CN108958919A - More DAG task schedule expense fairness assessment models of limited constraint in a kind of cloud computing - Google Patents

More DAG task schedule expense fairness assessment models of limited constraint in a kind of cloud computing Download PDF

Info

Publication number
CN108958919A
CN108958919A CN201810766931.0A CN201810766931A CN108958919A CN 108958919 A CN108958919 A CN 108958919A CN 201810766931 A CN201810766931 A CN 201810766931A CN 108958919 A CN108958919 A CN 108958919A
Authority
CN
China
Prior art keywords
dag
cost
task
virtual machine
fairness
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
Application number
CN201810766931.0A
Other languages
Chinese (zh)
Other versions
CN108958919B (en
Inventor
龙赛琴
周思恒
刘子浩
曾令斌
幸运
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xiangtan University
Original Assignee
Xiangtan University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xiangtan University filed Critical Xiangtan University
Priority to CN201810766931.0A priority Critical patent/CN108958919B/en
Publication of CN108958919A publication Critical patent/CN108958919A/en
Application granted granted Critical
Publication of CN108958919B publication Critical patent/CN108958919B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06F9/4806—Task transfer initiation or dispatching
    • 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
    • G06F9/4806—Task transfer initiation or dispatching
    • G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The present invention devises a kind of more DAG task schedule expense fairness assessment models of limited constraint in cloud computing, in more DAG scheduling schemes of limited constraint, the DAG that different scheduling schemes wants the loose degree of seeking time different the time limit will appear some fairness problems when being scheduled.The model requires stringent task to require loose task using different calculation formula from the time limit time limit, and the task urgent for time requirement suitably increases " punishment ", and the task loose to time requirement also sets minimum execution cost.User is submitted to the actual motion expense of the task computation of system its estimated execution cost, user's predictable cost and system, the scheduling scheme is judged whether to user friendly, fair, thus the satisfaction of the better scheduling scheme raising user of selection one by the relationship of this three.

Description

A kind of more DAG task schedule expense fairness assessment of limited constraint in cloud computing Model
Technical field
The present invention relates to DAG task schedule fields more in cloud computing, and more DAG task schedules are taken under predominantly limited constraint With the foundation of fairness assessment models.
Background technique
In recent years, due to the rapidly growth of global metadata total amount, the efficiency of data retrieval also declines therewith, physics, day The scientific algorithm amount of literature, the fields such as biology is also increasingly huge.It is high-performance, cheap when carrying out these data and calculating Distributed computer cluster is instead of single computer once.Cloud computing receives the concern of numerous researchers, Cloud computing technology also enters Rapid development stage.Nowadays, it there is also some problems in field of cloud calculation and need to be solved.Its Middle Mission Scheduling is just always to be difficult to the problem of balancing in distributed computing field.In Mission Scheduling, according to appoint It whether there is dependence between business and task, task schedule is divided into Independent Task Scheduling and dependence task schedule.It is this Dependence task schedule is usually indicated with directed acyclic graph DAG.In the huge data of calculating complexity, the work of DAG task If can optimize on software view in stream task distribution, allow task processing at a reasonable time, reasonable expense Interior completion will bring great convenience to user and more preferably experience.
At present about in the scheduling strategy of DAG, the technical solution of the target and use that need to solve is also different.By This several years development, present list DAG scheduling strategy has been achieved for numerous achievements, but the scheduling problem about multiple DAG is also It is to have certain defect, most of more DAG dispatching algorithms are optimized on the basis of single DAG dispatching algorithm.More Time limit constraint is added in DAG scheduling problem, can effectively increase user quality, promote the Experience Degree of user.It solves limited More DAG scheduled throughput maximization problems of constraint, for the handling capacity, the execution cost of task, the utilization rate of resource of system The problems such as optimization can play the role of it is good.
Summary of the invention
In more DAG scheduling schemes of limited constraint, the deadline pressing degree of DAG task determines this DAG DAG time limit task time that the dispatching priority of subtask, i.e. user are submitted is shorter, then illustrates that the DAG is just urgent, need By priority scheduling, (when the task that multiple users submit is all very urgent, in order to improve the handling capacity of system, need an extension part DAG).In this scheduling scheme, only consider to complete within the time limit, without fairness constraint, will lead to long times of deadline time Priority of being engaged in is lower, may be deferred in scheduling to improve handling capacity.It may also arrive deadline at scheduling and close on Shi Caineng is assigned to resource.This is run since one arithmetic speed faster processor of Systematic selection it is pressed for time, can be made Business.In cloud cluster, the better processor of performance, processing cost is higher.In this case, a time requirement is loose Task not only to handle the time long, and be also possible to along with higher expense, this is clearly inequitable for a user. The fairness of entire scheduling strategy is promoted, then first the fairness of scheduling scheme must be assessed, could preferably be improved Scheduling scheme.
In order to assess the fairness in scheduling scheme, the present invention devises a kind of more DAG of limited constraint in cloud computing The constraint in time limit is added in task schedule expense fairness assessment models, suitably requires user it is pressed for time to increase the time limit and " punishes Penalize ", to one minimum execution cost of user setting that the time limit wants seeking time loose, guarantee the balance and applicability of model.
Specific embodiment
Step 1 sets the virtual machine for having n free time in cloud service cluster, and the processing speed of each virtual machine is expressed as SVM= {svm1,svm2,...,svmi,...,svmn, unit is million instructions per second；The processing cost of each virtual machine is expressed as COST={ cost1,cost2,...,costi,...,costn, unit is the expense of each second when virtual machine is run.
Step 2 sets and submits the m DAG workflow tasks with the different completion time limits, each DAG table with user orientation server It is shown as DAG={ G1,G2,...,Gx,...,Gm, GxCut-off complete the time limit be deadlinex, subtask quantity is k, the DAG The task amount of subtask is expressed as W={ wx1,wx2,...,wxj,...,wxk, unit is million instructions.
Step 3 calculates GxThe estimated charges costEST of completion is executed in cloud service clusterx.After user submits DAG, meter Calculate the estimated charges of this time scheduling, it is contemplated that expense is unrelated with scheduling scheme, deadline, the execution with the task amount, system of DAG Expense is related.
GxExecuting the estimated charges completed is costESTx:
Step 4 calculates GxThe desired execution cost of user under time limit constraint.Requirement according to user to deadline, Time limit it is long task flow expense it is lower, the time limit it is short task flow it is costly.
1) the average treatment speed of each virtual machine is calculated
2) the average execution cost per second of virtual machine is calculated
3) G is calculatedxExpectation execution cost cost per secondx:
4) by costxIt is compared with virtual machine list of charges, an expense is selected to be greater than or equal to costxVirtual machine Execute the DAG.For example, if costxSize be 2.4, virtual machine list of charges are as follows: 1,1.5,2,3,3.8,5.Then select expense The task is executed for 3 virtual machine.
5) according to step 4), if the virtual machine execution cost cost of serial number ccWith costxIt is corresponding, virtual machine processing Speed is svmc, GxThe desired execution cost costEXP of user under time limit constraintxAre as follows:
Step 5 calculates GxActual schedule scheme execute spend costACTx.Different scheduling schemes will lead to different Execution cost will evaluate and test the fairness of scheduling scheme, it is necessary to learn the actual result of dispatching algorithm.If alpha-beta has mapping relations, Task α is dispatched in virtual machine β by expression to be run, then the DAG of actual schedule scheme executes cost are as follows:
Step 6 calculates GxFair function Fair (Gx).Since it is expected that expense costESTxWith scheduling scheme, deadline It is unrelated, to a certain extent, it is contemplated that expense costESTxIt can be used as the fair expense of user.As estimated charges costESTxIt is small In desired expense costEXPxWhen, illustrate that user is more demanding to deadline, formula model is related to desired expense, when estimated Expense costESTxGreater than desired expense costEXPxWhen, it is lower to illustrate that user requires deadline, formula model and pre- charging With correlation.Formula is as follows:
Fair function Fair (Gx) value it is higher, show that the cost of user does not meet expection more, for a user more It is unfair.Functional value is lower, then illustrates that user effort meets expection, just seem fair relatively.
Step 7, the population mean justice value averFair for calculating the scheduling scheme.In step 3-6, calculating is single The reasonability of the parameters of DAG, scheduling scheme should be evaluated from all user perspectives, and population mean justice value model is such as Under:
The lower expection for illustrating scheduling scheme and meeting most users of the value is added standard deviation STDEVFair evaluation and uses Expected discrepancy between family:
As can be seen that standard deviation is smaller from formula, the expected discrepancy between user and user is with regard to smaller.
Different scheduling schemes can be compared by the overall fair value of scheduling scheme from standard deviation, so as to A more reasonable scheme is selected, the satisfaction of user is improved.

Claims (6)

1. more DAG task schedule expense fairness assessment models of limited constraint in a kind of cloud computing, it is characterised in that described Method at least includes the following steps:
The processing speed and their processing cost of all virtual machines in step 1, setting cluster；
Step 2, each user of setting are submitted to the attribute of DAG task in cluster；
Step 3 calculates the estimated execution cost of DAG in the cluster；
Step 4 calculates the desired execution cost of DAG user；
Step 5 calculates execution cost of the DAG in actual schedule；
Step 6, the fairness for calculating DAG scheduling；
Step 7, the population mean justice value and standard deviation for calculating scheduling scheme.
2. more DAG task schedule expense fairness assessment models of limited constraint in cloud computing according to claim 1, It is characterized in that the DAG expects that the calculating process of execution cost is as follows in the cluster:
If the DAG of serial number x is expressed as Gx, wherein costESTxIt is expressed as GxExecute the estimated charges completed, wxjIndicate GxMiddle sequence Number for j subtask task amount, svmiIndicate the execution speed of virtual machine i, costiThen indicate the execution cost of virtual machine i, n Indicate the number of virtual machine in cloud system cluster.
3. more DAG task schedule expense fairness assessment models of limited constraint in cloud computing according to claim 1, It is characterized in that the DAG user it is expected the calculating process of execution cost, it is at least further comprising the steps of:
1) the average treatment speed of each virtual machine is calculated
2) the average execution cost per second of virtual machine is calculated
3) G is calculatedxExpectation execution cost cost per secondx:
Wherein deadlinexIndicate time limit deadline of the DAG of serial number x；
4) by costxIt is compared with virtual machine list of charges, an expense is selected to be greater than or equal to costxVirtual machine execute Gx；
5) G is calculatedxThe desired execution cost costEXP of user under time limit constraintx:
Wherein costcExpression and GxThe virtual machine execution cost of corresponding serial number c, svmcFor the processing speed of the virtual machine.
4. more DAG task schedule expense fairness assessment models of limited constraint in cloud computing according to claim 1, It is characterized in that DAG calculating process of execution cost in actual schedule is as follows:
Calculate execution cost costACT of the DAG in actual schedulex；
Alpha-beta has mapping relations, and task α is dispatched in virtual machine β by expression to be run.
5. more DAG task schedule expense fairness assessment models of limited constraint in cloud computing according to claim 1, It is characterized in that the calculating process of the DAG scheduling fairness is as follows:
Calculate fairness function Fair (Gx):
6. more DAG task schedule expense fairness assessment models of limited constraint in cloud computing according to claim 1, It is characterized in that the calculating process of the scheduling scheme population mean justice value and standard deviation, at least further includes following step It is rapid:
1) scheduling scheme population mean justice value averFair is calculated:
2) standard deviation STDEVFair is calculated:
Wherein m is the quantity of virtual machine in cluster.
CN201810766931.0A 2018-07-13 2018-07-13 Multi-DAG task scheduling cost fairness evaluation method with deadline constraint in cloud computing Active CN108958919B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810766931.0A CN108958919B (en) 2018-07-13 2018-07-13 Multi-DAG task scheduling cost fairness evaluation method with deadline constraint in cloud computing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810766931.0A CN108958919B (en) 2018-07-13 2018-07-13 Multi-DAG task scheduling cost fairness evaluation method with deadline constraint in cloud computing

Publications (2)

Publication Number Publication Date
CN108958919A true CN108958919A (en) 2018-12-07
CN108958919B CN108958919B (en) 2023-07-25

Family

ID=64483121

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810766931.0A Active CN108958919B (en) 2018-07-13 2018-07-13 Multi-DAG task scheduling cost fairness evaluation method with deadline constraint in cloud computing

Country Status (1)

Country Link
CN (1) CN108958919B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110209467A (en) * 2019-05-23 2019-09-06 华中科技大学 A kind of flexible resource extended method and system based on machine learning
CN113283692A (en) * 2021-03-19 2021-08-20 东南大学 Intelligent man-machine cooperation scheduling method and system for monitoring resource allocation of bulk commodity trading market
CN114860397A (en) * 2022-04-14 2022-08-05 深圳清华大学研究院 Task scheduling method, device and equipment

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104021040A (en) * 2014-06-04 2014-09-03 河海大学 Cloud computing associated task scheduling method and device based on time constraint
WO2016165392A1 (en) * 2015-04-17 2016-10-20 华南理工大学 Genetic algorithm-based cloud computing resource scheduling method
WO2017045211A1 (en) * 2015-09-16 2017-03-23 国云科技股份有限公司 Cloud computing task scheduling method constrained by multiple quality-of-service requirements
CN106934539A (en) * 2017-03-05 2017-07-07 北京工业大学 It is a kind of with limited and expense restriction workflow schedule method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104021040A (en) * 2014-06-04 2014-09-03 河海大学 Cloud computing associated task scheduling method and device based on time constraint
WO2016165392A1 (en) * 2015-04-17 2016-10-20 华南理工大学 Genetic algorithm-based cloud computing resource scheduling method
WO2017045211A1 (en) * 2015-09-16 2017-03-23 国云科技股份有限公司 Cloud computing task scheduling method constrained by multiple quality-of-service requirements
CN106934539A (en) * 2017-03-05 2017-07-07 北京工业大学 It is a kind of with limited and expense restriction workflow schedule method

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110209467A (en) * 2019-05-23 2019-09-06 华中科技大学 A kind of flexible resource extended method and system based on machine learning
CN110209467B (en) * 2019-05-23 2021-02-05 华中科技大学 Elastic resource expansion method and system based on machine learning
CN113283692A (en) * 2021-03-19 2021-08-20 东南大学 Intelligent man-machine cooperation scheduling method and system for monitoring resource allocation of bulk commodity trading market
CN113283692B (en) * 2021-03-19 2024-04-16 东南大学 Intelligent man-machine cooperation scheduling method and system for supervising resource allocation in bulk commodity trade market
CN114860397A (en) * 2022-04-14 2022-08-05 深圳清华大学研究院 Task scheduling method, device and equipment

Also Published As

Publication number Publication date
CN108958919B (en) 2023-07-25

Similar Documents

Publication Publication Date Title
CN110737529B (en) Short-time multi-variable-size data job cluster scheduling adaptive configuration method
CN107659433B (en) Cloud resource scheduling method and equipment
CN107273209B (en) Hadoop task scheduling method based on minimum spanning tree clustering improved genetic algorithm
Ben-Yehuda et al. Expert: Pareto-efficient task replication on grids and a cloud
CN108182109B (en) Workflow scheduling and data distribution method in cloud environment
Patel et al. Priority based job scheduling techniques in cloud computing: a systematic review
CN111381950A (en) Task scheduling method and system based on multiple copies for edge computing environment
CN103729246B (en) Method and device for dispatching tasks
CN107341041B (en) Cloud task multidimensional constraint backfill scheduling method based on priority queue
CN103942102A (en) Real-time task scheduling method based on dual priority
CN109799956B (en) Memory controller and IO request processing method
CN108737462A (en) A kind of cloud computation data center method for scheduling task based on graph theory
CN108958919A (en) More DAG task schedule expense fairness assessment models of limited constraint in a kind of cloud computing
Thaman et al. Green cloud environment by using robust planning algorithm
US8281313B1 (en) Scheduling computer processing jobs that have stages and precedence constraints among the stages
CN109710372A (en) A kind of computation-intensive cloud workflow schedule method based on cat owl searching algorithm
CN106991006A (en) Support the cloud workflow task clustering method relied on and the time balances
Ding et al. QoS-constrained service selection for networked microservices
CN114610474A (en) Multi-strategy job scheduling method and system in heterogeneous supercomputing environment
Singh et al. A comparative study of various scheduling algorithms in cloud computing
CN110262896A (en) A kind of data processing accelerated method towards Spark system
CN111736959B (en) Spark task scheduling method considering data affinity under heterogeneous cluster
Yang et al. Budget and SLA aware dynamic workflow scheduling in cloud computing with heterogeneous resources
CN112099932A (en) Optimal pricing method and system for soft-hard deadline task offloading in edge computing
CN115098240B (en) Multiprocessor application scheduling method and system and storage 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
GR01 Patent grant
GR01 Patent grant