CN106844020A - Different task travels through the high-efficient treatment method and system of a large amount of shared datas - Google Patents
Different task travels through the high-efficient treatment method and system of a large amount of shared datas Download PDFInfo
- Publication number
- CN106844020A CN106844020A CN201611087441.5A CN201611087441A CN106844020A CN 106844020 A CN106844020 A CN 106844020A CN 201611087441 A CN201611087441 A CN 201611087441A CN 106844020 A CN106844020 A CN 106844020A
- Authority
- CN
- China
- Prior art keywords
- broken
- task
- single component
- management module
- unit
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to digital processing technology field, and in particular to a kind of different task travels through the high-efficient treatment method and system of a large amount of shared datas.The system includes broken Single Component Management module and several task modules, and shared data is split into several broken units, additions and deletions are carried out to broken unit by described broken Single Component Management module, and each broken unit is numbered, and preserves the information of broken unit；Several described task modules can carry out operation treatment to different broken cell datas simultaneously；Broken Single Component Management module changes broken unit information according to each task module to the result of different broken cell datas, and it is that the broken unit for not having task to process carries out operation treatment to state only to allow task module.The present invention can greatly improve the efficiency that multitask is operated to big shared data.
Description
Technical field
The invention belongs to digital processing technology field, and in particular to a kind of different task travels through the height of a large amount of shared datas
Effect processing method and system.
Background technology
At present in many application and developments, many places business will be deleted shared data according to certain rule, increased
Deng operation treatment.It is in general, many in order to ensure the security of data when multithreading carries out operation treatment to shared data
Task cannot carry out operation simultaneously, after a task can only be waited to be disposed shared data, next to shared data
Business could then start to process identical shared data, the like.
In some application scenarios, because task is a lot, shared data amount is very huge, and a task will be disposed institute
The time for having shared data to need is more long, when this allows for needing to wait more long when next task processes identical shared data
Between, or even could not also be when processing shared data in next task, next one task also comes into queuing etc.
In treating, this multitask is to the processing method of shared data not only inefficiency, it could even be possible to causing procedure service to be delayed machine.
The content of the invention
Defect it is an object of the invention to be directed to prior art, there is provided a kind of different task travels through a large amount of shared datas
Processing method and system so that multiple tasks process the different sections with portion shared data simultaneously, to realize multitask to same
The efficient operation treatment of big shared data.
Technical scheme is as follows：A kind of different task travels through the high-efficient treatment method of a large amount of shared datas, including
Following steps：
(1) shared data is split into several broken units, and sets up broken Single Component Management module；
(2) task A obtained broken unit by broken Single Component Management module first before being operated to the shared data
Total quantity；
(3) one numbering of broken unit of the random acquirements of task A, and this is numbered to the broken Single Component Management module request
Broken cell data is operated；
(4) the broken Single Component Management module checks the broken location mode of corresponding numbering, and judges whether task A can be broken to this
Unit carries out operation treatment；
(5) after task A receives the judgement information of broken Single Component Management module, determine it is to current volume according to the judgement information
Number broken cell data processed, the still new broken element number of random request one again；If necessary at random please again
A new broken element number is sought, then return to step (3)；
(6) after task A is to being currently able to the completion operation treatment of the broken cell data for the treatment of, from the sum of the broken unit for obtaining
The numbering of the broken unit is deleted in amount, and data have been processed in notifying the broken unit of the broken Single Component Management module numbering
Finish；Broken Single Component Management module is updated to the status information of the broken unit of the numbering；
(7) task A updates the total quantity of untreated broken unit, and return to step (3) is circulated operation, until completing
The treatment of whole shared data.
Further, different task as described above travels through the high-efficient treatment method of a large amount of shared datas, described in step (4)
Broken Single Component Management module checks the broken location mode of corresponding numbering, it is determined whether have other tasks operating the broken unit of this numbering
The broken location mode, if the broken cell data of this numbering does not have task to process, is changed to task and taken, data manipulation by data
Authority gives task A treatment；If this broken unit is occupied state, notify the task A broken cell datas by other tasks
Take.
Further, if the judgement presentation of information that task A receives broken Single Component Management module return has other in step (5)
Broken cell data of the task to corresponding numbering takes, then the new broken element number of random request one again.
Further, different task as described above travels through the high-efficient treatment method of a large amount of shared datas, wherein, if step
(7) total quantity of the untreated broken units of task A less than setting value, (such as 10, it is proposed that 5), compile by the random broken unit of acquirement one in
Number send data processing request to broken Single Component Management module, broken Single Component Management module return information shows the broken unit number of this numbering
A volume is obtained at random again according to a period of time (depending on according to circumstances, generally 1-3 seconds) in other tasks occupancy, is then postponed
Number send request to broken Single Component Management module.
Further, different task as described above travels through the high-efficient treatment method of a large amount of shared datas, wherein, multiple tasks
The operation of the step (2)-(7) is carried out simultaneously.
Further, different task as described above travels through the high-efficient treatment method of a large amount of shared datas, wherein, if task
B does not include broken element number to the operation requests that broken Single Component Management module sends, then show that task B please to broken Single Component Management module
Seeking shared data increases, and now, broken Single Component Management module searches are in the minimum broken unit of free state data total amount and set
Task seizure condition, and task B operations are given this broken unit, after task B completes shared data increase operation, notify broken unit
Management module end of operation, should be free state this broken location mode, and update after broken Single Component Management module is connected to instruction
This broken unit information, is easy to other tasks to operate the data.
A kind of different task travels through the efficient process system of a large amount of shared datas, including broken Single Component Management module and several
Shared data is split into several broken units by task module, described broken Single Component Management module, and additions and deletions are carried out to broken unit, and
Each broken unit is numbered, the information of broken unit is preserved；Several described task modules can simultaneously to different broken
Cell data carries out operation treatment；Broken Single Component Management module according to each task module to the result of different broken cell datas more
Change broken unit information, and it is that the broken unit for not having task to process carries out operation treatment to state only to allow task module.
Further, different task as described above travels through the efficient process system of a large amount of shared datas, described broken unit
Information including broken unit total quantity, the state of each broken unit, the total amount of each broken cell data.
Further, different task as described above travels through the efficient process system of a large amount of shared datas, wherein, described is broken
The total quantity of unit is more than five times of task module quantity.
Beneficial effects of the present invention are as follows：Shared data is divided into many parts by the present invention, allows multiple tasks to carry out at every portion
Reason, after being disposed, then randomly selects lower a treatment, until all shared datas are disposed.The present invention can allow many
Business is operated to a large amount of shared datas simultaneously, realizes that multiple tasks process the different sections with portion shared data, treatment simultaneously
Process is without waiting for the mode of random process can reduce the probability of the multitask same data segment of same time operation, so that significantly
Improve the efficiency that multitask is operated to big shared data.
Brief description of the drawings
Fig. 1 is broken relation schematic diagram of the Single Component Management module (referred to as broken pipe) with broken unit (referred to as broken) of the invention；
Fig. 2 be technical solution of the present invention in number the process chart taken by other tasks；
Fig. 3 is task (additions and deletions change) peration data flow figure in technical solution of the present invention；
Fig. 4 is an increase data task flow chart in technical solution of the present invention.
Specific embodiment
The present invention is described in detail with reference to the accompanying drawings and examples.
The present invention is to process big shared data based on multiple tasks to solve simultaneously to patrol one kind that shared data additions and deletions are processed
The method of collecting, is when multiple tasks are operated to big shared data, at one kind that big shared data can not read while write simultaneously
Reason method.It is mainly used in shared data to be stored in server memory, when multithreading cannot read while write operation to shared data
A solution.
Cardinal principle of the present invention is exactly that shared data is divided into many parts, allows multiple tasks to carry out every a treatment, has been processed
Bi Hou, then lower a treatment is randomly selected, until all shared datas are disposed.Multiple tasks are reached to process simultaneously with a
Shared data difference section, purpose of the multitask to same big shared data efficient operation.In principle of the invention and following life
Example it is exactly the same, such as 10 books are regarded as an entirety, there are 10 students to read this 10 books, a student can only see
One book, another nine books are in free time, should just allow other nine students respectively to select a books to read, rather than grade
After life all reads this 10, then read to next student, allow other students to be in without in the readable wait of book.
The present invention will realize the multitask a large amount of shared datas of efficient process simultaneously, and multiple tasks treatment shared data must expire
Following two premises of foot,
1) shared data task treatment does not differentiate between sequencing；
2) multitasking shared data operation does not influence other tasks to process.
The complete shared data duration of each task traversal processing is not necessarily equal, and each task is in same time unit internal trigger
Number of times is also inconsistent, and we obtain the probability treatment mechanism of certain number evidence using task to shared data of traversal at random, here
A concept of broken unit is proposed, and is realized by setting up broken Single Component Management module.
From from the perspective of system, different task provided by the present invention travels through the efficient process system of a large amount of shared datas
Including broken Single Component Management module and several task modules.
Shared data is split into several broken units (referred to as broken) by described broken Single Component Management module (referred to as broken pipe), right
Broken unit carries out additions and deletions, and each broken unit is numbered, and for distinguishing the identity of each broken unit, preserves the letter of broken unit
Breath, the total quantity of these information including broken unit, the state of each broken unit, the total amount of each broken cell data etc., such as Fig. 1 institutes
Show.
Several described task modules can carry out operation treatment to different broken cell datas simultaneously；Broken Single Component Management
Module changes broken unit information according to each task module to the result of different broken cell datas, and only allows task module pair
State is that the broken unit for not having task to process carries out operation treatment.In general the total quantity of described broken unit is task module
At five times or so of quantity, treatment effeciency is higher, and the total quantity of broken unit is designed as task module number by this method general recommendations
Five times or more than five times of amount.
Below by taking a specific task A as an example, some specific tasks is how to travel through in introducing the method for the present invention
A large amount of shared datas carry out processing operation, the present invention when realizing, be multiple tasks simultaneously performing this operating process,
The operation of one specific tasks comprises the following steps：
(1) shared data is split into several broken units, and sets up broken Single Component Management module；
(2) task A obtained broken unit by broken Single Component Management module first before being operated to the shared data
Total quantity；
(3) one numbering of broken unit of the random acquirements of task A, and this is numbered to the broken Single Component Management module request
Broken cell data is operated；
(4) the broken Single Component Management module checks the broken location mode of corresponding numbering, it is determined whether have other tasks in behaviour
Make the broken cell data of this numbering, if the broken cell data of this numbering does not have task to process, the broken location mode is changed to appoint
Business is taken, and task A treatment is given data manipulation authority；If this broken unit is occupied state, the task A broken lists are notified
Metadata is taken by other tasks；
(5) after task A receives the judgement information of broken Single Component Management module, determine it is to current volume according to the judgement information
Number broken cell data processed, the still new broken element number of random request one again；If task A receives broken unit
The judgement presentation of information that management module is returned has broken cell data of other tasks to corresponding numbering to take, then return
Step (3)；
(6) after task A is to being currently able to the completion operation treatment of the broken cell data for the treatment of, from the sum of the broken unit for obtaining
The numbering of the broken unit is deleted in amount, and data have been processed in notifying the broken unit of the broken Single Component Management module numbering
Finish；Broken Single Component Management module is updated to the status information of the broken unit of the numbering；
(7) task A updates the total quantity of untreated broken unit, and return to step (3) is circulated operation, until completing
The treatment of whole shared data.
Fig. 2 shows the handling process of the example that a broken element number is taken by other tasks.As shown in Fig. 2 task A
A numbering for broken unit 3 is obtained at random, and the broken cell data of this numbering 3 is carried out to the broken Single Component Management module request
Operation；Broken Single Component Management module checks the broken location mode of numbering 3, and the status display of the broken unit is occupied, then notify to appoint
The business A broken cell datas are taken by other tasks；Task A obtains a numbering for broken unit 5 at random again, and to the broken list
First management module request is operated to the broken cell data of this numbering 5；Broken Single Component Management module checks the broken cell-like of numbering 5
State, the status display of the broken unit is vacant, then the broken location mode is changed to task and taken, and data manipulation authority is given
Task A treatment；Task A is to carry out treatment operation to the broken unit of numbering 5.
Specifically, Fig. 3 shows a flow for task operating data.As shown in figure 3, task A is to described shared
Before data are operated, the total quantity of broken unit is obtained by broken Single Component Management module first；Acquirement one is broken at random for task A
The numbering 5 of unit, and the broken cell data of this numbering 5 is operated to the broken Single Component Management module request；The broken list
First management module checks the broken location mode of corresponding numbering 5, and determining the broken cell data of this numbering 5 does not have task to process, then
The broken location mode is changed to task occupancy, and task A treatment is given data manipulation authority；Task A receives broken Single Component Management module
Judgement information after, according to the judgement information determine be that the broken cell data of current number 5 is processed；Task A is to working as
After the broken cell data of preceding numbering 5 completes operation treatment, the numbering of the broken unit is deleted from the total quantity of the broken unit for obtaining,
And data have been disposed in notifying the broken unit of the broken Single Component Management module numbering 5；Broken Single Component Management module is to this
The status information of the broken unit of numbering 5 is updated as vacant, in order to other task calls；Task A updates untreated
The total quantity of broken unit, obtains a numbering for broken unit at random again, aforesaid operations is repeated, if the broken unit quilt of the numbering
Take, then mode shown in reference picture 2 is performed, until completing the treatment of whole shared data.
If in above-mentioned handling process, the total quantity of the untreated broken units of task A less than setting value (such as 10, build
5) view, data processing request, broken Single Component Management module is sent when a broken element number is obtained at random to broken Single Component Management module
Return information show this numbering broken cell data be in other tasks take in, then postpone a period of time (depending on according to circumstances,
Generally 1-3 seconds) again it is random obtain one and number send request to broken Single Component Management module.Task can be so avoided only to be left
A small amount of broken cell data is untreated, and the broken unit is in other task seizure conditions, and frequent requests can give broken unit pipes
Reason module causes server stress to increase, and efficiency is reduced on the contrary.
A kind of special task treatment mode of operation is described below, as shown in Figure 4.If in many task modules same
When broken Single Component Management module is carried out data interaction operate when, the operation requests that a task B sends to broken Single Component Management module
Not comprising broken element number, then show that task B increases to broken Single Component Management module request shared data.Now, broken Single Component Management
Module searches are in the minimum broken unit of free state data total amount, and set the broken unit for task seizure condition, and handle
This broken unit gives task B operations, after task B completes shared data increase operation, notifies broken Single Component Management module end of operation,
Should be free state this broken location mode after broken Single Component Management module is connected to instruction, and update this broken unit information, be easy to
Other tasks are operated to the data.
To sum up, the features of the present invention is mainly reflected in following 6 key elements., mass data is divided into some sections at first point.
Second point, it is ensured that multitask is operated to the data simultaneously, after waiting a task to be disposed, next task
The start to process data, reach efficient purpose.Thirdly, task random number processing method, it is ensured that the same time behaviour of multitask
The probability for making the same broken unit of numbering is very low.4th point, the quantity of broken unit should be five times or so of task quantity, efficiency compared with
It is high.5th point, broken Single Component Management module will ensure broken unit operation thread-safe, it is ensured that same broken unit will not be by two tasks
Process simultaneously, it is ensured that its security.6th point, even if certain the broken cell processing data generation of task treatment mass data is abnormal,
Simply the treatment of this one piece of data is problematic, causes whole data exception rather than because of abnormal, then process again, it is only necessary to again
Process the broken unit and include data, so as to reach the shared big data purpose of multitask efficient process.
Obviously, those skilled in the art can carry out various changes and modification without deviating from essence of the invention to the present invention
God and scope.So, if these modifications of the invention and modification are belonged to the model of the claims in the present invention and its equivalent technology
Within enclosing, then the present invention is also intended to comprising these changes and modification.
Claims (10)
1. a kind of different task travels through the high-efficient treatment method of a large amount of shared datas, comprises the following steps：
(1) shared data is split into several broken units, and sets up broken Single Component Management module；
(2) task A obtained the total of broken unit by broken Single Component Management module first before being operated to the shared data
Quantity；
(3) task A obtains a numbering for broken unit, and the broken list numbered to this to the broken Single Component Management module request at random
Metadata is operated；
(4) the broken Single Component Management module checks the broken location mode of corresponding numbering, and judges whether task A can be to the broken unit
Carry out operation treatment；
(5) after task A receives the judgement information of broken Single Component Management module, determine it is to current number according to the judgement information
Broken cell data is processed, still the new broken element number of random request one again；If necessary to random request one again
Individual new broken element number, then return to step (3)；
(6) after task A is to being currently able to the completion operation treatment of the broken cell data for the treatment of, from the total quantity of the broken unit for obtaining
The numbering of the broken unit is deleted, and data have been disposed in notifying the broken unit of the broken Single Component Management module numbering；
Broken Single Component Management module is updated to the status information of the broken unit of the numbering；
(7) task A updates the total quantity of untreated broken unit, and return to step (3) is circulated operation, until completing whole
The treatment of shared data.
2. different task as claimed in claim 1 travels through the high-efficient treatment method of a large amount of shared datas, it is characterised in that：Step
(4) broken Single Component Management module described in checks the broken location mode of corresponding numbering, it is determined whether have other tasks operating this to compile
Number broken cell data, if this numbering broken cell data do not there is task to process, the broken location mode be changed to task take,
Task A treatment is given data manipulation authority；If this broken unit is occupied state, the task A broken cell datas are notified
Taken by other tasks.
3. different task as claimed in claim 2 travels through the high-efficient treatment method of a large amount of shared datas, it is characterised in that：Step
(5) if the judgement presentation of information that task A receives broken Single Component Management module return has other tasks to corresponding numbering in
Broken cell data is taken, then the new broken element number of random request one again.
4. different task as claimed in claim 1 travels through the high-efficient treatment method of a large amount of shared datas, it is characterised in that：If
The total quantity of the untreated broken units of task A is less than setting value in step (7), and a broken element number is obtained at random to broken unit
Management module sends data processing request, and broken Single Component Management module return information shows that the broken cell data of this numbering is in other
During task takes, then postpone a period of time again it is random obtain one and number send request to broken Single Component Management module.
5. different task as claimed in claim 4 travels through the high-efficient treatment method of a large amount of shared datas, it is characterised in that：It is described
Setting value take 5, described delay is for a period of time 1-3 seconds.
6. the different task as described in any one in claim 1-5 travels through the high-efficient treatment method of a large amount of shared datas, its
It is characterised by：Multiple tasks carry out the operation of the step (2)-(7) simultaneously.
7. different task as claimed in claim 6 travels through the high-efficient treatment method of a large amount of shared datas, it is characterised in that：If
Task B does not include broken element number to the operation requests that broken Single Component Management module sends, then show task B to broken Single Component Management mould
Block request shared data increases, and now, broken Single Component Management module searches are in the minimum broken unit of free state data total amount
Setting task seizure condition, and task B operations are given this broken unit, after task B completes shared data increase operation, notify broken
Single Component Management module end of operation, should be free state this broken location mode after broken Single Component Management module is connected to instruction, and
This broken unit information is updated, is easy to other tasks to operate the data.
8. a kind of different task travels through the efficient process system of a large amount of shared datas, it is characterised in that：Including broken Single Component Management mould
Block and several task modules, shared data is split into several broken units by described broken Single Component Management module, to broken unit
Additions and deletions are carried out, and each broken unit is numbered, preserve the information of broken unit；Several described task modules can be simultaneously
Operation treatment is carried out to different broken cell datas；Broken Single Component Management module is according to each task module to different broken cell datas
Result changes broken unit information, and it is that the broken unit for not having task to process is carried out at operation to state only to allow task module
Reason.
9. different task as claimed in claim 8 travels through the efficient process system of a large amount of shared datas, it is characterised in that：It is described
Broken unit information including broken unit total quantity, the state of each broken unit, the total amount of each broken cell data.
10. different task as claimed in claim 8 or 9 travels through the efficient process system of a large amount of shared datas, it is characterised in that：
The total quantity of described broken unit is more than five times of task module quantity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611087441.5A CN106844020A (en) | 2016-12-01 | 2016-12-01 | Different task travels through the high-efficient treatment method and system of a large amount of shared datas |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611087441.5A CN106844020A (en) | 2016-12-01 | 2016-12-01 | Different task travels through the high-efficient treatment method and system of a large amount of shared datas |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106844020A true CN106844020A (en) | 2017-06-13 |
Family
ID=59145443
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611087441.5A Pending CN106844020A (en) | 2016-12-01 | 2016-12-01 | Different task travels through the high-efficient treatment method and system of a large amount of shared datas |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106844020A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100095303A1 (en) * | 2008-10-09 | 2010-04-15 | International Business Machines Corporation | Balancing A Data Processing Load Among A Plurality Of Compute Nodes In A Parallel Computer |
| CN101753608A (en) * | 2008-12-09 | 2010-06-23 | 中国移动通信集团公司 | Dispatching method and system of distributed system |
| CN102104617A (en) * | 2010-11-30 | 2011-06-22 | 厦门雅迅网络股份有限公司 | Method for storing massive picture data by website operating system |
| CN102170466A (en) * | 2011-03-29 | 2011-08-31 | 中国人民解放军国防科学技术大学 | Data processing method and system |
| CN103677752A (en) * | 2012-09-19 | 2014-03-26 | 腾讯科技（深圳）有限公司 | Distributed data based concurrent processing method and system |
| US20160321777A1 (en) * | 2014-06-20 | 2016-11-03 | Tencent Technology (Shenzhen) Company Limited | Data parallel processing method and apparatus based on multiple graphic processing units |
-
2016
- 2016-12-01 CN CN201611087441.5A patent/CN106844020A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100095303A1 (en) * | 2008-10-09 | 2010-04-15 | International Business Machines Corporation | Balancing A Data Processing Load Among A Plurality Of Compute Nodes In A Parallel Computer |
| CN101753608A (en) * | 2008-12-09 | 2010-06-23 | 中国移动通信集团公司 | Dispatching method and system of distributed system |
| CN102104617A (en) * | 2010-11-30 | 2011-06-22 | 厦门雅迅网络股份有限公司 | Method for storing massive picture data by website operating system |
| CN102170466A (en) * | 2011-03-29 | 2011-08-31 | 中国人民解放军国防科学技术大学 | Data processing method and system |
| CN103677752A (en) * | 2012-09-19 | 2014-03-26 | 腾讯科技（深圳）有限公司 | Distributed data based concurrent processing method and system |
| US20160321777A1 (en) * | 2014-06-20 | 2016-11-03 | Tencent Technology (Shenzhen) Company Limited | Data parallel processing method and apparatus based on multiple graphic processing units |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109873904B (en) | User message state reporting processing method, device and storage medium for message pushing | |
| CN102473122B (en) | fault-tolerant batch processing | |
| CN112446025A (en) | Federal learning defense method and device, electronic equipment and storage medium | |
| CN107608773A (en) | task concurrent processing method, device and computing device | |
| CN107590277A (en) | Method of data synchronization, device, electronic equipment and storage medium | |
| CN111984426A (en) | Task scheduling method and device, electronic equipment and storage medium | |
| CN112559535A (en) | Multithreading-based asynchronous task processing method, device, equipment and medium | |
| CN103716384A (en) | Method and device for realizing cloud storage data synchronization in cross-data-center manner | |
| CN111694843B (en) | Missing number detection method and device, electronic equipment and storage medium | |
| WO2023040145A1 (en) | Artificial intelligence-based text classification method and apparatus, electronic device, and medium | |
| CN114169303A (en) | Method, device, equipment and medium for editing table based on vue.js | |
| CN113222408A (en) | Online inquiry service monitoring method, device, equipment and storage medium | |
| CN106844020A (en) | Different task travels through the high-efficient treatment method and system of a large amount of shared datas | |
| CN113849520B (en) | Intelligent recognition method and device for abnormal SQL, electronic equipment and storage medium | |
| CN107622059A (en) | A kind of method and system for improving database search efficiency | |
| CN115495620A (en) | Data management method based on graph structure and related equipment | |
| CN113449037B (en) | AI-based SQL engine calling method, device, equipment and medium | |
| CN110119301A (en) | A kind of virtual machine migration method and system | |
| CN114626103A (en) | Data consistency comparison method, device, equipment and medium | |
| CN114490137A (en) | Service data real-time statistical method and device, electronic equipment and readable storage medium | |
| CN113220992A (en) | Information flow content recommendation method, system and medium | |
| CN113076069A (en) | Double-cache method and device, electronic equipment and computer readable storage medium | |
| US10747626B2 (en) | Method and technique of achieving extraordinarily high insert throughput | |
| CN106250243B (en) | The processing method and processing device of banking system application based on poll tupe | |
| CN111143456A (en) | Spark-based Cassandra data import method, device, equipment and medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170613 |