CN103678589B - Database kernel query optimization method based on equivalence class - Google Patents
Database kernel query optimization method based on equivalence class Download PDFInfo
- Publication number
- CN103678589B CN103678589B CN201310681482.7A CN201310681482A CN103678589B CN 103678589 B CN103678589 B CN 103678589B CN 201310681482 A CN201310681482 A CN 201310681482A CN 103678589 B CN103678589 B CN 103678589B
- Authority
- CN
- China
- Prior art keywords
- equivalence class
- eclass
- rel
- col
- chained list
- 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
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
- G06F16/24534—Query rewriting; Transformation
- G06F16/24542—Plan optimisation
- G06F16/24544—Join order optimisation
Abstract
The invention belongs to the field of software, and provides a database kernel query optimization method based on an equivalence class. The method comprises the steps of using a function for processing description after the keyword 'from' and before the keyword 'where' in sql sentences after inquire statements enter a sub query optimizer, processing a nest tree structure with explicit connection or a linked list with no connection mode assigned, extracting list information, obtaining list objects, initializing the Col and Eclass bit variables contained by the list objects, and filling a linked list RELS; processing the restricted condition linked list 'quals' after the 'on' and after the 'where' in the sql sentences, and carrying out transmission on the restricted conditions according to the Eclass bit variables; entering an enumerator, enlarging planed searching space according to the Eclass bit variables, enumerating more equivalence plans, and entering a cost calculator to count cost; obtaining the plan with the minimum cost as a final execution plan, wherein the execution plan is compiled into an execution tree, and the execution plan is sent to an actuator to be executed. The database kernel query optimization method based on the equivalence class has the advantages of being high in performance.
Description
Technical field
The invention belongs to software field, more particularly, to a kind of database kernel query optimization method based on equivalence class.
Background technology
The overall architecture of the data base management system of early stage, either from safety or internal memory hardware environment consider,
It is all to be designed with taking the angle of less internal memory.So the data retrieval execution operation of most of traditional relational is all
Employ the streamline mechanism (a demand-pull pipeline mechanism) of a demand pull: system is each
Call plan node (plan node), this will obtain a record or null record from following node.Resident in internal memory
Data page number at most only have query tree leaf node number, this mechanism can significantly save the use of internal memory, and it can
To be efficiently applied to the work such as simple extraction, the checking of single table data.But the multi-table join of relational database management system
Or during complicated calculations, generally require mass data terminate-and-stay-resident.The memory techniques high speed development of computer now, computer is joined
The memory size put increases continuous, and database kernel is also updated with this.Streamline in original demand pull
Under mechanism, traditional database kernel starts to allow the data of middle leaf node resident.So, new problem occurs, that is, central
Between leaf node data too big and when exceeding internal memory, the data of internal memory just can be interacted with disk, and that is, disk is as virtual interior
Presence, under this scene, Database Systems make performance reduce rapidly because frequently io operates, and therefore, reduce as far as possible
The not still a large amount of minimizings to cpu below of the data volume of middle leaf node, are also important optimization means of io.
Using equivalence class, the equivalent table connecting is marked in traditional relational, the equivalent connection of equivalence class labelling
Set, is placed in the bottom of threaded tree, is preferentially attached, advantage of this is that and avoid cartesian product (cart-
Prod) connect and first carry out, so, the heuristic strategies based on equivalence class can largely reduce the meter of query optimization module
Draw search space.On the other hand, equivalence class is the key concept in discrete mathematics, itself has reflexivity, symmetry and transmission
Property；It is used as the important means of Data Reduction in a large number in fields such as knowledge discovery in database, but, available data library management
System does not make full use of the attribute reduction characteristic of equivalence class, and the filtering rod of class members of equal value in query optimization module
Part shared.These characteristics of equivalence class insufficient using directly result in database kernel lose early cross filter data machine
Meeting, so that intermediate result is excessive, leads to the consumption of cpu and io huge, ultimately causes the decline of data base querying performance.
Furthermore, database management system is using cost-based query optimization method (cost based mostly
Optimization).Rule-based optimizing device to enumerate as far as possible institute planned, calculate each plan cost,
And the scheduled transfer of final choice Least-cost is to executor.Traditional database does not make full use of equivalence class and excavates potential information
Function, missed the inquiry plan of more optimization, directly resulted in the reduction of final query execution performance.For example, optimizer
Not making full use of equivalence class so that not getting the potential characteristic of relation data base table, may be such that base table can not filter foot
Enough data, make optimizer not start index then, ultimately result in recall precision and reduce.
Content of the invention
The purpose of the embodiment of the present invention is to provide a kind of database kernel query optimization method based on equivalence class, its solution
The certainly low problem of some query performances of data base of prior art presence.
On the one hand the embodiment of the present invention is achieved in that, provides a kind of inquiry of the database kernel based on equivalence class excellent
Change method, methods described includes:
After triggering query optimization, after entering subquery optimizer, process key word from sql sentence using function
Description afterwards and before key word where: enter the explicit nested tree construction connecting or specified connection
The chained list of mode, extracts table information, obtains table object, initializes col the and eclass bit variable that it comprises, and fills chained list
rels.
Process the restrictive condition chained list quals after in sql sentence；
Process the description after where in sql sentence；
Traversal distinct, order by, group by belongs to the chained list of row characteristic, finds it according to eclass variable right
The equivalence class answered, these characteristics is risen to the characteristic of whole equivalence class；These characteristics are delivered to it each by equivalence class
In member；Each member of parity price class adds these characteristics and can operate with for optimization below；
Quals chained list in traversal sql, finds corresponding row to the qual of each monocular, according to row object rel's
Eclass object finds corresponding only one equivalence class, finds other row according to bitmap variable col of this equivalence class, will
Monocular constraints qual is sent among these row members；
Enter enumerator, whenever producing a pair of input condition, judge that a pair of input condition whether there is in equivalence class, such as
Exist, enter cost calculator；Obtain minimum cost is intended to be final implement plan, and implement plan is compiled into execution
Tree, is sent to execution in executor.
Optionally, the rule that described a pair of input condition of judgement whether there is in equivalence class specifically includes:
Calculate the phase of bitmap variable eclass in two table object rel in a pair of input condition and operating result, if
It is not zero with operating result, determine in equivalence class, there is equivalent association；Otherwise, there is not equivalent association.
Optionally, the restrictive condition chained list quals after in described process sql sentence specifically includes:
Obtain row object, initialize its comprising variable rel and eclass；Rel according to existing rel object value, according to
Col generates a class object ec of equal value, and rel is generated as, according to this ec, the eclass assignment that it comprises, and final col fills cols chain
Table；Meanwhile, the equivalence class object of this generation is itself rel and col variable assignments according to existing rels and cols.
Optionally, the equivalence class object of described generation is itself rel and col variable assignments bag according to existing rels and cols
Include:
Find the effectively equivalent connection that inner connects；If there is effective equivalent connection it should check that ecs looks to relate to
And two equivalence classes arriving whether individualism, in this way, then this two equivalence classes are merged in an equivalence class, another etc.
Valency class will be disallowable, simultaneously the eclass variable in rel and col to be changed.
Optionally, the description after where in described process sql sentence includes:
Return to the target chained list in sql sentence, before select, after from, arrange other column informations, generate object
Col, assignment rel and eclass variable, continue to fill up cols chained list；And add corresponding equivalence class.
In embodiments of the present invention, the technical scheme that the present invention provides makes full use of the transitivity of equivalence class, by equivalence class
One of the attribute of member be delivered to other all members of this equivalence class, carry out basis for optimizing, and finally realize new
Optimisation strategy, so it has the advantages that to improve data base querying performance.
Brief description
Fig. 1 is a kind of flow chart of database kernel query optimization method based on equivalence class that the present invention provides；
Fig. 2 is the broken line graph of the execution time that the embodiment of the present invention provides different filterconditions.
Specific embodiment
In order that the objects, technical solutions and advantages of the present invention become more apparent, below in conjunction with drawings and Examples, right
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only in order to explain the present invention, and
It is not used in the restriction present invention.
First, the specific embodiment of the present invention needs to define three overall chained list rels, cols, ecs and is respectively intended to deposit
Table object, row object and class object of equal value that storage gradually generates during query statement analysis.Rels stores table object
(rel), rel comprises two bit variable col and eclass, represents this table object rel and other row objects or class object of equal value
Association；Cols stores row object col, and row object comprises two bit variable rel and eclass, represent this row object col and other
Table object or the association of class object of equal value；Ecs storage class object ec of equal value, ec comprise two bit variable rel and col, and representing should
The associating of class object ec of equal value and other table objects or row object.
The specific embodiment of the invention provides a kind of database kernel query optimization method based on equivalence class, and the method is such as
Shown in Fig. 1, comprising:
101st, after triggering query optimization, after entering subquery optimizer, using function (process_from_
Clause, this function can also be the function of other forms) process in sql sentence (or sub- sql sentence) key word from it
Description afterwards and before key word where: enter the explicit nested tree construction connecting or specified connection side
The chained list of formula, extracts table information, obtains table object (rel), initializes col the and eclass bit variable that it comprises, and fills chained list
rels.
102nd, connect if there is explicit, process the restrictive condition chained list after in sql sentence (or sub- sql sentence)
quals；Obtain row object (col), initialize its comprising variable rel and eclass；Rel is according to existing rel object value, root
Generate a class object ec of equal value according to col, rel is generated as, according to this ec, the eclass assignment that it comprises, final col fills cols
Chained list；Meanwhile, the equivalence class object of this generation is itself rel and col variable assignments according to existing rels and cols.Find
The effectively equivalent connection that inner connects.Non- inner connects the row being designed into and can not add and forms new equivalence class pair in equivalence class
As eclass.If there is effective equivalent connection it should check that ecs looks at whether be related to two equivalence classes are individually deposited
In this way, then this two equivalence classes are being merged in an equivalence class, another equivalence class will be disallowable.Rel to be changed simultaneously
With the eclass variable in col.
103rd, process the description after where in sql sentence.Return to target chained list in sql sentence (before select,
After from), arrange other column informations, generate object col, assignment rel and eclass variable, continue to fill up cols chained list；
And adding corresponding equivalence class, this equivalence class is if newly-generated, then typically with regard to only one of which equivalence class members.
104th, traversal distinct, order by, group by(after can as sequence process) etc. belong to row characteristic
Chained list, finds its corresponding equivalence class according to eclass variable, these characteristics is risen to the characteristic of whole equivalence class.Equivalence class
These characteristics are delivered in its each member (arranging)；Each member of parity price class adds these characteristics can be for below
Optimization operate with.
105th, the quals chained list in traversal sql, finds corresponding row to the qual of each monocular.According to row object rel
Eclass object find corresponding only one equivalence class, other row are found according to bitmap variable col of this equivalence class.
By monocular constraints qual(such as a > 2) it is sent among these row members.
106th, enter enumerator, whenever producing a pair of input condition, judge whether a pair of input condition is deposited in equivalence class
Such as exist, entering cost calculator.The rule judging is the bitmap calculating in two table object rel in a pair of input condition
The phase of variable eclass and operating result, if be not zero with operating result, then means that this two tables can carry out equivalence
Association, can be attached operating；Otherwise, there is not equivalent association and be attached operation and can cause an equivalence class node
Descartes and connects in bottom child node, the bad plan that final one intermediate result of generation will be very huge.
107th, enter cost calculator, obtain minimum cost is intended to be final implement plan, and implement plan is compiled
It is translated into and executes tree, be sent to execution in executor.
Embodiment
Assume there is table t1, t2 and t3, t1, t2 and t3 have two row col1 and col2 respectively.For query statement:
“select t1.col1,t2.col2,t3.col2from t1,t2,t3where t1.col1=t2.col2and t2.col2=
T3.col1and t1.col1 > 0.9order by t2.col2 ", according to the thinking of the present invention, specific flow process is as follows:
1st, enter inquiry plan device, be recycled into subquery planner, overall chained list rels, cols, ecs are respectively intended to deposit
Table object, row object and class object of equal value that storage generates during query statement analysis.Initialization chained list is as shown in table 1.When
, more than eight, bitmap variable (as rel col eclass) can be with dynamic load, merely just using one for the number of object structure
The position of individual byte come to illustrate illustrate.Overall bit variable record each object id numbering, and the rel in object col
Eclass record and the association between the corresponding of global variable and variable.
Table 1: initialization overall chained list rels, cols, ecs
2nd, after entering subquery optimizer, after processing key word from sql sentence, before key word where
Table t1, t2 and t3.Global variable rels initializes, as shown in table 2.Continue with clause, because this example does not have on clause,
Can not find the information of row and the information of inner join, so col and eclass does not change.
Table 2: global variable rels initialization
3rd, enter in the process of where clause, ergodic condition chained list, the equivalent description t1.col1=connecting occurs
T2.col2, so the value changing overall situation equivalence class variable eclass here is 10000000.That is, occur equivalence class
T1.col1t2.col2 }.Changing col is 11000000, and the col variable in t1, t2 and t3 respectively 10000000,
01000000 and 00000000, t1 and t2 in eclass variable be all that eclass variable in 10000000, t3 is
00000000.As shown in table 3.
Table 3: process the change of rels, cols, ecs of where
4th, continue traversal, second equivalent description t2.col2=t3.col1 connecting occurs, adds new row t3.col1.
Due to there are other equivalence classes { t1.col1t2.col2 }, and the member of this equivalence class be present in equivalent connection description it
In, so needing to consider the merging of equivalence class.Will t3.col1 add in first equivalence class, class variable eclass of equal value is not
Become, but it member increase, id number be 0 equivalence class set side be { t1.col1t2.col2t3.col1 }.Change it
The value of rel and col is 11100000, and changing the col variable in t3 is 00100000, and the eclass variable changing in t3 is
10000000.Adjustment id is that rel and eclass of 2 col object is respectively 00100000 and 10000000.As shown in table 5.
Table 5: process the change of rels, cols, ecs of where
Continue traversal, constrained condition t1.col1 occurs 0.9, by restrictive condition record in restrictive condition chained list
quals.As shown in table 6:
Table 6: the record of restriction condition
5th, return to the target chained list (before select, after from) in sql sentence: row t1.col1 and t2.col2 is
Exist, and t3.col2 is non-existent, generate new object col, assignment rel and eclass variable, continue to fill up cols chained list；
And add corresponding equivalence class, this equivalence class only one of which equivalence class members.As shown in table 8:
Table 8: the interpolation of objective attribute target attribute
6th, record distinct, order by, group by characteristic.Exist in this example " order by t2.col2 ",
So shown in the related amendments information of cols such as table 9 highlights.
Table 9: the interpolation of objective attribute target attribute
7th, travel through quals chained list, find monocular constraints t1.col1 0.9, t1 is according to its local variable eclass(value
For 10000000) find equivalence class { t1.col1t2.col2t3.col1 } belonging to t1.col1.By col > 0.9 constraint bar
Part travels to each member of this equivalence class, can obtain t2.col2 > 0.9 and t3.col1 > 0.9.And this constraints is hung
It is downloaded in the restriction condition of table t2 and table t3.Simultaneously as there is the characteristic of order by t2.col2, this equivalence class
In other members also share this characteristic, so all can have the characteristic of sequence.Prominent in cols after modification such as table 10
Shown in display:
Table 10: the characteristic transmission of row
8th, enter plan enumerator.When generation t1 and t3 is attached prior to t2, the local using t1 and t2 becomes
Amount eclass(be 10000000) enter line position is 10000000 with operating result, that is, exist equivalence class
T1.col1t2.col2t3.col1 }, although there is no explicit t1.col1=t3.col1, still can be according to this equivalence class
Obtain this condition of contact, and can pass in cost calculator and calculated.The plan of final choice Least-cost, is delivered to execution
Device executes inquiry operation.
In order to verify the effectiveness of the new framework route of the application proposition, in latest edition 9.2.4 of postgresql
Middle realization.This place, by the test data random using a group, employs the contrast test of 1000000 random data, test
Create sentence as follows:
create table rdmt1(random float);
create table rdmt2(random float);
insert into rdmt1select random()from generate_series(1,1000000);
insert into rdmt2select random()from generate_series(1,1000000);
Wherein random () is the system function of postgresql, randomly generates 0～1 random number；
Generate_series (1,1000000) is the sequence number producing from 1 to 1000000.Insertion sentence represents to table
Rdmt1 and rdmt2 inserts 1000000 random floating point (0～1) respectively.Before and after the data of generation is directed respectively into modification
Data base in.Execute query statement: " explain select count (*) from client psql of postgresql
rdmt1,rdmt2where rdmt1.random=rdmt2.random and rdmt1.random>0.9;”.Amended
The inquiry plan that postgresql obtains is as mentioned below.Table rdmt2 is created with a conventional b tree index rdmt2_random_
Idx, and check inquiry plan respectively, not improved postgresql produces inquiry plan and does not change.But it is improved
The inquiry plan of postgresql has very big change as mentioned below.The inquiry meter that unmodified postgresql obtains
Draw as follows:
The inquiry plan that amended postgresql obtains is as follows:
Improved postgresql is as follows to the inquiry plan having index data:
Show in the inquiry plan that unmodified postgresql obtains, because postgresql9.2.4 is not fully sharp
With the transitivity of equivalence class so that constraint filtercondition (random > 0.9::double precision) does not have equivalence to be delivered to
On master meter rdmt2, and in amended postgresql, constraint filtercondition can be for delivery on master meter rdmt2, inquiry plan
Highlight part.And, after increase a b tree index to master meter rdmt2, the inquiry of not improved postgresql
Plan is still constant.But, the inquiry plan of improved postgresql becomes improved postgresql further to there being rope
The inquiry plan (highlighting part) of argument evidence, after that is, filtercondition equivalence has been delivered to master meter rdmt2, triggers index
Use.
In order to avoid the impact of operating system memory scheduling strategy and postgresql backstage checkpoint cycle grasp
The impact made, each plan performs 20 times, and records the wherein short time.The execution of three query statements spends above
Time is 495.871ms, 280.945ms and 216.502ms respectively.It can be seen that the time of execution is ever-reduced.Due to
The random floating point that random () produces is almost equally distributed, and restrictive condition " random > 0.9 " is probably to have filtered very
Nine data, so improve after postgresql execution efficiency to improve a lot.Filter size of data for checking to execution
The impact of performance, has done eight groups of tests herein again, the data of filtration has been gradually reduced.The filtercondition of random is reduced every time
The time result of 0.1 survey is as shown in table 11 below:
Table 11: the execution time (unit: ms) of different filterconditions
The broken line graph that the data of table 11 is drawn as is as shown in Figure 2.In table 8 we be clear that improved
Postgresql execution efficiency is significantly better than not improved postgresql.And improved postgresql is in certain bar
The use of index can be triggered under part.When limiting numerical value for 0.7,0.8,0.9, index is added into that is to say, that optimizing
The cost that device is computed addition index is smaller, and " optimum " path just adds index.So it can be seen in the drawing that limiting number
It is worth between 0.1 to 0.6, as having indexless improvement postgresql execution efficiency to be almost.
As can be seen from Table 11, when limiting numerical value and be 0.9, the query time not having an improved plan is
The execution time of 495.871ms, the postgresql after improvement is 280.945ms, and performance improves 43%.After adding index,
The original plan does not still change, and does not use index.And amended postgresql employs new index, and
And execution time is 216.502ms, performance improves 56%.As can be seen that performance has than larger raising it was demonstrated that herein
The high efficiency of the new architecture method for equivalence class of design.
One of ordinary skill in the art will appreciate that realizing all or part of step in the various embodiments described above method is can
Completed with the hardware instructing correlation by program, corresponding program can be stored in a computer read/write memory medium
In, described storage medium, such as rom/ram, disk or CD etc..
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention
Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.
Claims (3)
1. a kind of database kernel query optimization method based on equivalence class is it is characterised in that methods described includes:
After triggering query optimization, after entering subquery optimizer, after processing key word from sql sentence using function
And the description before key word where: enter the explicit nested tree construction connecting or specified connected mode
Chained list, extract table information, obtain table object, col bit variable and eclass bit variable that initial table object comprises, fill chain
Table rels；
Process the restrictive condition chained list quals after in sql sentence, the restrictive condition after in described process sql sentence
Chained list quals specifically includes:
Obtain row object col, initialize rel bit variable and the eclass bit variable that it comprises；Rel is according to existing table object
Rel assignment, generates a class object ec of equal value according to col, and rel is generated as, according to this ec, the eclass bit variable that it comprises and assigns
Value, final col fills cols chained list；Meanwhile, the equivalence class object of this generation according to existing rels and cols be its rel and
Col bit variable assignment,
The equivalence class object of described generation includes for itself rel and col bit variable assignment according to existing rels and cols:
Find the effectively equivalent connection that inner connects；If there is effective equivalent connection it should check that chained list ecs looks to relate to
And two equivalence classes arriving whether individualism, in this way, then this two equivalence classes are merged in an equivalence class, another etc.
Valency class will be disallowable, simultaneously the eclass bit variable in rel and col to be changed；
Process the description after where in sql sentence；
Traversal distinct, order by, group by belongs to the chained list of row characteristic, finds its correspondence according to eclass bit variable
Equivalence class, these characteristics are risen to the characteristic of whole equivalence class；These characteristics are delivered to its each one-tenth by equivalence class
In member；Each member of parity price class adds these characteristics and can operate with for optimization below；
Quals chained list in traversal sql, finds corresponding row to the qual of each monocular, according to the eclass of table object rel
Bit variable finds corresponding only one equivalence class, finds other row according to the col bit variable of this equivalence class, by monocular about
Bundle condition qual is sent among these row members；
Enter enumerator, whenever producing a pair of input condition, judge that a pair of input condition whether there is in equivalence class, such as deposit
Entering cost calculator；Obtain minimum cost is intended to be final implement plan, and implement plan is compiled into execution
Tree, is sent to execution in executor.
2. method according to claim 1 is it is characterised in that described judge whether a pair of input condition is deposited in equivalence class
Rule specifically include:
Calculate the phase of eclass bit variable in two table object rel in a pair of input condition and operating result, if with operation
Result is not zero, and determines there is equivalent association in equivalence class；Otherwise, there is not equivalent association.
3. method according to claim 1 is it is characterised in that description bag after where in described process sql sentence
Include:
Return to the target chained list in sql sentence, before select, after from, arrange other column informations, generate col, assignment
Rel and eclass bit variable, continues to fill up cols chained list；And add corresponding equivalence class.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310681482.7A CN103678589B (en) | 2013-12-12 | 2013-12-12 | Database kernel query optimization method based on equivalence class |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310681482.7A CN103678589B (en) | 2013-12-12 | 2013-12-12 | Database kernel query optimization method based on equivalence class |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103678589A CN103678589A (en) | 2014-03-26 |
| CN103678589B true CN103678589B (en) | 2017-02-01 |
Family
ID=50316134
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310681482.7A Active CN103678589B (en) | 2013-12-12 | 2013-12-12 | Database kernel query optimization method based on equivalence class |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103678589B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105335403B (en) * | 2014-07-23 | 2020-02-14 | 华为技术有限公司 | Database access method and device and database system |
| BR112018010857B1 (en) | 2015-11-30 | 2023-02-07 | Huawei Technologies Co., Ltd | DATA CONSULTATION METHOD AND APPLIANCE, AND DATABASE SYSTEM |
| CN106897343B (en) * | 2016-07-20 | 2020-08-07 | 阿里巴巴集团控股有限公司 | Searching method, storing method and device for execution plan |
| US11250004B2 (en) * | 2016-09-27 | 2022-02-15 | Nippon Telegraph And Telephone Corporation | Secure equijoin system, secure equijoin device, secure equijoin method, and program |
| CN108304517A (en) * | 2018-01-23 | 2018-07-20 | 西南大学 | Efficient nested querying method based on Complex event processing system |
| CN110659200B (en) * | 2018-06-29 | 2023-05-02 | 中国航发商用航空发动机有限责任公司 | Method and system for comparing and analyzing source code and target code of aviation onboard software |
| CN110866033B (en) * | 2018-08-28 | 2022-06-21 | 北京国双科技有限公司 | Feature determination method and device for predicting query resource occupancy |
| CN113111058B (en) * | 2020-01-10 | 2023-03-28 | 华为技术有限公司 | Database processing method and device |
| CN112395304B (en) * | 2020-10-30 | 2024-01-02 | 迅鳐成都科技有限公司 | Data security calculation method, system and storage medium based on data behavior simulation |
| CN114925092B (en) * | 2022-05-09 | 2023-05-30 | 北京达佳互联信息技术有限公司 | Data processing method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102930003A (en) * | 2012-10-24 | 2013-02-13 | 浙江图讯科技有限公司 | Database query plan optimization system and method |
| CN103365885A (en) * | 2012-03-30 | 2013-10-23 | 国际商业机器公司 | Method and system for optimizing database inquiry |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8429151B2 (en) * | 2010-11-22 | 2013-04-23 | Ianywhere Solutions, Inc. | Highly adaptable query optimizer search space generation process |
-
2013
- 2013-12-12 CN CN201310681482.7A patent/CN103678589B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103365885A (en) * | 2012-03-30 | 2013-10-23 | 国际商业机器公司 | Method and system for optimizing database inquiry |
| CN102930003A (en) * | 2012-10-24 | 2013-02-13 | 浙江图讯科技有限公司 | Database query plan optimization system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103678589A (en) | 2014-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103678589B (en) | Database kernel query optimization method based on equivalence class | |
| CN113228001B (en) | Eliminating query fragment duplication in complex database queries | |
| KR102074026B1 (en) | Profiling data with location information | |
| US9092481B2 (en) | Method and system for processing graph queries | |
| CN106471489B (en) | Manage the data with flexible modes | |
| JP3640346B2 (en) | Set predicates and retrieval in database management systems | |
| Meimaris et al. | Extended characteristic sets: graph indexing for SPARQL query optimization | |
| CN102270232B (en) | Semantic data query system with optimized storage | |
| US20080222087A1 (en) | System and Method for Optimizing Query Access to a Database Comprising Hierarchically-Organized Data | |
| US20100235344A1 (en) | Mechanism for utilizing partitioning pruning techniques for xml indexes | |
| CN103729392A (en) | Method for optimizing query and query complier | |
| Giannakouris et al. | MuSQLE: Distributed SQL query execution over multiple engine environments | |
| US11693883B2 (en) | Techniques for ordering predicates in column partitioned databases for query optimization | |
| US20130290292A1 (en) | Augmented Query Optimization by Data Flow Graph Model Optimizer | |
| CN108536692A (en) | A kind of generation method of executive plan, device and database server | |
| US20060161525A1 (en) | Method and system for supporting structured aggregation operations on semi-structured data | |
| CN104778236A (en) | ETL (Extract-Transform-Load) realization method and system based on metadata | |
| CN109325029A (en) | RDF data storage and querying method based on sparse matrix | |
| CN104081397A (en) | Creating an archival model | |
| CN102495834A (en) | Incremental data cleaning method based on memory mapping | |
| WO2011106006A1 (en) | Optimization method and apparatus | |
| Rusu et al. | In-depth benchmarking of graph database systems with the Linked Data Benchmark Council (LDBC) Social Network Benchmark (SNB) | |
| Saleem | Storage, indexing, query processing, and benchmarking in centralized and distributed RDF engines: a survey | |
| Salehpour et al. | A comparative analysis of knowledge graph query performance | |
| Wu et al. | Answering XML queries using materialized views revisited |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 100094 Beijing city Haidian District North Road No. 68, UFIDA Software Park Applicant after: Yonyou Network Technology Co., Ltd. Address before: 100094 Beijing city Haidian District North Road No. 68, UFIDA Software Park Applicant before: UFIDA Software Co., Ltd. |
|
| COR | Change of bibliographic data | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |