CN117520385B - Database query optimization method based on exploration value and query cost - Google Patents
Database query optimization method based on exploration value and query cost Download PDFInfo
- Publication number
- CN117520385B CN117520385B CN202410018425.9A CN202410018425A CN117520385B CN 117520385 B CN117520385 B CN 117520385B CN 202410018425 A CN202410018425 A CN 202410018425A CN 117520385 B CN117520385 B CN 117520385B
- Authority
- CN
- China
- Prior art keywords
- query
- plan
- cost
- database
- execution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000009826 distribution Methods 0.000 claims abstract description 24
- 238000012216 screening Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010801 machine learning Methods 0.000 abstract description 10
- 230000007547 defect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2453—Query optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2453—Query optimisation
- G06F16/24534—Query rewriting; Transformation
- G06F16/24542—Plan optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Operations Research (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a database query optimization method based on exploration value and query cost, which is characterized in that future query template distribution is inferred through load prediction, and the query cost of each query is estimated by carrying out base estimation through machine learning. The invention optimizes the strategy of only carrying out cost estimation on the current query by the traditional learning, so that the query estimation is combined with the current and the future to select the optimal query execution plan. The definition of the exploration value provided by the query method enables the query plan of the current query to be capable of being represented for the value of the future predicted query, and further provides deeper insight and decision support for database query optimization.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a database query optimization method based on exploration values and query costs.
Background
Conventional database query optimization techniques typically rely on a query plan generator and a cost estimator. The performance of these methods is limited by the quality of query plan generation rules and statistics, especially in large data and complex query environments, rule-based erroneous statistics may cause the query optimizer to generate suboptimal query plans, thereby affecting query execution efficiency. Furthermore, conventional optimizers are static and do not have learning, resulting in their inability to accommodate dynamically changing query loads and data distributions. Accordingly, there is a need to provide a more efficient method to optimize database queries.
In the field of database query optimization, machine learning technology has been successfully used in query optimization at present, for example, a neural network-based radix estimation method can provide more accurate radix estimation, and can enable an optimizer to have learning property by learning the radix of historical queries, so that the optimizer can select a better execution plan for future queries, the execution efficiency of the queries is improved, and the resource consumption of a system is reduced. However, this method only satisfies the current optimization, i.e., only guarantees the current query execution optimization when selecting the execution plan, and does not consider the execution performance of future queries. In real environments, such as cloud environments, it is often desirable that the overall query load perform optimally. For example, if a query plan of a current query, while suboptimal, is reused in future queries, the value of the query plan is far beyond the performance in the current model, because executing the suboptimal query may collect more valid historical data, thereby improving the execution efficiency of the future query. Thus, there is a great need for a query optimization technique that is known to future loads.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a database query optimization method based on exploration value and query cost, so that query optimization not only considers the current query, but also is more in the future, the combination of prediction and learning is realized, the future query template distribution is inferred through load prediction, the base estimation is carried out through machine learning, and the query cost of each query is estimated.
The invention aims to overcome the defect that the traditional database query optimizer does not have learning, and aims to solve the defect that the current machine learning-based optimizer ignores future queries, and provides a database query optimization technology based on exploration values and query costs.
The technical scheme adopted by the invention for solving the technical problems comprises the following specific steps:
Step 1: an execution engine in a database stores an execution plan of historical execution and real base data corresponding to the execution plan into a cache pool, and then carries out offline training and modeling on the data in the cache pool by using a regression model to obtain a base model M;
Step 2: the method comprises the steps that an optimizer obtains current query sentences from a database management system, the optimizer generates all corresponding query plans based on the current query sentences, carries out base estimation on all query plans by utilizing a base model M trained in the step 1 and a base estimation module of a database, and carries out cost estimation on each execution plan by adopting a cost estimation module of the database, so that estimated cost of each query plan is obtained;
Step 3: the method comprises the steps that records and logs of query operations are executed through a database management system, historical query of the database management system is obtained, the historical query comprises codes and templates of query sentences executed in a historical mode, time spent for executing the query operations and result data returned by the query operations, and future query prediction is conducted through the historical query to obtain query template distribution P;
Step 4: according to all the query plans obtained in the step 2, decomposing each execution plan into a plurality of sub-execution plans, and respectively estimating the lifting degree V (V 1,v2,…,vi,…,vn) of each sub-plan node on each query template, wherein V i refers to the lifting degree of the sub-plan node i on the query template p i, and n is the number of the query templates;
Step 5: combining the query template distribution P (P 1,p2,…,pi,…,pn) obtained in the step 3 and the lifting degree V (V 1,v2,…,vi,…,vn) obtained in the step 4, obtaining the exploration value of each sub-Plan node through , wherein is the exploration value, plan refers to each sub-Plan node, and accumulating the exploration values of all sub-Plan nodes on the execution Plan as the cost estimation to obtain the exploration value of the execution Plan, namely:
。
Wherein Query is an execution plan;
Step 6: and screening all the query plans according to the optimization conditions to obtain an optimal query plan, namely selecting an execution plan with the maximum exploration value under the condition of ensuring that the cost of the selected execution plan is lower than a given threshold value, and sending the query plan to an execution engine.
The regression model employs a locally weighted regression model LWPR.
In the step 3, the queries are generated based on the query templates, so that the prediction of the future query is equivalent to the distribution of the query templates, and the query template distribution P (P 1,p2,…,pi,…,pn) is obtained by using a Markov model, wherein P i refers to the query templates to which the next query may obey.
The predicting step of the future query is: assuming that the query is loaded with n query templates, the time when the next query statement arrives is set as t, the historical query template set before the time t is { X t-1,Xt-2,Xt-3,...,Xt-m }, wherein X t-1 is the query template at the time t-1, and the distribution of the query templates at the time t is calculated to be P (P 1,p2,…,pi,…,pn) by using the trained m-order Markov model and the historical query.
The step of calculating the lifting degree V of a certain sub-plan node to each query template is as follows:
V (V 1,v2,…,vi,…,vn) wherein V i is calculated from the latest k historical queries of the sub-plan node and the ith possible obeying query template p i by the LWPR (Local Weighted Projection Regression) model, LWPR model being a machine learning algorithm for nonlinear regression and function approximation; specifically, the calculation process of each v i is as follows:
Let the history query of the i-th query template p i be , where represents the last k history queries of the query template p i, and the LWPR model has M receptive fields RECEPTIVE FIELD (RF) (RF 1,RF2,…,RFM), and the center point corresponding to the M receptive fields is C (C 1,c2,…cM); firstly, calculating the distribution P i of the RF center point C by using Q i, and then, calculating the lifting degree V i,Vi of each linear model through the linear model of each RF in the LWPR model to describe the prediction uncertainty of each linear model of each RF; finally is obtained.
And screening all the inquiry plans according to the optimization conditions, wherein the optimization conditions are as follows:
Wherein denotes maximizing the maximum value of the exploration value of the selected execution plan, denotes the query cost of the query plan, denotes the query cost of the optimal execution plan selected by the optimizer, denotes that the performance of the current execution plan is not lower than times of the current optimal estimated cost, and the optimization condition is that the query plan with the maximum exploration value is selected as the query plan to be executed in all query plans with the query cost less than or equal to the query cost of the optimal execution plan selected by the optimizer.
The invention has the beneficial effects that:
and (3) personalized optimization: a personalized query optimization is achieved by identifying, through the LWPR model, which query plans are often used for a portion of multiple queries. This means that the system can better adapt to the execution mode of different queries, improving the query performance.
The learning ability: the machine learning regression model is used for training a historical query plan and a query base, and a base model which is used for incremental learning along with data updating is established.
Load prediction: the future query templates are predicted through the historical query template distribution, and the value of the current query on the future predicted query templates is calculated so as to select the query plan optimal for the future query, thereby reducing the uncertainty of query execution and improving the performance.
Exploring value: the method optimizes the query by a load prediction and machine learning method and provides a concept of exploration value, so that the target of query optimization is not limited to the query cost of each real-time query, and the learning and prediction modes are realized by considering some connection operations possibly needed by the future query and the value of the current query to the future query.
Real-time performance: the machine learning model can continuously learn according to the provided historical query plan, and the system can dynamically optimize according to the new execution condition, so that the real-time update of the query is ensured.
Resource saving: the method can more accurately select the query execution plan, and the system can more effectively utilize hardware resources, thereby reducing resource waste and improving performance and efficiency.
The query method of the invention uses the technology of combining load prediction and machine learning, optimizes the strategy that the traditional learning only carries out cost estimation on the current query, and enables the query estimation to select the optimal query execution plan for the combination of the future and the current. The definition of the exploration value provided by the query method enables the query plan of the current query to be capable of being represented for the value of the future predicted query, and further provides deeper insight and decision support for database query optimization.
Drawings
FIG. 1 is a flowchart illustrating a database query optimization technique based on exploration value and query cost according to the present invention.
Fig. 2 is a diagram showing an example of a frame structure of the present invention.
FIG. 3 is an exemplary diagram of the calculation of probability distribution P i by historical queries in accordance with the present invention.
FIG. 4 is a graph illustrating the degree of elevation of each RF obtained by LWPR model in accordance with the present invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
A database query optimization method based on exploration value and query cost comprises the following specific steps: the execution engine in the database stores the data of historical execution into a cache pool, and then carries out offline training and modeling on the data in the cache pool by using a regression model to obtain a base model M; the optimizer acquires current query sentences from the database management system, generates all corresponding query plans based on the current query sentences, carries out base estimation on all query plans by utilizing the base model M trained in the step 1, and finally carries out cost estimation on the base estimation result by using the traditional cost model to obtain query cost estimation of each query plan; acquiring historical query template distribution of a database management system, and predicting future query template distribution; according to all the obtained query plans, calculating the lifting degree V of the query plan X for executing the current query statement on the current query template; combining the query template distribution P (X t) obtained before and the lifting degree V of each query template to obtain the exploration value of each query plan of the current query statement; and screening all the query plans according to the optimization conditions, and sending the query plans to be executed to an execution engine.
Fig. 1 is a flowchart illustrating a database query optimization technique based on exploration value and query cost, and fig. 2 is a diagram illustrating a framework structure of the present invention. As shown in fig. 1, the method of the present invention includes steps S101 to S106.
Step S101: and the execution engine in the database stores the execution plans of the historical execution and the real base data corresponding to the historical execution plans into a cache pool, and then performs offline training and modeling on the data in the cache pool by using a local weighted regression model LWPR to obtain a base model M.
Step S102: the method comprises the steps that an optimizer obtains current query sentences from a database management system, the optimizer generates all corresponding query plans based on the current query sentences, carries out base estimation on all query plans by utilizing a base model M trained in the step S101 and a base estimation module of a database, and carries out cost estimation on each execution plan by adopting a cost estimation module of the database, so that estimated cost of each query plan is obtained;
The radix estimation module of the database is a very important component of the database optimizer, in which radix refers to the number of different elements or unique values in a set, and radix estimation is used to estimate the number of different values in a set, rather than calculating a specific number of all values. The radix estimation module is thus a tool in the database optimizer for estimating the number of unique values (radix), such as using statistical techniques (e.g., histograms) or probabilistic estimation algorithms (e.g., hyperLogLog) to obtain the number of unique values (radix).
The cost estimation module of the database is very critical in the query processing of the database, aiming at evaluating the cost of performing the query operation. The database may have multiple execution plans for a query, it being necessary to compare the execution costs of the execution plans based on the (estimated) costs of the different execution plans and to select the best execution plan. The cost estimation module measures the cost of query execution in the form of a variety of resources, including disk access, CPU time used to execute a query, and communication costs in parallel and distributed database systems.
Step S103: recording and logging of query operations performed by the database management system to obtain historical queries of the database management system, comprising: the code and text of the historically executed query statement, the time it takes for the query operation to execute, and the result data returned by the query operation, are used to make predictions of future queries: assuming that the query is loaded with n query templates, the time when the next query statement arrives is set as t, the historical query template set before the time t is { X t-1,Xt-2,Xt-3,...,Xt-m }, wherein X t-1 is the query template at the time t-1, and the query template distribution P (P 1,p2,…,pi,…,pn) at the time t is calculated by using the trained m-order Markov model and the historical query.
Step S104: according to all the query plans obtained in the step S102, each query plan can be decomposed into a plurality of sub-execution plans, each sub-execution plan is composed of a plurality of sub-plan nodes, the lifting degree V (V 1,v2,…,vi,…,vn) of each sub-plan node on each query template is estimated, wherein V i refers to the lifting degree of each sub-plan node on the query template p i, V i is calculated by LWPR models from the sub-plan node of the query plan and the latest k historical queries of the ith query template p i, LWPR models are machine learning algorithms for nonlinear regression and function approximation, the historical queries of the ith query template p i are , represents the latest k historical queries of the query template p i, LWPR models have M receptive fields RECEPTIVE FIELD (RF) (RF 1,RF2,…,RFM), and the corresponding central points of the M receptive fields are C (C 1,c2,…cM); firstly, calculating the distribution P i of the RF center points C by using Q i, as shown in FIG. 3, and then, calculating the lifting degree V i,Vi of each linear model through the linear model of each RF in the LWPR model to describe the prediction uncertainty of each linear model of each RF; finally is obtained.
An example of calculating the probability distribution P i by historical queries in the present invention is shown in fig. 3, and an example of obtaining the degree of elevation of each RF by LWPR model is shown in fig. 4.
Step S105: combining the query template distribution P (P 1,p2,…,pi,…,pn) of the next query obtained in the step S103 and the lifting degree V (V 1,v2,…,vi,…,vn) of each sub-Plan node for querying each query template obtained in the step S104, obtaining the exploration value of each sub-Plan node through , wherein represents the exploration value of the invention, plan refers to each sub-Plan node, and accumulating the exploration values of all sub-Plan nodes on an execution Plan as well as cost estimation to obtain the exploration value of the execution Plan, namely .
Step S106: screening all the query plans according to the optimization conditions to obtain an optimal query plan, namely selecting an execution plan with the largest exploration value under the condition of ensuring that the cost of the selected execution plan is lower than a given threshold value, and sending the query plan to an execution engine, wherein the optimization conditions are as follows:
Wherein denotes maximizing the maximum value of the exploration value of the selected execution plan, denotes the query cost of the query plan, denotes the query cost of the optimal execution plan selected by the optimizer, denotes that the performance of the current execution plan is not lower than times of the current optimal estimated cost, and the optimization condition is that the query plan with the maximum exploration value is selected as the query plan to be executed from all query plans with the query cost smaller than the query cost of the optimal query plan selected by the optimizer.
Claims (6)
1. The database query optimization method based on the exploration value and the query cost is characterized by comprising the following steps of:
Step 1: an execution engine in a database stores an execution plan of historical execution and real base data corresponding to the execution plan into a cache pool, and then carries out offline training and modeling on the data in the cache pool by using a regression model to obtain a base model M;
Step 2: the method comprises the steps that an optimizer obtains current query sentences from a database management system, the optimizer generates all corresponding query plans based on the current query sentences, carries out base estimation on all query plans by utilizing a base model M trained in the step 1 and a base estimation module of a database, and carries out cost estimation on each execution plan by adopting a cost estimation module of the database, so that estimated cost of each query plan is obtained;
Step 3: the method comprises the steps that records and logs of query operations are executed through a database management system, historical query of the database management system is obtained, the historical query comprises codes and templates of query sentences executed in a historical mode, time spent for executing the query operations and result data returned by the query operations, and future query prediction is conducted through the historical query to obtain query template distribution P;
Step 4: according to all the query plans obtained in the step 2, decomposing each execution plan into a plurality of sub-execution plans, and respectively estimating the lifting degree V (V 1,v2,…,vi,…,vn) of each sub-plan node on each query template, wherein V i refers to the lifting degree of the sub-plan node i on the query template p i, and n is the number of the query templates;
Step 5: combining the query template distribution P (P 1,p2,…,pi,…,pn) obtained in the step 3 and the lifting degree V (V 1,v2,…,vi,…,vn) obtained in the step 4, obtaining the exploration value of each sub-Plan node through , wherein/> is the exploration value, plan refers to each sub-Plan node, and accumulating the exploration values of all sub-Plan nodes on the execution Plan as the cost estimation to obtain the exploration value/> of the execution Plan, namely:
；
Wherein Query is an execution plan;
Step 6: and screening all the query plans according to the optimization conditions to obtain an optimal query plan, namely selecting an execution plan with the maximum exploration value under the condition of ensuring that the cost of the selected execution plan is lower than a given threshold value, and sending the query plan to an execution engine.
2. The database query optimization method based on exploration value and query cost of claim 1, wherein:
the regression model employs a locally weighted regression model LWPR.
3. The database query optimization method based on exploration value and query cost of claim 1, wherein:
In the step 3, the queries are generated based on the query templates, so that the prediction of the future query is equivalent to the distribution of the query templates, and the query template distribution P (P 1,p2,…,pi,…,pn) is obtained by using a Markov model, wherein P i refers to the query templates to which the next query may obey.
4. The database query optimization method based on exploration value and query cost of claim 1, wherein:
The predicting step of the future query is: assuming that the query is loaded with n query templates, the time when the next query statement arrives is set as t, the historical query template set before the time t is { X t-1,Xt-2,Xt-3,...,Xt-m }, wherein X t-1 is the query template at the time t-1, and the distribution of the query templates at the time t is calculated to be P (P 1,p2,…,pi,…,pn) by using the trained m-order Markov model and the historical query.
5. The database query optimization method based on exploration value and query cost of claim 1, wherein:
the calculation step of the lifting degree V (V 1,v2,…,vi,…,vn) of each query template is as follows:
V i in V (V 1,v2,…,vi,…,vn) is calculated by LWPR model from the latest k historical queries of the sub-plan node and the ith possible obeying query template p i, specifically, the calculation process of each V i is as follows:
Setting the history query of the ith query template p i as , wherein/> represents the latest k history queries of the query template p i, the LWPR model has M receptive fields (RF 1,RF2,…,RFM), and the center point corresponding to the M receptive fields is C (C 1,c2,…cM); firstly, calculating the distribution P i of the RF center point C by using Q i, and then, calculating the lifting degree V i,Vi of each linear model through the linear model of each RF in the LWPR model to describe the prediction uncertainty of each linear model of each RF; finally,/> was obtained.
6. The database query optimization method based on exploration value and query cost of claim 1, wherein:
And screening all the inquiry plans according to the optimization conditions, wherein the optimization conditions are as follows:
；
Wherein denotes maximizing the maximum value of the search value of the selected execution plan, denotes the search cost of the search plan, denotes the search cost of the optimal execution plan selected by the optimizer, denotes that the performance of the current execution plan is not lower than/> times of the current optimal estimated cost, and the optimization condition is that the search plan with the maximum search value is selected as the search plan to be executed in all the search plans with the search cost less than or equal to the search cost of the optimal execution plan selected by the optimizer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410018425.9A CN117520385B (en) | 2024-01-05 | 2024-01-05 | Database query optimization method based on exploration value and query cost |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410018425.9A CN117520385B (en) | 2024-01-05 | 2024-01-05 | Database query optimization method based on exploration value and query cost |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117520385A CN117520385A (en) | 2024-02-06 |
| CN117520385B true CN117520385B (en) | 2024-04-16 |
Family
ID=89746059
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410018425.9A Active CN117520385B (en) | 2024-01-05 | 2024-01-05 | Database query optimization method based on exploration value and query cost |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117520385B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114911823A (en) * | 2022-03-29 | 2022-08-16 | 哈尔滨工业大学 | Deep learning cost estimation system, method and equipment for cloud-edge-side collaborative query |
| CN116244333A (en) * | 2023-02-16 | 2023-06-09 | 山东大学 | Database query performance prediction method and system based on cost factor calibration |
| CN116521719A (en) * | 2023-04-15 | 2023-08-01 | 复旦大学 | Query optimization system based on cost estimation |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11036735B2 (en) * | 2018-01-16 | 2021-06-15 | Oracle International Corporation | Dimension context propagation techniques for optimizing SQL query plans |
-
2024
- 2024-01-05 CN CN202410018425.9A patent/CN117520385B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114911823A (en) * | 2022-03-29 | 2022-08-16 | 哈尔滨工业大学 | Deep learning cost estimation system, method and equipment for cloud-edge-side collaborative query |
| CN116244333A (en) * | 2023-02-16 | 2023-06-09 | 山东大学 | Database query performance prediction method and system based on cost factor calibration |
| CN116521719A (en) * | 2023-04-15 | 2023-08-01 | 复旦大学 | Query optimization system based on cost estimation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117520385A (en) | 2024-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163410B (en) | Line loss electric quantity prediction method based on neural network-time sequence | |
| CN110163429B (en) | Short-term load prediction method based on similarity day optimization screening | |
| CN112363896B (en) | Log abnormality detection system | |
| Cai et al. | On-device image classification with proxyless neural architecture search and quantization-aware fine-tuning | |
| CN111984514B (en) | Log anomaly detection method based on Prophet-bLSTM-DTW | |
| CN112463337B (en) | Workflow task migration method used in mobile edge computing environment | |
| CN106776316A (en) | A kind of power information software defect prevention method | |
| May et al. | Data analytics for time constraint adherence prediction in a semiconductor manufacturing use-case | |
| CN117520385B (en) | Database query optimization method based on exploration value and query cost | |
| CN111913887B (en) | Software behavior prediction method based on beta distribution and Bayesian estimation | |
| CN111582567B (en) | Wind power probability prediction method based on hierarchical integration | |
| CN116955335B (en) | Address data management method and system based on big data model algorithm | |
| CN112801356A (en) | Power load prediction method based on MA-LSSVM | |
| CN116821643A (en) | HPC job power consumption prediction method and system based on online increment learning | |
| CN115423149A (en) | Incremental iterative clustering method for energy internet load prediction and noise level estimation | |
| CN115081609A (en) | Acceleration method in intelligent decision, terminal equipment and storage medium | |
| CN111046321A (en) | Photovoltaic power station operation and maintenance strategy optimization method and device | |
| Candelieri et al. | Data efficient learning of implicit control strategies in Water Distribution Networks | |
| CN111931994A (en) | Short-term load and photovoltaic power prediction method, system, equipment and medium thereof | |
| CN117235448B (en) | Data cleaning method, terminal equipment and storage medium | |
| CN117931420B (en) | Cloud workload prediction method, device, equipment and storage medium | |
| Li et al. | Reliability-based design optimization based on weighted simulation and Kriging model | |
| CN115794405A (en) | Dynamic resource allocation method of big data processing framework based on SSA-XGboost algorithm | |
| CN118410910A (en) | Park energy scheduling amount prediction method and device | |
| CN116933917A (en) | Power grid material prediction method based on improved wild dog algorithm optimization neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |