CN104915253A - Work scheduling method and work processor - Google Patents
Work scheduling method and work processor Download PDFInfo
- Publication number
- CN104915253A CN104915253A CN201410090668.XA CN201410090668A CN104915253A CN 104915253 A CN104915253 A CN 104915253A CN 201410090668 A CN201410090668 A CN 201410090668A CN 104915253 A CN104915253 A CN 104915253A
- Authority
- CN
- China
- Prior art keywords
- task
- execution time
- queue
- processor
- job
- 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.)
- Granted
Links
Landscapes
- Debugging And Monitoring (AREA)
- Multi Processors (AREA)
Abstract
The invention discloses a work scheduling method and a work processor. The method comprises the steps that when new work is received, the latest executing time of the new work is determined according to the arrival time and the priority of the new work, and the new work is added to a work cache queue according to the latest executing time; the actual executing time of all pieces of work is determined according to the latest executing time of all the pieces of work in the work cache queue; according to the actual executing time of all the pieces of work in the work cache queue, a work executing queue is established, and task processors are allocated for tasks obtained by decomposing all the pieces of work in the work executing queue. By means of the work scheduling method and the work processor, the response time of the work with the high priority can be ensured, and the waiting time of a user is shortened.
Description
Technical field
The present invention relates to the dispatching technique of the communications field, particularly relate to a kind of method and job processor of job scheduling.
Background technology
Map stipulations (Map Reduce) technology obtains most large-scale IT company extensive utilization as the typical apply framework of cloud computing, Map Reduce simplifies the concurrent software programming mode on the super large cluster that is made up of common computer, can be used for the parallel computation of large-scale dataset.At present, the job scheduling in Map Reduce distributed computing system mainly contains following several mode: first-in first-out (FIFO, First Input First Output) method, HOD method and fair scheduling algorithm.Wherein, described fair scheduling algorithm has Fairshare method and minimum shared amount method; Described Fairshare method is: when an independent operation operationally, it will use whole cluster; After having new user's submit job, task groove (task slot) can be assigned to these new users by system, makes each user can obtain the cpu resource of roughly equivalent.Described minimum shared amount method is: each pond arranges the task groove number of a minimum shared amount, when dispatching, scheduler can guarantee that each pond can obtain its minimum shared amount whenever necessary; When the minimum shared amount in a pond does not use completely, idle task groove can be assigned to other pond.
Although what fair scheduling algorithm can ensure all user fairness obtains computational resource, but, the execution of a large amount of little task easily causes a queue scheduling problem: little job input data small scale, and on large-scale cluster, to account for the ratio of whole cluster less for the node of its Data distribution8.When little operation is scheduled for the head of scheduling queue according to minimum shared amount fair scheduling algorithm, whichever node have an idle groove all to require little Hand up homework task goes to run, thus cause the data locality of little task poor.In addition, fair scheduling algorithm also may cause groove viscous problem.Groove viscous problem is garrisoned for a long time by large operation and is caused in the cluster.When adopting fair scheduling algorithm, there will be same groove cycle assignment to the trend of same task.This just causes operation always operation task on identical groove, cannot ensure good data locality, the throughput of influential system.
Summary of the invention
In view of this, the object of the present invention is to provide a kind of method and job processor of job scheduling, in conjunction with the execution time of priority determination operation, the stand-by period of user can be reduced.
Embodiments provide a kind of method of job scheduling, described method comprises:
When receiving new operation, determine the execution time at the latest of described new operation according to time of arrival of described new operation and priority, and according to described new operation is added into operation buffer queue by the execution time at the latest;
According to the execution time at the latest of each operation in operation buffer queue, determine the actual execution time of each operation;
According to the actual execution time of each operation in described operation buffer queue, setting up Job execution queue, is the task matching task processor that each breakdown of operation in described Job execution queue obtains.
In such scheme, described basis is after described new operation is added into operation buffer queue by the execution time at the latest, and described method also comprises:
Judge whether there is the execution time at the latest of any one operation in described operation buffer queue early than current time, if so, then the execution time at the latest of described operation is postponed to specify duration.
In such scheme, describedly determine that the actual execution time of each operation comprises:
Utilize the execution time at the latest of each operation in operation buffer queue, and the resource requirement of each operation, determine the actual execution time of each operation.
In such scheme, described is the task matching task processor that in Job execution queue, each breakdown of operation obtains, and comprising:
According to the order preset, operation is extracted as current work successively from described Job execution queue, for described current work chooses the task processor meeting preset state, after described current work is decomposed at least one task, at least one task described is sent to the described task processor chosen; Wherein, described preset state be described task processor be in idle condition and computing power higher than preset value.
In such scheme, before the described task matching task processor obtained for each breakdown of operation in Job execution queue, described method also comprises:
Each task processor of real-time collecting to the execution result of each task, and adds up the execution time of each each task of task processor process;
According to the statistics of the execution time of each each task of task processor process, determine the computing power that each processor each is corresponding.
The embodiment of the present invention additionally provides a kind of job processor, and described job processor comprises: operation buffer unit, queue organization unit, job execution unit; Wherein,
Described operation buffer unit, for when receiving new operation, determines the execution time at the latest of described new operation according to time of arrival of described new operation and priority, and according to described new operation is added into operation buffer queue by the execution time at the latest;
Described queue organization unit, for the execution time at the latest according to each operation in operation buffer queue, determines the actual execution time of each operation, according to the actual execution time of each operation in described operation buffer queue, sets up Job execution queue;
Described job execution unit, for the task matching task processor obtained for each breakdown of operation in described Job execution queue.
In such scheme, whether described queue organization unit, also have the execution time at the latest of any one operation early than current time for judging in described operation buffer queue, if so, then the execution time at the latest of described operation postponed to specify duration.
In such scheme, described queue organization unit, specifically for utilizing the execution time at the latest of each operation in operation buffer queue, and the resource requirement of each operation, determine the actual execution time of each operation.
In such scheme, described job execution unit, for according to preset order, operation is extracted as current work successively from described Job execution queue, for described current work chooses the task processor meeting preset state, after described current work is decomposed at least one task, at least one task described is sent to the described task processor chosen; Wherein, described preset state be described task processor be in idle condition and computing power higher than preset value.
In such scheme, described job execution unit, also for each task processor of real-time collecting to the execution result of each task, and add up the execution time of each each task of task processor process; According to the statistics of the execution time of each each task of task processor process, determine the computing power that each processor each is corresponding.
The method of job scheduling provided by the present invention and job processor, can when the process of establishing of operation buffer queue, determines its execution time at the latest according to the priority of each operation; Determine the actual execution time of operation again according to the execution time at the latest of each operation in operation buffer queue, ensure that each operation in operation buffer queue all performed before its execution time at the latest; So, the execution time to key operation and the higher operation of priority can be ensured, reduce period of reservation of number.
Accompanying drawing explanation
Fig. 1 is the method flow schematic diagram of embodiment of the present invention job scheduling;
Fig. 2 is the composition structural representation of embodiment of the present invention job processor;
Fig. 3 is the exemplary system figure that the embodiment of the present invention uses job processor composition.
Embodiment
Below in conjunction with drawings and the specific embodiments, the present invention is further described in more detail.
Embodiment one,
The method of the job scheduling that the embodiment of the present invention proposes, as shown in Figure 1, comprising:
Step 101: when receiving new operation, determines the execution time at the latest of described new operation according to time of arrival of described new operation and priority, and according to described new operation is added into operation buffer queue by the execution time at the latest.
Here, described priority for arranging list in client (Client) according to the priority preset, can obtain the priority of the operation that will upload; Wherein, described priority arrange list can according to actual conditions preset.
This step can be complete by job trace device of the prior art (Job Tracker), such as: when job trace device (Job Tracker) receives the operation of i-th arrival, and the time of arrival of record i-th operation; According to time of arrival and the priority of i-th operation, calculate the execution time at the latest of i-th operation.
Describedly determine that the execution time at the latest of described new operation comprises: by the time of arrival of described new operation and Timing Advance sum corresponding to priority, as the execution time at the latest of described new operation;
Wherein, Timing Advance corresponding to described priority can be: utilize described priority to search corresponding time value from the priority corresponding lists preset, using the described time value that finds as Timing Advance.
According to execution time at the latest, described new operation is added into operation buffer queue to comprise: according to the sequencing of execution time at the latest corresponding to operation whole in current work buffer queue, new described operation is added in described operation buffer queue and obtains the operation buffer queue after upgrading.
Preferably, after above-mentioned steps 101, can also comprise: judge whether there is the execution time at the latest of any one operation in described operation buffer queue early than current time, if so, then the execution time at the latest of the operation early than current time be postponed to specify duration; Otherwise, perform step 102.
Wherein, described appointment duration can be a period of time length preset according to actual conditions, can be such as 1 second or 10 milliseconds.
Preferably, after the execution time at the latest of described operation is postponed to specify duration, can also comprise: whether the quantity judging the described execution time at the latest to postpone the operation of specifying duration, exceed default threshold value, if so, then carry out early warning; Otherwise, perform step 102.
Step 102: according to the execution time at the latest of each operation in operation buffer queue, determine the actual execution time of each operation.
Here, described actual execution time is early than the described execution time at the latest.
The described actual execution time determining each operation, Ke Yiwei: the execution time at the latest utilizing each operation in operation buffer queue, and the resource requirement of each operation, determine the actual execution time of each operation; Such as, suppose that i-th operation is represented by Ji, operation buffer queue is { J
0, J
1, J
2, J
3j
n, the execution time at the latest of Ji is Tdi; Ji resource requirement distribution situation is represented by Si; The actual execution time of Ji is Tsi; I represents the order of arrival of operation in operation buffer queue; Tsi<Tdi;
The selectable value of Tsi exhaustively can comprise Tsi0, Tsi1, Tsi2 ... Tsin;
One by one from operation buffer queue the actual execution time of all tasks exhaustive value in choose the Ts value of each task, utilize formula Tsum=Ts1+Ts2+Ts3+ ... + Tsn calculates the actual execution time sum that various combination obtains, and chooses the minimum value of actual execution time sum; And ensure the t guarantee at any time of each task in this combination S0+S1+S2+S3+ ... + Sn<S
total, S
totalrepresent that system resource meets the not overlapping distribution of each operation resource S; Finally solve and obtain operation buffer queue optimal execution time set F={Ts0, Ts1, Ts2, Ts3 ... Tsn}, make actual execution time sum minimum, realize ensureing that the task node of each operation in combination is not overlapping, and under each task all do not surmount the prerequisite of respective execution time Td, obtain job queue result.
Step 103: according to the actual execution time of each operation in described operation buffer queue, set up Job execution queue is the task matching task processor that each breakdown of operation in described Job execution queue obtains.
Here, each operation described reconfigures and obtains Job execution queue and can be by the actual execution time of each operation described in described basis: according to the sequencing of described actual execution time, each operation described is arranged, each operation after arrangement is formed described Job execution queue.
The described task matching task processor being in described Job execution queue, each breakdown of operation obtains comprises: according to the order preset, operation is extracted as current work successively from described Job execution queue, for described current work chooses the task processor meeting preset state, after described current work is decomposed at least one task, at least one task described is sent to the described task processor chosen; Wherein, described preset state be described task processor be in idle condition and computing power higher than preset value.
Wherein, determine that the method that described task processor is in idle condition is: after completing described step 103, when receiving the task request message that any one task processor is sent, determine that described task processor is in idle condition.
The defining method of the computing power of described task processor is: each task processor of real-time collecting to the execution result of each task, and adds up the execution time of each each task of task processor process; According to the statistics of the execution time of each each task of task processor process, determine the computing power that each processor each is corresponding.
Wherein, the statistics of the described execution time according to each task of each task processor process, calculating computing power corresponding to each task processor can for periodically performing, also for utilizing the current whole statisticses got, the computing power that each task processor is corresponding can periodically be calculated.
After adopting such scheme, can ensure that task performs at task processor place, data place, reduce Internet Transmission burden; And can ensure that in job queue, each operation all performed before its execution time at the latest; Utilize Nash Equilibrium Theory, perform if many breakdown of operation gained task can be distributed in different task processor at synchronization, then these operations can executed in parallel.Thus improve the job parallelism implementation rate of overall Map Reduce system to greatest extent, and overall calculation capacity utilization ratio; By applying Nash Equilibrium Theory in queue anabolic process, and task priority Time Calculation method, be advanced by the response time of mission critical, decrease period of reservation of number;
In addition, by the statistics task execution time, periodically assessment TaskTracker Mission Capability, facilitates the performance of dynamic monitoring system.
Embodiment two,
The job processor of the method for the job scheduling of the execution embodiment one that the embodiment of the present invention provides comprises as shown in Figure 2: operation buffer unit, queue organization unit, job execution unit; Wherein,
Described operation buffer unit, for when receiving new operation, determines the execution time at the latest of described new operation according to time of arrival of described new operation and priority, and according to described new operation is added into operation buffer queue by the execution time at the latest;
Described queue organization unit, for the execution time at the latest according to each operation in operation buffer queue, determines the actual execution time of each operation, according to the actual execution time of each operation in described operation buffer queue, sets up Job execution queue;
Described job execution unit, for the task matching task processor obtained for each breakdown of operation in described Job execution queue.
Described priority can be: each client (Client) arranges list according to the priority preset, and obtains the priority of the operation that will upload; Wherein, described priority arrange list can according to actual conditions preset.
The time of arrival of the operation that described record i-th arrives; According to the time of arrival of i-th operation and the priority of i-th operation, calculate the execution time at the latest of task.
Described operation buffer unit, specifically for by the time of arrival of described new operation and Timing Advance sum corresponding to priority, as the execution time at the latest of described new operation;
Wherein, Timing Advance corresponding to described priority can be: utilize described priority to search corresponding time value from the priority corresponding lists preset, using the described time value that finds as Timing Advance.
Described operation buffer unit, specifically for the sequencing according to execution time at the latest corresponding to operation whole in current work buffer queue, is added into new described operation in described operation buffer queue and obtains the operation buffer queue after upgrading.
Whether institute's operation buffer unit, also have the execution time at the latest of any one operation early than current time for judging in described operation buffer queue, if so, then the execution time at the latest of the operation early than current time postponed to specify duration; Otherwise, according to the execution time at the latest of each operation in the operation buffer queue after described renewal and the process resource of each operation, calculate the actual execution time of each operation.Wherein, described appointment duration can be a period of time length preset according to actual conditions, can be such as 1 second or 10 milliseconds.
Preferably, whether described operation buffer unit, also for judging the quantity being postponed the operation of specifying duration the described execution time at the latest, exceeding default threshold value, if so, then carrying out early warning; Otherwise, according to the execution time at the latest of each operation in the operation buffer queue after described renewal and the process resource of each operation, calculate the actual execution time of each operation.
Described queue organization unit, specifically for utilizing the execution time at the latest of each operation in operation buffer queue, and the resource requirement of each operation, when determining the actual execution of each operation.
Described job execution unit, specifically for the sequencing according to described actual execution time, arranges each operation described, and each operation after arrangement is formed described Job execution queue.
Described job execution unit, specifically for according to preset order, operation is extracted as current work successively from described Job execution queue, for described current work chooses the task processor meeting preset state, after described current work is decomposed at least one task, at least one task described is sent to the described task processor chosen; Wherein, described preset state be described task processor be in idle condition and computing power higher than preset value.
Described job execution unit, specifically for when receiving the task request message that any one task processor is sent, determines that described task processor is in idle condition.
Described job execution unit, also for each task processor of real-time collecting to the execution result of each task, and add up the execution time of each task of each task processor process; According to the statistics of the execution time of each task of each task processor process, calculate the computing power that each task processor is corresponding
Preferably, the job processor that the embodiment of the present invention provides may be used for, in MapReduce job scheduling system of the prior art, as shown in Figure 3, comprising: client, job processor, one or more task processor.
Client, for carrying out priority mark to job request;
Job processor, for when receiving new operation, determines the execution time at the latest of described new operation according to time of arrival of described new operation and priority, and according to described new operation is added into operation buffer queue by the execution time at the latest; According to the execution time at the latest of each operation in operation buffer queue, determine the actual execution time of each operation, according to the actual execution time of each operation in described operation buffer queue, set up Job execution queue; For the task matching task processor that each breakdown of operation in described Job execution queue obtains;
Task processor, for the treatment of receiving of task.
When there not being new operation to arrive, in job queue, each operation expection execution time is determined, in queue, each operation is by by the time determined and order, is sent to Job execution device and performs.Now, in job queue, each operation is in mutual containing, and queue is in Nash Equilibrium state.
When there being new operation to arrive, the mutual containing state of each operation in job queue Central Plains is broken.In job queue, each operation will be vied each other, and while guaranteeing oneself the execution time is met at the latest, meet the execution time at the latest of other operations to greatest extent, and will vie each other, strive for the Job execution time obtained as early as possible.
When All Jobs agrees to the Job execution time of oneself, new Nash Equilibrium state produces.This state will keep, until there is new operation to arrive.
The above, be only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.
Claims (10)
1. a method for job scheduling, is characterized in that, described method comprises:
When receiving new operation, determine the execution time at the latest of described new operation according to time of arrival of described new operation and priority, and according to described new operation is added into operation buffer queue by the execution time at the latest;
According to the execution time at the latest of each operation in operation buffer queue, determine the actual execution time of each operation;
According to the actual execution time of each operation in described operation buffer queue, setting up Job execution queue, is the task matching task processor that each breakdown of operation in described Job execution queue obtains.
2. method according to claim 1, is characterized in that, described basis is after described new operation is added into operation buffer queue by the execution time at the latest, and described method also comprises:
Judge whether there is the execution time at the latest of any one operation in described operation buffer queue early than current time, if so, then the execution time at the latest of described operation is postponed to specify duration.
3. method according to claim 1, is characterized in that, describedly determines that the actual execution time of each operation comprises:
Utilize the execution time at the latest of each operation in operation buffer queue, and the resource requirement of each operation, determine the actual execution time of each operation.
4. the method according to any one of claim 1-3, is characterized in that, described is the task matching task processor that in Job execution queue, each breakdown of operation obtains, and comprising:
According to the order preset, operation is extracted as current work successively from described Job execution queue, for described current work chooses the task processor meeting preset state, after described current work is decomposed at least one task, at least one task described is sent to the described task processor chosen; Wherein, described preset state be described task processor be in idle condition and computing power higher than preset value.
5. method according to claim 4, is characterized in that, before the described task matching task processor obtained for each breakdown of operation in Job execution queue, described method also comprises:
Each task processor of real-time collecting to the execution result of each task, and adds up the execution time of each each task of task processor process;
According to the statistics of the execution time of each each task of task processor process, determine the computing power that each processor each is corresponding.
6. a job processor, is characterized in that, described job processor comprises: operation buffer unit, queue organization unit, job execution unit; Wherein,
Described operation buffer unit, for when receiving new operation, determines the execution time at the latest of described new operation according to time of arrival of described new operation and priority, and according to described new operation is added into operation buffer queue by the execution time at the latest;
Described queue organization unit, for the execution time at the latest according to each operation in operation buffer queue, determines the actual execution time of each operation, according to the actual execution time of each operation in described operation buffer queue, sets up Job execution queue;
Described job execution unit, for the task matching task processor obtained for each breakdown of operation in described Job execution queue.
7. job processor according to claim 6, is characterized in that,
Whether described queue organization unit, also have the execution time at the latest of any one operation early than current time for judging in described operation buffer queue, if so, then the execution time at the latest of described operation postponed to specify duration.
8. job processor according to claim 6, is characterized in that,
Described queue organization unit, specifically for utilizing the execution time at the latest of each operation in operation buffer queue, and the resource requirement of each operation, determine the actual execution time of each operation.
9. the job processor according to any one of claim 6-8, is characterized in that,
Described job execution unit, for according to preset order, operation is extracted as current work successively from described Job execution queue, for described current work chooses the task processor meeting preset state, after described current work is decomposed at least one task, at least one task described is sent to the described task processor chosen; Wherein, described preset state be described task processor be in idle condition and computing power higher than preset value.
10. job processor according to claim 9, is characterized in that,
Described job execution unit, also for each task processor of real-time collecting to the execution result of each task, and add up the execution time of each each task of task processor process; According to the statistics of the execution time of each each task of task processor process, determine the computing power that each processor each is corresponding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410090668.XA CN104915253B (en) | 2014-03-12 | 2014-03-12 | A kind of method and job processor of job scheduling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410090668.XA CN104915253B (en) | 2014-03-12 | 2014-03-12 | A kind of method and job processor of job scheduling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104915253A true CN104915253A (en) | 2015-09-16 |
| CN104915253B CN104915253B (en) | 2019-05-10 |
Family
ID=54084334
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410090668.XA Active CN104915253B (en) | 2014-03-12 | 2014-03-12 | A kind of method and job processor of job scheduling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104915253B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106358054A (en) * | 2015-07-14 | 2017-01-25 | 杭州海康威视数字技术股份有限公司 | Method and system for analyzing cluster video |
| CN106790204A (en) * | 2017-01-03 | 2017-05-31 | 北京奇虎科技有限公司 | The conversion method and device of a kind of tunneling of flow data |
| CN106970874A (en) * | 2017-01-22 | 2017-07-21 | 阿里巴巴集团控股有限公司 | A kind of task processing method, device and electronic equipment |
| CN108052443A (en) * | 2017-10-30 | 2018-05-18 | 北京奇虎科技有限公司 | A kind of test assignment dispatching method, device, server and storage medium |
| CN109388482A (en) * | 2017-08-11 | 2019-02-26 | 中兴通讯股份有限公司 | Dispatching method, device and the storage medium of task |
| CN109408217A (en) * | 2018-11-13 | 2019-03-01 | 杭州数梦工场科技有限公司 | A kind of spark Runtime method of adjustment, device and equipment |
| CN110196761A (en) * | 2019-04-15 | 2019-09-03 | 北京达佳互联信息技术有限公司 | Delay task processing method and processing device |
| CN110333942A (en) * | 2019-07-05 | 2019-10-15 | 中国工商银行股份有限公司 | Electronic equipment and the task processing method executed by electronic equipment |
| CN111104168A (en) * | 2018-10-25 | 2020-05-05 | 杭州嘉楠耘智信息科技有限公司 | Calculation result submitting method and device |
| CN112685181A (en) * | 2020-12-29 | 2021-04-20 | 成都科来网络技术有限公司 | Push task scheduling method and system for balancing CPU resources |
| CN113138883A (en) * | 2021-05-14 | 2021-07-20 | 北京字节跳动网络技术有限公司 | Message processing method and device, storage medium and electronic equipment |
| CN116932227A (en) * | 2023-09-14 | 2023-10-24 | 西安华创马科智能控制系统有限公司 | Task scheduling method and device based on single thread |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101309208A (en) * | 2008-06-21 | 2008-11-19 | 华中科技大学 | Job scheduling system suitable for grid environment and based on reliable expense |
| CN102799957A (en) * | 2012-05-30 | 2012-11-28 | 武汉理工大学 | Scientific work flow scheduling method with safe perception under cloud calculation environment |
| CN103235640A (en) * | 2013-01-08 | 2013-08-07 | 北京邮电大学 | DVFS-based energy-saving dispatching method for large-scale parallel tasks |
| CN103473122A (en) * | 2013-08-21 | 2013-12-25 | 上海交通大学 | Workflow system resource scheduling method in cloud computing environment |
-
2014
- 2014-03-12 CN CN201410090668.XA patent/CN104915253B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101309208A (en) * | 2008-06-21 | 2008-11-19 | 华中科技大学 | Job scheduling system suitable for grid environment and based on reliable expense |
| CN102799957A (en) * | 2012-05-30 | 2012-11-28 | 武汉理工大学 | Scientific work flow scheduling method with safe perception under cloud calculation environment |
| CN103235640A (en) * | 2013-01-08 | 2013-08-07 | 北京邮电大学 | DVFS-based energy-saving dispatching method for large-scale parallel tasks |
| CN103473122A (en) * | 2013-08-21 | 2013-12-25 | 上海交通大学 | Workflow system resource scheduling method in cloud computing environment |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106358054A (en) * | 2015-07-14 | 2017-01-25 | 杭州海康威视数字技术股份有限公司 | Method and system for analyzing cluster video |
| CN106790204A (en) * | 2017-01-03 | 2017-05-31 | 北京奇虎科技有限公司 | The conversion method and device of a kind of tunneling of flow data |
| CN106970874A (en) * | 2017-01-22 | 2017-07-21 | 阿里巴巴集团控股有限公司 | A kind of task processing method, device and electronic equipment |
| CN109388482A (en) * | 2017-08-11 | 2019-02-26 | 中兴通讯股份有限公司 | Dispatching method, device and the storage medium of task |
| CN108052443A (en) * | 2017-10-30 | 2018-05-18 | 北京奇虎科技有限公司 | A kind of test assignment dispatching method, device, server and storage medium |
| CN111104168A (en) * | 2018-10-25 | 2020-05-05 | 杭州嘉楠耘智信息科技有限公司 | Calculation result submitting method and device |
| CN109408217A (en) * | 2018-11-13 | 2019-03-01 | 杭州数梦工场科技有限公司 | A kind of spark Runtime method of adjustment, device and equipment |
| CN110196761A (en) * | 2019-04-15 | 2019-09-03 | 北京达佳互联信息技术有限公司 | Delay task processing method and processing device |
| CN110196761B (en) * | 2019-04-15 | 2021-10-19 | 北京达佳互联信息技术有限公司 | Delayed task processing method and device |
| CN110333942A (en) * | 2019-07-05 | 2019-10-15 | 中国工商银行股份有限公司 | Electronic equipment and the task processing method executed by electronic equipment |
| CN112685181A (en) * | 2020-12-29 | 2021-04-20 | 成都科来网络技术有限公司 | Push task scheduling method and system for balancing CPU resources |
| CN112685181B (en) * | 2020-12-29 | 2024-06-04 | 科来网络技术股份有限公司 | Push task scheduling method and system for balancing CPU resources |
| CN113138883A (en) * | 2021-05-14 | 2021-07-20 | 北京字节跳动网络技术有限公司 | Message processing method and device, storage medium and electronic equipment |
| CN113138883B (en) * | 2021-05-14 | 2024-02-02 | 北京字节跳动网络技术有限公司 | Message processing method and device, storage medium and electronic equipment |
| CN116932227A (en) * | 2023-09-14 | 2023-10-24 | 西安华创马科智能控制系统有限公司 | Task scheduling method and device based on single thread |
| CN116932227B (en) * | 2023-09-14 | 2023-12-22 | 西安华创马科智能控制系统有限公司 | Task scheduling method and device based on single thread |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104915253B (en) | 2019-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104915253A (en) | Work scheduling method and work processor | |
| US9916183B2 (en) | Scheduling mapreduce jobs in a cluster of dynamically available servers | |
| Kato et al. | Semi-partitioned scheduling of sporadic task systems on multiprocessors | |
| CN104765640B (en) | A kind of intelligent Service dispatching method | |
| CN108984267B (en) | Micro-kernel architecture control system of industrial server and industrial server | |
| Xie et al. | Mixed real-time scheduling of multiple dags-based applications on heterogeneous multi-core processors | |
| CN110569252B (en) | Data processing system and method | |
| CN103235835A (en) | Inquiry implementation method for database cluster and device | |
| Albers et al. | On multi-processor speed scaling with migration | |
| CN112130966A (en) | Task scheduling method and system | |
| Moulik et al. | SEAMERS: A semi-partitioned energy-aware scheduler for heterogeneous multicore real-time systems | |
| Moulik | RESET: A real-time scheduler for energy and temperature aware heterogeneous multi-core systems | |
| CN108132840B (en) | Resource scheduling method and device in distributed system | |
| Yun et al. | An integrated approach to workflow mapping and task scheduling for delay minimization in distributed environments | |
| Agrawal et al. | Scheduling parallelizable jobs online to minimize the maximum flow time | |
| Ghouma et al. | Context aware resource allocation and scheduling for mobile cloud | |
| CN108153583A (en) | Method for allocating tasks and device, real-time Computational frame system | |
| Garibay-Martínez et al. | On the scheduling of fork-join parallel/distributed real-time tasks | |
| CN105159752A (en) | Real-time task and resource scheduling method with function of machine startup time perception in virtualized clouds | |
| CN106534312B (en) | A kind of service request selection of facing mobile apparatus and dispatching method | |
| Natarajan | Parallel queue scheduling in dynamic cloud environment using backfilling algorithm | |
| CN109189581B (en) | Job scheduling method and device | |
| CN109445847B (en) | Industrial server microkernel architecture implementation method | |
| Chishiro et al. | Optimal multiprocessor real-time scheduling based on run with voltage and frequency scaling | |
| Desirena-Lopez et al. | On-line scheduling in multiprocessor systems based on continuous control using timed continuous petri nets |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |