problem of optimal parameter configuration information corresponding to different weight coefficient combinations.

[0066] Accordingly, adjusting the parameter configuration of the current weight coefficient combination according to the target training sample set to determine the target parameter configuration information corresponding to the current weight coefficient combination may include:

[0067] Calculate the performance return of each current weight coefficient combination according to at least one training sample, and determine the return value corresponding to each of the at least one training sample;

[0068] Constructing a fitting function model according to the state variable value, parameter configuration information and the corresponding reward value in at least one training sample;

[0069] The optimal solution is calculated based on the fitting function model to determine the target parameter configuration information; wherein the fitting function model is used to characterize the functional relationship between the parameter configuration information, the state variable value and the return value.

[0070] It should be noted that in the embodiment of the present application, based on at least one training sample, the reward values ​​corresponding to multiple sets of parameter configuration information of the current weight coefficient combination can be calculated by formula (1). After determining the reward values ​​corresponding to each of the multiple sets of parameter configuration information, the maximum reward value can be determined by comparing them two by two, or by sorting them to determine the maximum reward value. The embodiment of the present application does not make specific limitations.

[0071] It should also be noted that, in the embodiment of the present application, the target parameter configuration information is the parameter configuration information corresponding to the maximum reward value.

[0072] It should also be noted that in the embodiments of the present application, the fitting function model can be expressed as R = f(C, I). In some embodiments, to calculate the optimal solution based on the fitting function model, a random parameter configuration information C can be first generated; the parameter configuration information C can then be applied to the target database to obtain the value of each state variable in the target database to form the state variable value I; the performance index values ​​can then be tested using the multi-threaded performance testing tool Sysbench, with the throughput performance index value being denoted as r1 and the average latency performance index value being denoted as r2; finally, a fitting search can be performed on the current weight coefficient combination according to the method in CDBTune to obtain the target parameter configuration information.

[0073] It should also be noted that in the embodiment of the present application, the target parameter configuration information corresponding to the current weight coefficient combination can be obtained according to the method in CDBTune, or the target parameter configuration information can be obtained according to other methods, without specific limitation.

[0074] In some embodiments, adjusting the parameter configuration of the current weight coefficient combination according to the target training sample set to determine the target parameter configuration information corresponding to the current weight coefficient combination may include:

[0075] Input the current weight coefficient combination into the parameter adjustment model, and output the target parameter configuration information through the parameter adjustment model;

[0076] Among them, the parameter adjustment model is obtained by training the neural network model based on the target training sample set.

[0077] It should be noted that in the embodiment of the present application, the parameter adjustment model is a data model that generates parameter adjustment data based on the current data state indicator value. The parameter adjustment model can be a deep reinforcement learning model. The parameter adjustment data is the data based on which the current parameter configuration information of the database is adjusted. The parameter adjustment data includes the adjustment direction for each parameter configuration information, and the adjustment direction can be any of increase, unchanged, and decrease. The parameter adjustment model is obtained by training the neural network model based on the target training sample set. The neural network model is cyclically trained and continuously converged through model training until the training termination condition is met. The neural network model finally trained is determined as the parameter adjustment model.

[0078] This embodiment provides a parameter configuration adjustment method, which determines the weight coefficient corresponding to each of at least one performance indicators in a target database; discretizes the at least one determined weight coefficient to determine at least one group of weight coefficient combinations; wherein each group of weight coefficient combinations is composed of a discrete weight value of at least one weight coefficient; from the at least one group of weight coefficient combinations, determines the current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment; determines first parameter configuration information corresponding to the first weight coefficient combination, and determines a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination; performs parameter configuration adjustment on the current weight coefficient combination based on the target training sample set, and determines the target parameter configuration information corresponding to the current weight coefficient combination. In this way, by discretizing the weight coefficients, it is possible to flexibly support weight coefficient combinations of different granularities; then, the optimal parameter configuration information corresponding to the weight coefficient combination that has a neighbor relationship and has been solved is selected as the target training sample set of the regression problem corresponding to the current weight coefficient combination to guide the solution process of the current weight coefficient combination, thereby not only saving the amount of calculation, accelerating the solution process, and improving the solution efficiency, but also recommending to users parameter configuration information under different weight coefficient combinations for multiple performance indicators, thereby improving practicality.

[0079] In another embodiment of the present application, see Figure 2, which shows a flow chart of another parameter configuration adjustment method provided by an embodiment of the present application. Figure 2 As shown, discretization processing is performed on each of the at least one determined weight coefficient to determine at least one set of weight coefficient combinations. The method may include:

[0080] S201 : Discretize each weight coefficient in at least one weight coefficient to obtain a preset number of discrete weight values ​​corresponding to each weight coefficient in at least one weight coefficient.

[0081] It should be noted that in the embodiment of the present application, by discretizing each weight coefficient of at least one weight coefficient, a finite number of discrete weight values ​​corresponding to each weight coefficient can be determined, thereby saving the amount of calculation.

[0082] It should also be noted that, in the embodiment of the present application, discretizing the weight coefficient means dividing the weight coefficient into a preset number of discrete weight values, wherein the weight coefficient can be divided evenly or unevenly, which is not specifically limited in the embodiment of the present application.

[0083] It should also be noted that, in the embodiment of the present application, the number of preset quantities is determined by the user himself, and the preset number can be 10, 20, 30, etc., and the embodiment of the present application does not make any specific limitation.

[0084] S202. Combining weight coefficients according to a preset number of discrete weight values ​​corresponding to each weight coefficient in at least one weight coefficient to obtain multiple groups of candidate weight coefficient combinations; wherein each candidate weight coefficient combination includes a discrete weight value for each of at least one weight coefficient.

[0085] It should be noted that in the embodiment of the present application, after determining the preset number of discrete weight values ​​corresponding to each weight coefficient, a discrete weight value is selected from the preset number of discrete weight values ​​corresponding to each weight coefficient to obtain a candidate weight coefficient combination.

[0086] S203: Perform a legality judgment on multiple groups of candidate weight coefficient combinations. If a first candidate weight coefficient combination meets a legality condition, the first candidate weight coefficient combination is determined as the weight coefficient combination to obtain at least one group of weight coefficient combinations.

[0087] It should be noted that in an embodiment of the present application, at least one group of weight coefficient combinations is determined by performing a legitimacy judgment on multiple groups of candidate weight coefficient combinations; wherein the first candidate weight coefficient combination refers to any one candidate weight coefficient combination among the multiple groups of candidate weight coefficient combinations.

[0088] In some embodiments, the method may further include:

[0089] If the sum of at least one discrete weight value included in the first candidate weight coefficient combination is equal to 1, it is determined that the first candidate weight coefficient combination meets the legality condition.

[0090] It should be noted that, in the embodiment of the present application, a function W can be set, and this function is for each w i Specify a specific weight value W(w i ), when formula (2) is satisfied, it means that the first candidate weight coefficient combination meets the legality condition.

[0091] W(w1)+W(w2)+...+W(w k )=1.0 (2)

[0092] Among them, w i Represents any performance indicator m i The corresponding weight coefficient (i∈{1,...,k}).

[0093] In another embodiment of the present application, see Figure 3 , which shows a flow chart of another parameter configuration adjustment method provided by an embodiment of the present application. The method may include:

[0094] S301: Determine a weight coefficient corresponding to at least one performance indicator in a target database.

[0095] S302 . Discretize the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of the at least one weight coefficient.

[0096] S303: Determine a second weight coefficient combination having a preset mark from at least one group of weight coefficient combinations; wherein the preset mark is used to indicate that parameter configuration adjustment has been performed on the second weight coefficient combination.

[0097] S304: If there is only one discrete weight value different between the second weight coefficient combination and the current weight coefficient combination, determine that the second weight coefficient combination and the current weight coefficient combination are neighbors, and use the second weight coefficient combination as the first weight coefficient combination.

[0098] S305: Determine first parameter configuration information corresponding to the first weight coefficient combination, and determine a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination.

[0099] S306: Adjust the parameter configuration of the current weight coefficient combination according to the target training sample set, and determine the target parameter configuration information corresponding to the current weight coefficient combination.

[0100] It should be noted that in the embodiment of the present application, steps S301, S302, S305, and S306 correspond to steps S101, S102, S104, and S105 in the aforementioned embodiment, respectively, and for the sake of brevity, they are not further described here. In addition, steps S303 and S304 are specific implementations of step S103 in the aforementioned embodiment, and are described in detail below.

[0101] It can be understood that, for step S303, firstly, the second weight coefficient combination for which parameter configuration adjustment has been performed is determined, and then it is determined whether there is a neighbor relationship between the second weight coefficient combination and the current weight coefficient combination.

[0102] It should be noted that in an embodiment of the present application, a list can be set, which is an empty list when initialized. When a weight coefficient combination has performed parameter configuration adjustment, it is added to this list to indicate that the weight coefficient combination has a preset mark, and the weight coefficient combination that has performed parameter configuration adjustment is recorded as the second weight coefficient combination. Then, if the current weight coefficient combination is compared with the second weight coefficient combination in this list and it is found that the weight coefficient combination already exists in the list, it is considered that the current weight coefficient combination has performed parameter configuration adjustment; when the comparison is found that the weight coefficient combination does not exist in the list, it is considered that the current weight coefficient combination has not performed parameter configuration adjustment.

[0103] It can also be understood that for step S306, after determining the second weight coefficient combination, it is necessary to further determine whether there is only one discrete weight value that is different between the second weight coefficient combination and the current weight coefficient combination. If there is only one discrete weight value that is different, it is determined that the second weight coefficient combination and the current weight coefficient combination have a neighbor relationship, and the second weight coefficient combination is used as the first weight coefficient combination; if there are two or more discrete weight values ​​that are different, it is determined that the second weight coefficient combination and the current weight coefficient combination do not have a neighbor relationship.

[0104] Furthermore, in an embodiment of the present application, if there is no weight coefficient combination with which the current weight coefficient combination has a neighboring relationship, then parameter configuration adjustment is performed according to the initial training sample set corresponding to the current weight coefficient combination to determine the target parameter configuration information corresponding to the current weight coefficient combination.

[0105] It should also be noted that, in an embodiment of the present application, if there is no weight coefficient combination that has a neighboring relationship with the current weight coefficient combination, then the initial training sample set of the current weight coefficient combination is an empty sample set at the time of initialization; if there is a first weight coefficient combination that has a neighboring relationship with the current weight coefficient combination and has performed parameter configuration adjustment, then the initial training sample set of the current weight coefficient combination has the first parameter configuration information corresponding to the first weight coefficient combination, and the initial training sample set with the first parameter configuration information added is determined as the target training sample set.

[0106] In some embodiments, determining, from at least one set of weight coefficient combinations, a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment may further include:

[0107] Dividing at least one group of weight coefficient combinations to determine a first set of weight coefficient combinations and a second set of weight coefficient combinations; wherein the first set of weight coefficient combinations includes at least one group of weight coefficient combinations for which parameter configuration adjustment has not been performed, and the second set of weight coefficient combinations includes at least one group of weight coefficient combinations for which parameter configuration adjustment has been performed;

[0108] In the second weight coefficient combination set, if there is only one discrete weight value that is different between the second weight coefficient combination and the current weight coefficient combination, then it is determined that the second weight coefficient combination and the current weight coefficient combination are in a neighbor relationship, and the second weight coefficient combination is used as the first weight coefficient combination;

[0109] The second weight coefficient combination is any one in the second weight coefficient combination set.

[0110] In another embodiment of the present application, see Figure 4 , which shows a flow chart of another parameter configuration adjustment method provided by an embodiment of the present application. The method may include:

[0111] S401: Determine a weight coefficient corresponding to at least one performance indicator in a target database.

[0112] S402 . Discretize the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of the at least one weight coefficient.

[0113] S403: Determine, from at least one group of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment.

[0114] S404: Determine first parameter configuration information corresponding to the first weight coefficient combination, and determine a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination.

[0115] S405: Adjust the parameter configuration of the current weight coefficient combination according to the target training sample set, and determine the target parameter configuration information corresponding to the current weight coefficient combination.

[0116] S406: Add a preset mark to the current weight coefficient combination to indicate that parameter configuration adjustment has been performed on the current weight coefficient combination.

[0117] S407: After parameter configuration adjustment has been performed on all of at least one set of weight coefficient combinations, target parameter configuration information corresponding to each of at least one set of weight coefficient combinations is displayed to recommend the target parameter configuration information.

[0118] It should be noted that in the embodiment of the present application, steps S401, S402, S403, S404, and S405 correspond to steps S101, S102, S103, S104, and S105 in the aforementioned embodiment, respectively, and for the sake of brevity, they are not repeated here. In addition, steps S406 and S407 are a specific implementation method after determining the target parameter configuration information corresponding to the current weight coefficient combination, and are described in detail below.

[0119] It can be understood that, for step S406, after determining the target parameter configuration information corresponding to the current weight coefficient combination, it indicates that the weight coefficient combination has been subjected to parameter configuration adjustment, and thus a preset mark is added thereto.

[0120] In some embodiments, after adjusting the parameter configuration of the current weight coefficient combination according to the target training sample set, the method may further include:

[0121] Adding a preset mark to the current weight coefficient combination, and moving the current weight coefficient combination from the first weight coefficient combination set to the second weight coefficient combination set;

[0122] If there are weight coefficient combinations in the first weight coefficient combination set, one of them is used as the current weight coefficient combination, and the step of determining the first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has performed parameter configuration adjustment is continued until the first weight coefficient combination set is empty.

[0123] It should be noted that, in the embodiment of the present application, the first set of weight coefficient combinations includes at least one set of weight coefficient combinations that have not undergone parameter configuration adjustment, and the second set of weight coefficient combinations includes at least one set of weight coefficient combinations that have undergone parameter configuration adjustment.

[0124] It can be understood that, for step S407, after all weight coefficient combinations have been subjected to parameter configuration adjustment, the target parameter configuration information corresponding to at least one set of weight coefficient combinations is displayed to the user for selection and adoption.

[0125] In summary, this embodiment provides a parameter configuration adjustment method, specifically a multi-objective method for intelligent parameter adjustment of a database management system. This method can be applied to China Mobile's public cloud service, Mobile Cloud. Mobile Cloud's "Cloud Database" service is a high-availability database service hosted by Mobile Cloud and available to tenants out-of-the-box. It includes dozens of database products, such as MySQL and PostgreSQL. Tenants of each cloud database product typically handle different types of data access loads, resulting in different requirements for database parameter configuration information. Adjusting and optimizing database parameter configuration information for tenants' specific data access loads requires the services of an experienced database administrator (DBA). However, with the rapid growth of tenants for mobile cloud database products, the demand for DBAs has increased dramatically, and the labor cost of providing such tuning services is high for mobile cloud. Furthermore, due to the wide variety of tenants serving public cloud services, different tenants have different requirements for balancing tuning performance indicators. Therefore, providing only one set of parameter configuration information optimization solutions often cannot meet the needs of customers. This solution aims to automatically provide multiple sets of parameter configuration information optimization solutions, allowing technical personnel to flexibly choose from them during communication with tenants, providing more robust practicality.

[0126] In a specific embodiment, for the parameter configuration adjustment method described in the above embodiment, it is assumed that the target for parameter configuration adjustment of the database management system includes k (k is greater than or equal to 1) performance indicators, which are respectively denoted as m1, ..., m k After the user's data access business test, the performance index value of each performance index is recorded as r1,...,r k ; The weight coefficients corresponding to each performance indicator are denoted as w1,...,w k ; There are n (n is greater than or equal to 1) adjustable parameters, which are denoted as b1,...,b n ; Parameter configuration information C represents setting specific values ​​for each parameter, using C(b i ) indicates that the parameter configuration information C is parameter b iThe specific value specified (i∈{1,...,n}); the reward value corresponding to the parameter configuration information C is expressed as a function R with C as the variable, as shown in formula (1). The larger the value of R(C), the better C is. The internal state characteristics of the DBMS, as the dependent variable for analyzing the performance indicator value after configuring C, are recorded as the state variable value I (for example, one of the state variable values ​​of MySQL InnoDB is buffer_data_reads, which records the total number of bytes read from the buffer of the InnoDB storage system). Therefore, the multi-objective problem of intelligent parameter adjustment of the database management system is to solve the problem based on the training samples (I, C, r1,..., r k ) constitutes the target training sample set, and outputs performance indicators m1,...,m k The problem of optimal parameter configuration information corresponding to different weight coefficient combinations can be specifically addressed by the following steps:

[0127] Step 1: Set w1,...,w k Each w i Divide evenly into p+1 parts.

[0128] For example: if w1 takes the value [0,1] and p = 10, then [0,1] will be divided into {0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0}.

[0129] A combination of weight coefficients W, expressed as each w i Specify a specific weight value W(w i ), which is legal must satisfy formula (2). The set of all legal weight coefficient combinations is denoted as W P .

[0130] Step 2: From the weight coefficient combination set W P Select a combination that has not been tried now , execute W now For the target weight coefficient combination (i.e. the current weight coefficient combination), set the DBMS configuration parameter b i The optimized configuration is divided into the following steps:

[0131] Step 2-1: Add the optimal parameter configuration information of the weight coefficient combination of the tried parameter configuration adjustment (the parameter configuration information with the largest R value among all the tried parameter configuration information is the optimal configuration) to the initial training sample set of this solution. The added rule is: Let W h It is a weight coefficient configuration that has been tried (i.e., the second weight coefficient combination), and W is obtained through CDBTune regression prediction. h The corresponding optimal parameter configuration information C h(ie, the first parameter configuration information); if W now and W h There is only one discrete weight value that is different, then W h It's W now neighbors; if W h W now Neighbors, then W h The corresponding optimal parameter configuration information C h Add to W now The training sample set T now (i.e. the initial training sample set).

[0132] Step 2-2, with W now for m1,...,m k The weight coefficient combination defines the reward function, with T now Perform a fitting search for the initial training sample set, obtain the optimal parameter configuration, and record it.

[0133] Initial training sample set T now The form of a single training sample in is: (I,C,r1,...,r k )

[0134] The performance reward function is designed as follows:

[0135] R=W now (w1)*r1+W now (w2)*r2+...+W now (w k )*r k (3)

[0136] The function to be fitted is: R = f(C, I)

[0137] For example, using CDBTune, we adopt a reinforcement learning method based on a deep learning model, with R as the reward function and T as the reward function. now For the initial training sample set, iterative execution:

[0138] 1) Generate a random parameter configuration information C;

[0139] 2) Apply C to the target database, obtain the value of each state variable in the target database, form the state variable value I, use Sysbench to test the performance index value, record the performance index value of throughput as r1, and the performance index value of average delay as r2;

[0140] 3) Use the method in the related technology CDBTune to design a deep neural network model, use 2) to form samples (C, I, r1, r2), and train the deep neural network.

[0141] 4) Repeat 1) to 3) until the feedback training in 3) is completed.

[0142] The above process is for the current weight coefficient combination W now After the fitting search is completed, the recommended optimal parameter configuration information is recorded as C now (i.e. target parameter configuration information), and W now The combination of weight coefficients that has been tried is marked (ie, the second combination of weight coefficients).

[0143] Step 2-3, if W P If there are any untried weight coefficient combinations, go to step 2-1; otherwise, go to step 3.

[0144] Step 3: W P Each weight coefficient combination W in , and the optimal parameter configuration information (i.e., target parameter configuration information) C corresponding to W, are displayed to the user.

[0145] In summary, the technical solution of the embodiment of the present application proposes a method framework, specifically, it transforms the intelligent parameter adjustment method of multi-objective fixed weights into an intelligent parameter adjustment method of multi-objective multi-weights. Here, by discretizing the weight coefficients to divide the number p, the weight coefficients of different granularities are flexibly supported to consider the needs; and the historical solution results corresponding to a weight coefficient combination of multi-objective intelligent parameter adjustment are added to the subsequent solution process of the new weight coefficient combination, and the historical solution results are used as training samples to guide the solution of the new weight coefficient combination; but not all historical solution results are used to guide the fitting search of the new weight coefficient combination, but the neighbor relationship of the weight coefficient combination is defined, and the historical results with the neighbor relationship are used as training sample sets to guide the solution of the new weight coefficient combination; a weight coefficient combination represents the user's degree of attention to each performance indicator. The embodiment of the present application displays the optimal parameter configuration information corresponding to all the listed weight coefficient combinations to the user for selection and adoption, which can achieve multi-objective optimization recommendation.

[0146] This embodiment provides a parameter configuration adjustment method. The specific implementation of the aforementioned embodiment is elaborated in detail based on the aforementioned embodiment. It can be seen that according to the technical solution of the aforementioned embodiment, on the one hand, the related technology can only recommend DBMS parameter configuration information suggestions corresponding to one weight coefficient combination for multiple performance indicators to the user, while the present solution can recommend multiple parameter configuration information suggestions for multiple performance indicators under different weight coefficient combinations, thereby providing the user with more DBMS parameter configuration information candidates; on the other hand, the solution process of the related technology does not have a mechanism for using the DBMS parameter configuration information recommendation results of its historical solution process to construct samples to guide the new solution process. The present solution defines the neighbor relationship between different weight coefficient combinations of multiple objectives, selects the parameter configuration information corresponding to the weight coefficient combination with the neighbor relationship and has been solved, as the fitting sample of the regression problem corresponding to the new weight coefficient combination, and guides the new solution process. This not only saves computing time in sample construction, but also can continue to recommend better parameter configuration information based on historical experience by guiding the new solution process from the sample.

[0147] In another embodiment of the present application, see Figure 5 , which shows a schematic diagram of the structure of a parameter configuration adjustment device 50 provided in an embodiment of the present application. Figure 5 As shown, the parameter configuration adjustment device 50 may include a determination unit 501, a discretization unit 502, and an adjustment unit 503, wherein:

[0148] A determining unit 501 is configured to determine a weight coefficient corresponding to at least one performance indicator in a target database;

[0149] The discretization unit 502 is configured to discretize the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of the at least one weight coefficient;

[0150] The determining unit 501 is further configured to determine, from at least one set of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment; and determine first parameter configuration information corresponding to the first weight coefficient combination, and determine a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination;

[0151] The adjusting unit 503 is configured to perform parameter configuration adjustment on the current weight coefficient combination according to the target training sample set, and determine target parameter configuration information corresponding to the current weight coefficient combination.

[0152] In some embodiments, the discretization unit 502 is further configured to discretize each weight coefficient in at least one weight coefficient separately to obtain a preset number of discrete weight values ​​corresponding to each weight coefficient in at least one weight coefficient; and to combine weight coefficients according to the preset number of discrete weight values ​​corresponding to each weight coefficient in at least one weight coefficient to obtain multiple groups of candidate weight coefficient combinations; wherein each candidate weight coefficient combination includes a discrete weight value for each of at least one weight coefficient; and to perform legitimacy judgment on multiple groups of candidate weight coefficient combinations. If the first candidate weight coefficient combination meets the legitimacy condition, the first candidate weight coefficient combination is determined as the weight coefficient combination to obtain at least one group of weight coefficient combinations.

[0153] In some embodiments, the discretization unit 502 is further configured to determine that the first candidate weight coefficient combination meets the legality condition if the sum of at least one discrete weight value included in the first candidate weight coefficient combination is equal to 1.

[0154] In some embodiments, the determination unit 501 is further configured to determine a second weight coefficient combination with a preset mark from at least one group of weight coefficient combinations; wherein the preset mark is used to indicate that the second weight coefficient combination has performed parameter configuration adjustment; and if there is only one discrete weight value different between the second weight coefficient combination and the current weight coefficient combination, it is determined that the second weight coefficient combination and the current weight coefficient combination have a neighbor relationship, and the second weight coefficient combination is used as the first weight coefficient combination.

[0155] In some embodiments, as Figure 5 As shown, the parameter configuration adjustment device 50 may include a calculation unit 504, which is configured to perform performance return calculations on multiple groups of parameter configuration information of the first weight coefficient combination, determine the return values ​​corresponding to each of the multiple groups of parameter configuration information; and select the maximum return value from the multiple return values, and use the parameter configuration information corresponding to the maximum return value as the first parameter configuration information.

[0156] In some embodiments, the target training sample set includes at least one training sample, and each training sample includes a state variable value, parameter configuration information, and a performance indicator value; accordingly, the determination unit 501 is further configured to perform performance return calculations on the current weight coefficient combination according to the at least one training sample, and determine the return value corresponding to each of the at least one training sample; and construct a fitting function model based on the state variable value, parameter configuration information, and the corresponding return value in the at least one training sample; and calculate the optimal solution based on the fitting function model to determine the target parameter configuration information; wherein the fitting function model is used to characterize the functional relationship between the parameter configuration information, the state variable value, and the return value.

[0157] In some embodiments, as Figure 5As shown, the parameter configuration adjustment device 50 may include an input unit 505, which is configured to input the current weight coefficient combination into a parameter adjustment model and output target parameter configuration information through the parameter adjustment model; wherein the parameter adjustment model is obtained by training the neural network model based on the target training sample set.

[0158] In some embodiments, the adjusting unit 503 is further configured to add a preset mark to the current weight coefficient combination to indicate that the parameter configuration adjustment has been performed on the current weight coefficient combination.

[0159] In some embodiments, the adjustment unit 503 is further configured to display the target parameter configuration information corresponding to at least one set of weight coefficient combinations after all the at least one set of weight coefficient combinations have undergone parameter configuration adjustment, so as to recommend the target parameter configuration information.

[0160] It is understood that in this embodiment, a "unit" can be a portion of a circuit, a portion of a processor, a portion of a program or software, etc., and can also be a module or a non-modular system. Furthermore, the various components in this embodiment can be integrated into a single processing unit, or each unit can exist physically separately, or two or more units can be integrated into a single unit. The aforementioned integrated units can be implemented in the form of hardware or software functional modules.

[0161] If the integrated unit is implemented as a software functional module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, or the portion that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0162] Therefore, this embodiment provides a computer storage medium storing a computer program. When the computer program is executed by at least one processor, the steps of the camouflage method described in any one of the aforementioned embodiments are implemented.

[0163] Based on the above-mentioned composition of the parameter configuration adjustment device 50 and the computer storage medium, see Figure 6, which shows a schematic diagram of the structure of an electronic device 60 provided in an embodiment of the present application. Figure 6 As shown, the electronic device 60 may include: a communication interface 601, a memory 602 and a processor 603; each component is coupled together via a bus system 604. It is understood that the bus system 604 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 604 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 6 Various buses are labeled as bus system 604. Among them, the communication interface 601 is used to receive and send signals in the process of sending and receiving information between other external network elements;

[0164] Memory 602, used to store computer programs that can be run on processor 603;

[0165] The processor 603 is configured to, when running the computer program, execute:

[0166] Determine a weight coefficient corresponding to at least one performance indicator in the target database;

[0167] Discretizing the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of the at least one weight coefficient;

[0168] Determine, from at least one set of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment;

[0169] Determining first parameter configuration information corresponding to the first weight coefficient combination, and determining a target training sample set based on the first parameter configuration information and an initial training sample set corresponding to the current weight coefficient combination;

[0170] The parameter configuration of the current weight coefficient combination is adjusted according to the target training sample set to determine the target parameter configuration information corresponding to the current weight coefficient combination.

[0171] It is understood that the memory 602 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced synchronous DRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DRRAM). The memory 602 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0172] Processor 603 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits or software instructions in processor 603. The above processor 603 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in memory 602, and processor 603 reads the information in memory 602 and, in conjunction with its hardware, completes the steps of the above method.

[0173] It is understood that the embodiments described herein may be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or a combination thereof.

[0174] For software implementation, the techniques described herein can be implemented by modules (e.g., procedures, functions, etc.) that perform the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or external to the processor.

[0175] Optionally, as another embodiment, the processor 603 is further configured to execute the method described in any one of the aforementioned embodiments when running the computer program.

[0176] In yet another embodiment of the present application, see Figure 7 , which shows a schematic diagram of the composition structure of another electronic device 60 provided in an embodiment of the present application. Figure 7 As shown, the electronic device 60 at least includes the parameter configuration adjustment device 50 according to any one of the aforementioned embodiments.

[0177] In an embodiment of the present application, for the electronic device 60, by discretizing the weight coefficients, it is possible to flexibly support weight coefficient combinations of different granularities; then, the optimal parameter configuration information corresponding to the weight coefficient combination that has a neighbor relationship and has been solved is selected as the target training sample set of the regression problem corresponding to the current weight coefficient combination to guide the solution process of the current weight coefficient combination, thereby not only saving the amount of calculation, accelerating the solution process, and improving the solution efficiency, but also recommending to the user parameter configuration information under different weight coefficient combinations for multiple performance indicators, thereby improving practicality.

[0178] The above description is merely a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application.

[0179] It should be noted that, in this application, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0180] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0181] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0182] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0183] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0184] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A parameter configuration adjustment method, characterized in that: The method comprises: Determine a weight coefficient corresponding to at least one performance indicator in the target database; Discretizing the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of each of the at least one weight coefficient; Determine, from the at least one set of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment; Determining first parameter configuration information corresponding to the first weight coefficient combination, and determining a target training sample set based on the first parameter configuration information and an initial training sample set corresponding to the current weight coefficient combination; Adjusting the parameter configuration of the current weight coefficient combination according to the target training sample set to determine target parameter configuration information corresponding to the current weight coefficient combination; The step of determining, from the at least one set of weight coefficient combinations, a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment, includes: Determining, from the at least one set of weight coefficient combinations, a second weight coefficient combination having a preset mark, wherein the preset mark is used to indicate that parameter configuration adjustment has been performed on the second weight coefficient combination; If there is only one discrete weight value different between the second weight coefficient combination and the current weight coefficient combination, it is determined that the second weight coefficient combination and the current weight coefficient combination have a neighbor relationship, and the second weight coefficient combination is used as the first weight coefficient combination.

2. The method according to claim 1, characterized in that The discretizing the determined at least one weight coefficient to determine at least one set of weight coefficient combinations includes: performing discretization processing on each weight coefficient of the at least one weight coefficient to obtain a preset number of discrete weight values ​​corresponding to each weight coefficient of the at least one weight coefficient; Combining weight coefficients according to a preset number of discrete weight values ​​corresponding to each weight coefficient of the at least one weight coefficient to obtain a plurality of candidate weight coefficient combinations; wherein each candidate weight coefficient combination includes a discrete weight value of each of the at least one weight coefficient; The legitimacy of the multiple groups of candidate weight coefficient combinations is judged. If a first candidate weight coefficient combination meets the legitimacy condition, the first candidate weight coefficient combination is determined as the weight coefficient combination to obtain the at least one group of weight coefficient combinations.

3. The method according to claim 2, characterized in that The method further comprises: If the sum of at least one discrete weight value included in the first candidate weight coefficient combination is equal to 1, it is determined that the first candidate weight coefficient combination meets the legality condition.

4. The method according to claim 1, wherein The determining the first parameter configuration information corresponding to the first weight coefficient combination includes: Performing performance return calculations on each of the multiple sets of parameter configuration information of the first weight coefficient combination to determine a return value corresponding to each of the multiple sets of parameter configuration information; A maximum reward value is selected from multiple reward values, and parameter configuration information corresponding to the maximum reward value is used as the first parameter configuration information.

5. The method according to claim 1, characterized in that The target training sample set includes at least one training sample, each training sample includes a state variable value, parameter configuration information and a performance indicator value; Accordingly, the adjusting the parameter configuration of the current weight coefficient combination according to the target training sample set to determine the target parameter configuration information corresponding to the current weight coefficient combination includes: Performing performance reward calculations on the current weight coefficient combination according to the at least one training sample, and determining reward values ​​corresponding to the at least one training sample; Constructing a fitting function model according to the state variable value, parameter configuration information and respective corresponding reward values ​​in the at least one training sample; An optimal solution is calculated according to the fitting function model to determine the target parameter configuration information; wherein the fitting function model is used to characterize the functional relationship between the parameter configuration information, the state variable value and the reward value.

6. The method according to claim 1, characterized in that The adjusting the parameter configuration of the current weight coefficient combination according to the target training sample set to determine target parameter configuration information corresponding to the current weight coefficient combination includes: Inputting the current weight coefficient combination into a parameter adjustment model, and outputting the target parameter configuration information through the parameter adjustment model; The parameter adjustment model is obtained by training the neural network model based on the target training sample set.

7. The method according to any one of claims 1 to 6, characterized in that After determining the target parameter configuration information corresponding to the current weight coefficient combination, the method further includes: A preset mark is added to the current weight coefficient combination to indicate that parameter configuration adjustment has been performed on the current weight coefficient combination.

8. The method according to claim 7, characterized in that The method further comprises: After parameter configuration adjustment has been performed on all of the at least one set of weight coefficient combinations, target parameter configuration information corresponding to each of the at least one set of weight coefficient combinations is displayed to recommend the target parameter configuration information.

9. A parameter configuration adjustment device, characterized in that: The parameter configuration adjustment device includes a determination unit, a discretization unit and an adjustment unit; wherein: The determining unit is configured to determine a weight coefficient corresponding to each of at least one performance indicator in the target database; The discretization unit is configured to discretize the at least one determined weight coefficient to determine at least one set of weight coefficient combinations; wherein each set of weight coefficient combinations is composed of a discrete weight value of each of the at least one weight coefficients; The determining unit is further configured to determine, from the at least one set of weight coefficient combinations, a current weight coefficient combination and a first weight coefficient combination that has a neighbor relationship with the current weight coefficient combination and has undergone parameter configuration adjustment; and determine first parameter configuration information corresponding to the first weight coefficient combination, and determine a target training sample set based on the first parameter configuration information and the initial training sample set corresponding to the current weight coefficient combination; The adjustment unit is configured to adjust the parameter configuration of the current weight coefficient combination according to the target training sample set, and determine the target parameter configuration information corresponding to the current weight coefficient combination; The determination unit is further configured to determine a second weight coefficient combination with a preset mark from the at least one group of weight coefficient combinations; wherein the preset mark is used to indicate that the second weight coefficient combination has performed parameter configuration adjustment; and if there is only one discrete weight value different between the second weight coefficient combination and the current weight coefficient combination, it is determined that the second weight coefficient combination and the current weight coefficient combination have a neighbor relationship, and the second weight coefficient combination is used as the first weight coefficient combination.

10. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein: The memory is used to store a computer program that can be run on the processor; The processor is configured to execute the method according to any one of claims 1 to 8 when running the computer program.

11. A computer storage medium, characterized in that The computer storage medium stores a computer program, and when the computer program is executed by at least one processor, the method according to any one of claims 1 to 8 is implemented.

Images