CN105740059B - A kind of population dispatching method towards Divisible task - Google Patents
A kind of population dispatching method towards Divisible task Download PDFInfo
- Publication number
- CN105740059B CN105740059B CN201410768968.9A CN201410768968A CN105740059B CN 105740059 B CN105740059 B CN 105740059B CN 201410768968 A CN201410768968 A CN 201410768968A CN 105740059 B CN105740059 B CN 105740059B
- Authority
- CN
- China
- Prior art keywords
- task
- subtask
- particle
- fitness
- population
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 239000002245 particle Substances 0.000 claims abstract description 71
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000005611 electricity Effects 0.000 claims description 6
- 230000002123 temporal effect Effects 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 5
- 230000002068 genetic effect Effects 0.000 description 3
- 238000005265 energy consumption Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Abstract
The present invention relates to a kind of population dispatching method towards Divisible task, it include: after task to be scheduled is divided into subtask, using the task allocation plan that is randomly generated as a particle, using the corresponding time performance of task allocation plan as the fitness of particle, the speed being mutually shifted between particle is calculated with the difference between particle fitness, multiple evolution is done to population, the best particle of fitness is selected from the result repeatedly evolved；Overhead value is finally combined, subtask scheduling is done in each subtask in task allocation plan corresponding to the particle best to fitness.
Description
Technical field
The present invention relates to computer networking technology, in particular to a kind of population dispatching method towards Divisible task.
Background technique
There are many common method for scheduling task, and a portion method is that task is considered as to the whole progress that can not be split
Scheduling, such methods can be listed below:
Min-Min algorithm predicts minimum completion of each task on each processor in current task queue first
Then task with minimum completion time is distributed to corresponding processor by the time, while updating corresponding processor just
The thread time is removed assigned task from task queue, such remaining task of duplicate allocation, until entire task queue is
It is empty.Easily there is load imbalance phenomenon in Min-Min algorithm.
Max-Min algorithm and Min-Min algorithm the difference is that, determining each task on each processor
After earliest finish time, the task with maximum earliest finish time is distributed into corresponding processor, and timely update
Corresponding processor ready time, reprocesses remaining task.Max-Min algorithm ratio in terms of load balancing
Min-Min algorithm makes moderate progress.
Promethee algorithm at task end, according to the customized standard of user (such as task scale, in current processor
On prediction execute one of used time, the indexs such as cost, be also possible to for many indexes to be weighted the synthesis that processing obtains
Performance indicator) pending task is subjected to priority ranking；At processor end, machine state is monitored in real time, once there is machine
There is idle state, is just sorted according to the task priority obtained in advance and the task of highest priority is assigned to current idle
Machine gets on.Emulation shows to suitably adjust the weight between each performance indicator, and algorithm can be made to realize various performances most
It is excellent.
Separately there is certain methods proposition that task is divided into multiple subtasks to be dispatched one by one, but the task object of analysis
It is only limitted to some specific tasks, be not related to high-volume task while being occurred, a variety of partitioning schemes and the situation deposited, this
Class method can be listed below:
Timing requirements between subtask are analyzed first to the genetic algorithm of timing correlator task Parallel Scheduling, to institute
There is the time depth value when execution of subtask to be ranked up.Then several " subtask-node " allocation matrix is generated at random, often
One kind " subtask-node " matrix is a kind of allocation plan.The thinking of algorithm is randomly generated several allocation plan and constitutes
Initial population, and to the individual in population carry out variation and screening operation, be allowed to by generation improve, thus obtain it is new, complete when
Between shorter scheme.By much for genetic algorithm after, it can be deduced that it is stable, preferably solve.But genetic algorithm is answered
Miscellaneous degree is higher, will cause very big calculation delay in the case that total task number is more in a network.
EDTS algorithm is the method for carrying out optimal scheduling for the N step task inside a task, and algorithm is predicted first
Each subtask executes the time it takes and energy consumption on all machines out, is then the total cut-off of this succession of task setting
Time, in conjunction with existing sequential relationship, finds out the subtask method of salary distribution as energy-efficient as possible under fixed total deadline,
But EDTS algorithm is split just for a task, is dispatched, and is accomplished that the best performance of a task itself, is worked as network
In when there is broad medium task, the mutual waiting time as caused by temporal constraint is longer between subtask, each task
Local optimum is contradictory with whole optimization.
Summary of the invention
It is an object of the invention to overcome, subtask scheduling method complexity in the prior art is high, calculation delay is big, energy
The defects of high is consumed, to provide a kind of subtask scheduling method of comprehensive measure time, energy consumption.
To achieve the goals above, the present invention provides a kind of population dispatching method towards Divisible task, comprising:
After task to be scheduled is divided into subtask, using the task allocation plan that is randomly generated as a particle, with
Fitness of the corresponding time performance of task allocation plan as particle is calculated between particle with the difference between particle fitness
The speed being mutually shifted does multiple evolution to population, and the best particle of fitness is selected from the result repeatedly evolved；Finally
In conjunction with overhead value, subtask scheduling is done in each subtask in task allocation plan corresponding to the particle best to fitness.
In above-mentioned technical proposal, this method is specifically included:
Step 1), the instruction strip number for collecting multiple subtasks after the instruction strip number of scheduler task, segmentation, subtask it
Between Temporal dependency relationship, the file block size of each task；
Step 2) acquires the speed of service MIPS of each server in cluster, unit time electricity charge expense CPS, currently bears
It carries, earliest idle moment EST, the bandwidth information between calculation server；
Step 3) will carry out descending sort according to instruction strip number to scheduler task in queue, for head of the queue wait dispatch
Task successively executes step 4)-step 10), until the needed scheduler task in queue has been processed into；
Step 4) is decomposed to current to scheduler task, obtains subtask structure chart, and to be described to scheduler task
N number of allocation plan is randomly generated, forms initial population；
Wherein, the subtask structure chart reflects the sequential relationship before being broken down into subtask to scheduler task；
The initial population includes the N number of particle P generated at random1,P2,...PN, each particle represents " a task-
Server " allocation plan, if the sum of server is M, each particle PnIt is expressed as a L*M matrix Sn: where 1≤n≤
The size of N, N are set according to the number of servers in cluster, the subtask number to include in scheduler task, and L is indicated wait adjust
The number of obtained subtask after degree Task-decomposing；
Step 5), the time performance Makespan for calculating N number of particle, obtain the fitness of N number of particle；
Step 6) calculates movement speed of the particle in new round iteration, and each particle is once evolved, is obtained down
Generation population；
Step 7) judges whether current evolution number is less than preset value, if so, re-executeing the steps 5), otherwise, executes
Step 8)；
Step 8) filters out the highest particle P of fitness from evolution results all beforebest；
Step 9), the corresponding weight P of setting overhead performance, carry out the task immigration based on expense Cost；
Step 10) exports the task immigration of step 9) as a result, carrying out subtask scheduling by this result.
In above-mentioned technical proposal, in step 5), time performance Makespan is calculated according to following principle:
It a, must be after the finish time of its all forerunner subtask at the beginning of subtask；
B, subtask must receive the output file of its all forerunner subtask as its input file；
C, subtask must start in the case where being currently located processor idle states；
D, subtask must start as early as possible in the case where meeting above-mentioned condition a-c condition；
E, subtask TaskjIt is Task_MI the time required to executingj/MIPSk；
F, a task complete time span Makespan be its first subtask at the beginning of with the last one
Time span between the finish time of subtask.
In above-mentioned technical proposal, in step 5), the fitness of n-th of particle is calculated according to following formula (2)
Fitnessn:
Fitnessn=max { Makespan1,Makespan2,...MakespanN}-MakespannFormula (2).
In above-mentioned technical proposal, the step 6) includes:
It is P by the maximum particle label of fitness valuebest, the corresponding allocation matrix of the particle is labeled as Sbest；
Define particle PnTo PbestMovement speed is Vn,best, then shown in for example following formula of its calculation formula (3):
Vn,best=(fitnessbest-fitnn)/fitnesszFormula (3)；
In the primary evolution of population, each particle PnAll with speed Vn,bestTo PbestShifting moves a step, and what is obtained is N number of new
Particle form next-generation population.
In above-mentioned technical proposal, the step 9) includes:
The best particle P of fitness in filtering out current particle groupbestAfterwards, for current subtask, it is calculated at it
The time performance and overhead performance executed on its any server, and the performance is calculated compared to PbestGain delta C and Δ FT,
Wherein C indicates the cost when executing on a certain server, and FT indicates the completion moment executed on a certain server, CbaseWith
FTbaseCalculated C and FT value before being task immigration；
For Servers-all, finds out and calculate the obtained maximum server of value according to the following formula (4), it then will be sub
On task assignment to the server:
P*ΔC/Cbase+(1-P)*ΔFT/FTbaseFormula (4).
The present invention has the advantages that
1, method of the invention has carried out the considerations of many-sided performance factor, when assessing scheduling scheme, not only considers it
Total time performance, and using indexs such as load balancing in the overhead of task execution, cluster as the factor considered, this to appoint
Business distribution is more reasonable, efficient；
2, the scene that method of the invention had both considered batch tasks request while having occurred, it is contemplated that between subtask
Temporal dependency, the cutting operation of subtask, which solves software license resource constraint bring on the virtual machine in cloud environment, to be limited to
Property.
Detailed description of the invention
Fig. 1 is the schematic diagram of network environment applied by the method for the present invention in one embodiment；
Fig. 2 is the schematic diagram of serial sequential relationship between subtask；
Fig. 3 is the schematic diagram of parallel sequential relationship between subtask；
Fig. 4 is the schematic diagram of mixed type sequential relationship between subtask；
Fig. 5 is the flow chart of the method for the present invention.
Specific embodiment
Now in conjunction with attached drawing, the invention will be further described.
Method of the invention is directed to batch, the alienable task requests that client is initiated, and is formed after the segmentation of consideration task
Subtask granularity and subtask between sequential relationship, and combine server cluster in different processor computing capability and function
The factors such as rate, Internet bandwidth are scheduled batch tasks, so that scheduling result is equal in total time and two aspect of charge costs
With preferable performance.
Before elaborating to method of the invention, an introduction is done to concept involved in the method for the present invention first.
Method of the invention should meet following three preconditions before use:
1, the scheduler in a region is responsible for the traffic control of all terminal tasks in the region, and the task that terminal is initiated needs
Wait in line the decision of scheduler；
2, the batch tasks handled by the present invention are alienable task (may be partitioned into multiple subtasks), and son is appointed
There is serial, parallel or mixed type sequential relationship between business.Forerunner subtask if it exists, one subtask, then need to wait for its institute
After the completion of having predecessor task, the subtask can be executed；
3, described in condition 2 as in the previous, meet the necessary condition that the sequential relationship between subtask is subtask operation, when
When one subtask and its some forerunner subtask are assigned to different server execution, needing will by internet or local area network
Server is as input file where the output file of forerunner subtask is transmitted to subsequent subtask；If some subtask and its before
It drives subtask and is assigned to same server, be then not required to carry out file transmission.
It is in figs. 2,3 and 4 respectively serial sequential relationship between subtask described in condition 2, parallel sequential relationship, mixed
The schematic diagram of mould assembly sequential relationship.
Method of the invention pursues following two performance indicators:
1, the overall deadline of current pending batch tasks；
2, overhead needed for executing batch tasks, including charge on traffic and the processor electricity charge.
Method of the invention mainly considers following three factors in performance evaluation:
1, influence of the different CPU for executing speed for task completion time；
2, influence of the network bandwidth that network topology structure is determined to I/O file propagation delay time；
3, influence of the CPU of different capacity for power consumption and the corresponding electricity charge.
Symbol involved in the method for the present invention is defined as follows:
Jobi: i-th of task；
Taskj: j-th of subtask；
Serverk: k-th of server；
Job_MIi: JobiInstruction strip number (million)；
Makespani: JobiTerminate practical time delay experienced from initiating to executing；
Task_MIj: TaskiInstruction strip number (million)；
MIPSk: ServerkThe instruction strip number (million/second) of operation per second；
CPSk: ServerkExpense per second in the operational mode；
ECTk: ServerkEarliest idle moment.
It in the embodiment shown in fig. 1, include 3 server clusters, cluster topology structure and bandwidth condition in network
As shown in Figure 1.
With reference to Fig. 5, the method for the present invention includes the following steps:
Step 1), the instruction strip number for collecting multiple subtasks after the instruction strip number of scheduler task, segmentation, subtask it
Between Temporal dependency relationship, the file block size of each task；
Step 2) acquires the speed of service MIPS of each server in cluster, unit time electricity charge expense CPS, currently bears
It carries, earliest idle moment EST, the bandwidth information between calculation server；
Step 3) will carry out descending sort according to instruction strip number to scheduler task in queue, for head of the queue wait dispatch
Task successively executes step 4)-step 10), until the needed scheduler task in queue has been processed into；
Step 4) is decomposed to current to scheduler task, obtains subtask structure chart, and to be described to scheduler task
N number of allocation plan is randomly generated, forms initial population；
The subtask structure chart reflects the sequential relationship before being broken down into subtask to scheduler task, i.e. forerunner-
Subsequent relationship, subtask structure chart can guarantee that in subsequent execution scheduling process, any subtask is all in its all forerunner
Start again after the completion of task.
In initial population, N number of particle P is generated at random1,P2,...PN, each particle represents " a task-service
Device " allocation plan, if the sum of server is M, each particle Pn(1≤n≤N) is represented by a L*M matrix Sn: where
The size of N is set according to the number of servers in cluster, the subtask number to include in scheduler task, and L is indicated wait dispatch
The number of obtained subtask after Task-decomposing；
Step 5), the Makespan for calculating N number of particle, obtain the fitness of N number of particle；
In this step, time performance Makespan is calculated according to following principle for each particle (i.e. each task):
It a, must be after the finish time of its all forerunner subtask at the beginning of subtask；
B, subtask must receive the output file of its all forerunner subtask as its input file；
C, subtask must start in the case where being currently located processor idle states；
D, subtask must start as early as possible in the case where meeting above-mentioned condition a-c condition；
E, subtask TaskjIt is Task_MI the time required to executingj/MIPSk；
F, a task complete time span Makespan be its first subtask at the beginning of with the last one
Time span between the finish time of subtask.
After the Makespan for obtaining N number of particle, the fitness of n-th of particle is calculated according to following formula (2)
Fitnessn:
Fitnessn=max { Makespan1,Makespan2,...MakespanN}-MakespannFormula (2)
The respective fitness of N number of particle can be obtained according to the formula.
Step 6) calculates movement speed of the particle in new round iteration, and each particle is once evolved, is obtained down
Generation population；
It, will the wherein maximum particle of fitness value after calculating the respective fitness value of N number of particle in a previous step
Labeled as Pbest, the corresponding allocation matrix of the particle is labeled as Sbest, define particle PnTo PbestMovement speed is
Vn,best, then shown in for example following formula of its calculation formula (3):
Vn,best=(fitnessbest-fitnn)/fitnessbFormula (3)
Vn,bestIt can be regarded as matrix SnIn element variation be SbestProbability, population it is primary evolution in, each
Particle PnAll with speed Vn,bestTo PbestShifting moves a step, and the N number of new particle obtained will form next-generation population.
Whether step 7), current evolution number are less than preset value, if so, re-executeing the steps 5), otherwise, execute step
8)；The preset value determines that subtask number, number of servers are more, the preset value according to subtask number, number of servers
With regard to the bigger of setting.
Step 8) filters out the highest particle P of fitness from evolution results all beforebest；
Step 9), the corresponding weight P of setting overhead performance (empirical value obtained according to multiple test result), are based on
The task immigration of expense Cost；
The best particle P of fitness in filtering out current particle groupbestAfterwards, for current subtask, it is calculated at it
The time performance and overhead performance executed on its any server, and the performance is calculated compared to PbestGain delta C and Δ FT,
Wherein C indicates the cost (including charge on traffic and processor electricity charge) when executing on a certain server, and FT is indicated in a certain clothes
The completion moment executed on business device, CbaseAnd FTbaseCalculated C and FT value before being task immigration.
For Servers-all, finds out and calculate the obtained maximum server of value according to the following formula (4), it then will be sub
On task assignment to the server:
P*ΔC/Cbase+(1-P)*ΔFT/FTbaseFormula (4)
Step 10) exports the task immigration of step 9) as a result, carrying out subtask scheduling by this result.
It should be noted last that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting.Although ginseng
It is described the invention in detail according to embodiment, those skilled in the art should understand that, to technical side of the invention
Case is modified or replaced equivalently, and without departure from the spirit and scope of technical solution of the present invention, should all be covered in the present invention
Scope of the claims in.
Claims (5)
1. a kind of population dispatching method towards Divisible task, comprising:
After task to be scheduled is divided into subtask, using the task allocation plan that is randomly generated as a particle, with task
Fitness of the corresponding time performance of allocation plan as particle is calculated between particle mutually with the difference between particle fitness
Mobile speed does multiple evolution to population, and the best particle of fitness is selected from the result repeatedly evolved；Finally combine
Subtask scheduling is done in overhead value, each subtask in task allocation plan corresponding to the particle best to fitness；
This method specifically includes:
Step 1), the instruction strip number for collecting multiple subtasks after the instruction strip number of scheduler task, segmentation, between subtask
Temporal dependency relationship, the file block size of each task；
The speed of service MIPS of each server in step 2), acquisition cluster, unit time electricity charge expense CPS, present load, most
Early idle moment EST, the bandwidth information between calculation server；
Step 3) will carry out descending sort according to instruction strip number to scheduler task in queue, for head of the queue to scheduler task,
Step 4)-step 10) is successively executed, until the needed scheduler task in queue has been processed into；
Step 4) is decomposed to current to scheduler task, and subtask structure chart is obtained, and is described random to scheduler task
N number of allocation plan is generated, initial population is formed；
Wherein, the subtask structure chart reflects the sequential relationship before being broken down into subtask to scheduler task；
The initial population includes the N number of particle P generated at random1,P2,...PN, each particle represents " a task-service
Device " allocation plan, if the sum of server is M, each particle PnIt is expressed as a L*M matrix Sn: where 1≤n≤N, N's
Size is set according to the number of servers in cluster, the subtask number to include in scheduler task, and L is indicated to scheduler task
The number of obtained subtask after decomposition；
TaskjFor j-th of subtask, ServerkFor k-th of server；
Step 5), the time span Makespan for calculating N number of particle, obtain the fitness of N number of particle；
Step 6) calculates movement speed of the particle in new round iteration, and each particle is once evolved, the next generation is obtained
Population；
Step 7) judges whether current evolution number is less than preset value, if so, re-executeing the steps 5), otherwise, executes step
8)；
Step 8) filters out the highest particle P of fitness from evolution results all beforebest；
Step 9), the corresponding weight P of setting overhead performance, carry out the task immigration based on expense Cost；
Step 10) exports the task immigration of step 9) as a result, carrying out subtask scheduling by this result.
2. the population dispatching method according to claim 1 towards Divisible task, which is characterized in that in step 5)
In, time span Makespan is calculated according to following principle:
It a, must be after the finish time of its all forerunner subtask at the beginning of subtask；
B, subtask must receive the output file of its all forerunner subtask as its input file；
C, subtask must start in the case where being currently located processor idle states；
D, subtask must start as early as possible in the case where meeting above-mentioned condition a-c condition；
E, subtask TaskjIt is Task_MI the time required to executingj/MIPSk；Task_MIjFor j-th of subtask TaskjInstruction
Item number；MIPSkFor ServerkThe instruction strip number of operation per second；
F, appoint at the beginning of the time span Makespan that a task is completed is its first subtask with last height
Time span between the finish time of business.
3. the population dispatching method according to claim 1 towards Divisible task, which is characterized in that in step 5)
In, the fitness Fitness of n-th of particle is calculated according to following formula (2)n:
Fitnessn=max { Makespan1,Makespan2,...MakespanN}-MakespannFormula (2)
MakespannFor the time span of n-th of particle, 1≤n≤N.
4. the population dispatching method according to claim 1 towards Divisible task, which is characterized in that the step 6)
Include:
It is P by the maximum particle label of fitness valuebest, the corresponding allocation matrix of the particle is labeled as Sbest；
Define particle PnTo PbestMovement speed is Vn,best, then shown in for example following formula of its calculation formula (3):
Vn,best=(fitnessbest-fitnessn)/fitnessbest
Formula (3)；fitnessbestFor the fitness of the maximum particle of fitness value；
In the primary evolution of population, each particle PnAll with speed Vn,bestTo PbestShifting moves a step, the N number of new grain obtained
Son forms next-generation population.
5. the population dispatching method according to claim 1 towards Divisible task, which is characterized in that the step 9)
Include:
The best particle P of fitness in filtering out current particle groupbestAfterwards, for current subtask, it is calculated other any
The time performance and overhead performance executed on server, and the performance is calculated compared to PbestGain delta C and Δ FT, wherein C
Indicate the cost when executing on a certain server, FT indicates the completion moment executed on a certain server, CbaseAnd FTbase
Calculated C and FT value before being task immigration；
For Servers-all, finds out and calculate the obtained maximum server of value according to the following formula (4), then by subtask
It is assigned on the server:
P*ΔC/Cbase+(1-P)*ΔFT/FTbaseFormula (4).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410768968.9A CN105740059B (en) | 2014-12-11 | 2014-12-11 | A kind of population dispatching method towards Divisible task |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410768968.9A CN105740059B (en) | 2014-12-11 | 2014-12-11 | A kind of population dispatching method towards Divisible task |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105740059A CN105740059A (en) | 2016-07-06 |
| CN105740059B true CN105740059B (en) | 2018-12-04 |
Family
ID=56240898
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410768968.9A Expired - Fee Related CN105740059B (en) | 2014-12-11 | 2014-12-11 | A kind of population dispatching method towards Divisible task |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105740059B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106371908A (en) * | 2016-08-31 | 2017-02-01 | 武汉鸿瑞达信息技术有限公司 | Optimization method for image/video filter task distribution based on PSO (Particle Swarm Optimization) |
| CN107729130A (en) * | 2017-09-20 | 2018-02-23 | 昆明理工大学 | A kind of time point based on information physical system does not know task-dynamic dispatching method |
| CN107613025B (en) * | 2017-10-31 | 2021-01-08 | 武汉光迅科技股份有限公司 | Message queue sequence reply-based implementation method and device |
| CN110928651B (en) * | 2019-10-12 | 2022-03-01 | 杭州电子科技大学 | Service workflow fault-tolerant scheduling method under mobile edge environment |
| CN111414198B (en) * | 2020-03-18 | 2023-05-02 | 北京字节跳动网络技术有限公司 | Request processing method and device |
| TWI749992B (en) * | 2021-01-06 | 2021-12-11 | 力晶積成電子製造股份有限公司 | Wafer manufacturing management method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101604258A (en) * | 2009-07-10 | 2009-12-16 | 杭州电子科技大学 | A kind of method for scheduling task of embedded heterogeneous multiprocessor system |
| CN102724220A (en) * | 2011-03-29 | 2012-10-10 | 无锡物联网产业研究院 | Method and apparatus for task cooperation, and system for internet of things |
| CN103019822A (en) * | 2012-12-07 | 2013-04-03 | 北京邮电大学 | Large-scale processing task scheduling method for income driving under cloud environment |
-
2014
- 2014-12-11 CN CN201410768968.9A patent/CN105740059B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101604258A (en) * | 2009-07-10 | 2009-12-16 | 杭州电子科技大学 | A kind of method for scheduling task of embedded heterogeneous multiprocessor system |
| CN102724220A (en) * | 2011-03-29 | 2012-10-10 | 无锡物联网产业研究院 | Method and apparatus for task cooperation, and system for internet of things |
| CN103019822A (en) * | 2012-12-07 | 2013-04-03 | 北京邮电大学 | Large-scale processing task scheduling method for income driving under cloud environment |
Non-Patent Citations (3)
| Title |
|---|
| Self-Adaptive Learning PSO-Based Deadline Constrained Task Scheduling for Hybrid IaaS Cloud;Xingquan Zuo等;《IEEE Transactions on Automation Science and Engineering》;20140430;第11卷(第2期);564-573 * |
| 基于改进粒子群算法的云计算任务调度算法;张陶等;《计算机工程与应用》;20131001;第49卷(第19期);68-72 * |
| 粒子群算法求解任务可拆分项目调度问题;邓林义等;《控制与决策》;20080615;第23卷(第6期);681-684 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105740059A (en) | 2016-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105740059B (en) | A kind of population dispatching method towards Divisible task | |
| CN108829494B (en) | Container cloud platform intelligent resource optimization method based on load prediction | |
| CN103605567B (en) | Cloud computing task scheduling method facing real-time demand change | |
| US10031774B2 (en) | Scheduling multi-phase computing jobs | |
| CN105718479B (en) | Execution strategy generation method and device under cross-IDC big data processing architecture | |
| CN103729248B (en) | A kind of method and apparatus of determination based on cache perception task to be migrated | |
| CN104765640B (en) | A kind of intelligent Service dispatching method | |
| CN103593323A (en) | Machine learning method for Map Reduce task resource allocation parameters | |
| CN111274036B (en) | Scheduling method of deep learning task based on speed prediction | |
| CN103699433B (en) | One kind dynamically adjusts number of tasks purpose method and system in Hadoop platform | |
| CN112685153A (en) | Micro-service scheduling method and device and electronic equipment | |
| CN110347602A (en) | Multitask script execution and device, electronic equipment and readable storage medium storing program for executing | |
| Quang-Hung et al. | Epobf: energy efficient allocation of virtual machines in high performance computing cloud | |
| CN115220898A (en) | Task scheduling method, system, device and medium based on deep reinforcement learning | |
| Fathalla et al. | Best-KFF: a multi-objective preemptive resource allocation policy for cloud computing systems | |
| CN102184124B (en) | Task scheduling method and system | |
| Iglesias et al. | A methodology for online consolidation of tasks through more accurate resource estimations | |
| CN111367632A (en) | Container cloud scheduling method based on periodic characteristics | |
| Yakubu et al. | Priority based delay time scheduling for quality of service in cloud computing networks | |
| CN111506407B (en) | Resource management and job scheduling method and system combining Pull mode and Push mode | |
| Haque et al. | A priority-based process scheduling algorithm in cloud computing | |
| CN106648866A (en) | KVM platform meeting task time limit requirements-based resource scheduling method | |
| Chen et al. | Pickyman: A preemptive scheduler for deep learning jobs on gpu clusters | |
| Bhutto et al. | Analysis of Energy and Network Cost Effectiveness of Scheduling Strategies in Datacentre | |
| Pan | Heterogeneous High-Performance System Algorithm Based on Computer Big Data Technology |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181204 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |