CN102360246B - Self-adaptive threshold-based energy-saving scheduling method in heterogeneous distributed system - Google Patents

Self-adaptive threshold-based energy-saving scheduling method in heterogeneous distributed system Download PDF

Info

Publication number
CN102360246B
CN102360246B CN201110312108.0A CN201110312108A CN102360246B CN 102360246 B CN102360246 B CN 102360246B CN 201110312108 A CN201110312108 A CN 201110312108A CN 102360246 B CN102360246 B CN 102360246B
Authority
CN
China
Prior art keywords
task
energy consumption
scheduling
processor
threshold
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
Application number
CN201110312108.0A
Other languages
Chinese (zh)
Other versions
CN102360246A (en
Inventor
刘伟
杜薇
尹行
段玉光
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.)
Wuhan University of Technology WUT
Original Assignee
Wuhan University of Technology WUT
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 Wuhan University of Technology WUT filed Critical Wuhan University of Technology WUT
Priority to CN201110312108.0A priority Critical patent/CN102360246B/en
Publication of CN102360246A publication Critical patent/CN102360246A/en
Application granted granted Critical
Publication of CN102360246B publication Critical patent/CN102360246B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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

Abstract

The invention relates to a self-adaptive threshold-based energy-saving scheduling method in an environment of supporting a heterogeneous distributed system, belonging to the technical field of parallel task scheduling of the heterogeneous distributed system. The method provided by the invention specifically comprises the steps of: reading a parallel task directed acyclic graph (DAG) file; obtaining an initial task scheduling sequence; obtaining an optimal threshold: dynamically obtaining an optimal threshold according to parallel task and system environment; grouping the tasks: controlling task duplication by using the optimal threshold, namely, selectively duplicating optimal precursors of the tasks to balance system performance and energy consumption to obtain approximate optimal grouping; mapping the tasks: scheduling all groups into a processor, wherein the processor is not distributed and has minimum energy consumption; and regulating the voltage of the processor: dynamically regulating the voltage of the processor by using a task idle time to reduce the energy consumption of the processor. According to the invention, requirements on system performance and energy consumption are comprehensively considered. The optical threshold in the method can self-adapt the parallel tasks and the system environment, and is used for controlling the task duplication for balancing the system performance and the energy consumption, so that the energy consumption is reduced as far as possible under the premise that the final scheduling result meets the requirement of the system performance.

Description

Energy-saving scheduling method based on adaptive threshold in a kind of heterogeneous distributed system
Technical field
The present invention relates to a kind of computer system energy-saving scheduling method, particularly relate to a kind of energy-saving scheduling method based on adaptive threshold in heterogeneous distributed system of supporting.
Background technology
Heterogeneous distributed system (HDS) is comprised of many computational resources with different disposal ability conventionally, and these computational resources are connected with each other by meeting the express network of various application requirements.In the past few years, heterogeneous distributed system has become popular computing platform, for computation-intensive and communications-intensive tasks provide various computation requirements.
Heterogeneous distributed system is used widely at numerous areas such as industry and commerce.Yet heterogeneous distributed system has also consumed huge energy when powerful calculating ability is provided.According to UN (Energy User News) report, now, cluster computing center energy requirement, from 150W/ft2 to 200W/ft2, will be increased to 200-300W/ft2 in the coming years.Environmental Protection Department of the United States Federal (EPA) has submitted a report about data center in August, 2009.This part of report title, the energy resource consumption of data center increased by one times from 2000 to 2006, expected 2011 and doubled.Huge energy consumption cost becomes the important bottleneck of the heterogeneous distributed System Development of restriction, is badly in need of effectively being solved.
With directed acyclic graph DAG (Directed Acyclic Graph), represent that Parallel Task Scheduling problem is the study hotspot of academia always, Parallel Task Scheduling problem has been proved to be np complete problem, has caused research widely at home and abroad.The target of scheduling is to meet under the prerequisite of priority of task restriction relation, according to suitable allocation strategy, determine a kind of assignment and execution sequence, task that can executed in parallel is reasonably allocated on processor and carries out in an orderly manner, to reach the object that reduces total execution time.
Scheduling strategy based on different, existing dispatching method mainly can be divided three classes: list scheduling method, the dispatching method that clusters, the dispatching method based on Task Duplication.Energy-conservation in heterogeneous distributed system can be taked traditional dispatching method, and operational scheme is also closely similar, just when scheduling, considered energy-conservation this target.Wherein, the selection of dispatching method has larger impact for scheduling result, because inappropriate dispatching method may make real powerful computing power can not get embodying, even can reduce system performance, increases system energy consumption.At present, existing energy-saving scheduling method great majority are based on above three kinds of dispatching methods, then in conjunction with dynamic electric voltage, regulate the power-saving technologies such as (DVS), dynamic power management (DPM) to carry out optimization system energy consumption.
It is a breakthrough of microprocessor low power dissipation design in recent years that dynamic electric voltage regulates (DVS) technology, and it allows dynamically to regulate microprocessor voltage and frequency, and the system that simultaneously guarantees is normally worked.At present, in embedded hardware circuit design, no matter be CPU, or peripheral I/O interface, all generally use CMOS logical circuit.In CMOS logical circuit, power consumption is directly proportional to the quadratic power of frequency and voltage.Although reduce the increase that frequency/voltage can cause task execution time, but according to the dependence exploration task free time between parallel task, utilize dynamic electric voltage to regulate (DVS) technology reasonably to regulate processor voltage, do not affecting under the prerequisite of other tasks carryings, can effectively reduce processor energy consumption.In voltage-regulation process, voltage management chip can be carried out small voltage adjustment, and can be within the extremely short time (in tens microseconds) complete the adjustment of voltage.
At present, both at home and abroad about the energy-saving scheduling method under heterogeneous distributed system environments mainly contains:
The people such as D.P.Agrawal of U.S. University of Cincinnati have proposed a kind of dispatching algorithm (TDS) based on Task Duplication in heterogeneous system, this algorithm as much as possible replication task forerunner with improved system performance.Then, proposed again a kind of dispatching algorithm (TANH) based on heterogeneous network, this algorithm has carried out improving with further sophisticated systems performance on the basis of TDS algorithm.
Dispatching method based on Task Duplication by copying certain or some tasks to nonidentical processor, is eliminated the communication overhead between the task with restriction relation, thereby is reduced whole task scheduling time span exactly.The method can reduce task scheduling length effectively, saves communication energy consumption, but also can increase the calculating energy consumption of replication task simultaneously, therefore in energy-saving distribution, needs careful consideration.
The people such as Zong Ziliang of U.S. Auburn University have proposed a kind of energy-saving distribution algorithm based on Task Duplication, and this algorithm has proposed a threshold value and come control task to copy, and optionally the best forerunner of replication task comes balance system performance and energy consumption two aspects.The people such as the Gong Bin of Shandong University have proposed a kind of power-economizing method that cluster dispatching method and Task Duplication dispatching method are effectively combined, and utilize dynamic power management (DPM) technology to save the idle energy consumption of processor.
The Lee of Sydney University has proposed a kind of heuristic dispatching method that regulates (DVS) technology based on dynamic electric voltage, the method considers aspect performance and processor energy consumption two in task scheduling process simultaneously, proposed a relative priority (RS) as evaluation index set the tasks dispatch processor and the execution voltage on this processor.The Kang Yan of Yunnan University has proposed a kind of power-economizing method in heterogeneous distributed system, first the method utilizes taboo (Tabu) strategy to obtain free time, then utilizes dynamic electric voltage to regulate the voltage of (DVS) Techniques For Reducing current task on processor to save CPU energy consumption.The people such as the Kang of Univ Florida USA have proposed a kind of scheduling strategy that regulates (DVS) technology based on dynamic electric voltage, rationally adjust processor voltage by effective allocating task free time, thus minimizing processor energy consumption.
In sum, dispatching method major part has in the past carried out improving or optimizing for certain part (system energy consumption comprises processor energy consumption and network energy consumption two parts) of performance or energy consumption.The weak point existing has: the method 1. having is only considered performance and ignored energy consumption completely; 2. some energy-saving scheduling method based on Task Duplication utilizes threshold value to control Task Duplication, but given threshold value is to arrange arbitrarily, can not regulate according to parallel task and system environments self-adaptation, causes scheduling result unstable; Although the method 3. having has not only been considered performance but also considered processor energy consumption, has ignored network service energy consumption.
Summary of the invention
The object of this invention is to provide the energy-saving scheduling method based on adaptive threshold in a kind of heterogeneous distributed system, the method regulates (DVS) technology to combine the Task Duplication dispatching method based on adaptive threshold and dynamic electric voltage, meeting under the prerequisite of system performance requirement, can effectively save energy consumption.
Technical scheme of the present invention comprises the following steps:
One, first read parallel task directed acyclic graph DAG file.
Two, obtain initiating task scheduling sequence.For parallel task collection V, total n=|V| task, from export task
Figure 684005DEST_PATH_IMAGE001
start, calculate the priority of each task until beginning task
Figure 900222DEST_PATH_IMAGE002
finish.Then according to task priority size ascending order, arrange, obtain initiating task scheduling sequence.
Three, obtain optimal threshold; First, from first task of initiating task scheduling sequence, start to travel through all tasks, calculate and copy current task
Figure 720411DEST_PATH_IMAGE003
best forerunner (
Figure 259845DEST_PATH_IMAGE004
) the energy consumption moreenergy that increases, the energy consumption moreenergy that wherein increased equals
Figure 936814DEST_PATH_IMAGE004
calculating energy consumption deduct it with
Figure 374749DEST_PATH_IMAGE003
between communication energy consumption, get minimum value wherein and maximal value respectively as minimum threshold min_threshold and max-thresholds max_threshold; Then, task scheduling length while utilizing the computing method of scheduling length to ask threshold value to be respectively max_threshold and min_threshold – 1 is minimum scheduling length and maximum scheduling length, the computing method of described scheduling length comprise following three steps: the first step, and under current threshold value, utilize the grouping strategy in step 4 to divide into groups to task; Second step, utilize scheduling strategy in step 5 by duty mapping to corresponding processor; The 3rd step, obtains the task scheduling length under current threshold value according to task scheduling result; Finally, specify one can meet the scheduling length that user performance requires, from min_threshold – 1, start to travel through all threshold values, and utilize dispatching method to ask task scheduling length under current threshold value, until the scheduling length of trying to achieve is less than, specify scheduling length just to finish traversal, current threshold value is optimal threshold;
Four, task grouping; From first task of initiating task scheduling sequence, carry out depth-first search until beginning task finishes; In task search procedure, if current task
Figure 998628DEST_PATH_IMAGE003
best forerunner (
Figure 205619DEST_PATH_IMAGE004
) grouping, distributed to the grouping at current task place, and be labeled as and distribute, otherwise to whether copying
Figure 7484DEST_PATH_IMAGE004
carry out; If copied
Figure 932715DEST_PATH_IMAGE004
, can increase replication task
Figure 157022DEST_PATH_IMAGE004
calculating energy consumption, reduce task scheduling length and
Figure 421782DEST_PATH_IMAGE004
with
Figure 706133DEST_PATH_IMAGE003
between communication energy consumption; Suppose that the time reducing is lesstime, the time wherein reducing equals the scheduling length reducing, and the final energy consumption increasing is that moreenergy(equals
Figure 40031DEST_PATH_IMAGE004
calculating energy consumption deduct it with
Figure 68030DEST_PATH_IMAGE003
between communication energy consumption), if lesstime>0, and moreenergy is less than or equal to optimal threshold, by replication task
Figure 249612DEST_PATH_IMAGE004
add current task to
Figure 642547DEST_PATH_IMAGE003
the grouping at place, otherwise exit current group and choose first in initiating task scheduling sequence and do not divide group task to carry out next one to divide into groups.
Five, duty mapping.From first grouping, calculate and be respectively grouped in energy consumption on all unappropriated processors, current group is mapped on the processor of energy consumption minimum, and this processor of mark is occupied, so circulation is gone down until all groupings are all assigned with away.
Six, processor voltage regulates.The free time of utilizing the dependence between task to produce, sets the tasks the execution time under each voltage on dispatch processor, makes the calculating energy consumption of generation minimum, then in task implementation, dynamically adjusts processor voltage.
Feature of the present invention is for the energy-optimised problem in heterogeneous distributed system, regulates (DVS) technology to combine the Task Duplication strategy based on adaptive threshold and dynamic electric voltage, optimization system energy consumption when meeting system performance requirement.First, the Task Duplication strategy based on adaptive threshold in the present invention can be according to parallel task and optimal threshold of system environments (mainly comprising processor and network parameter that system provides) Dynamic Acquisition, the dirigibility that has improved system; Then utilize this threshold value to copy control task, optionally the best forerunner of replication task comes balance system performance and energy consumption to obtain near-optimization grouping, has improved the stability of system; Finally, by each packet scheduling to processor unappropriated and that energy consumption is minimum, the free time that this processor produces according to the dependence between task, utilize dynamic electric voltage to regulate (DVS) technology dynamically to adjust processor voltage further to save processor energy consumption.
accompanying drawing explanation
Accompanying drawing is FB(flow block) of the present invention.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described in further detail.
Parallel task: the task that user submits to represents with directed acyclic graph DAG, is defined as
Figure 542370DEST_PATH_IMAGE005
.Wherein
Figure 799826DEST_PATH_IMAGE006
represent to forgive the task-set of n task.For each task in V, task
Figure 835915DEST_PATH_IMAGE007
at processor
Figure 399752DEST_PATH_IMAGE008
on execution time be
Figure 521292DEST_PATH_IMAGE009
, wherein maximum execution time is designated as
Figure 78044DEST_PATH_IMAGE010
,
Figure 234219DEST_PATH_IMAGE011
expression task
Figure 31273DEST_PATH_IMAGE007
execution time on the processor that is scheduled.For convenient, set forth, use respectively here
Figure 577792DEST_PATH_IMAGE012
with
Figure 485705DEST_PATH_IMAGE013
the voltage and the frequency set that represent each task place processor.Especially, work as task
Figure 496387DEST_PATH_IMAGE007
while there is free time, can utilize dynamic electric voltage to regulate (DVS) technology to regulate the execution voltage on the processor of task place, task is at voltage under execution time be designated as
Figure 248890DEST_PATH_IMAGE015
.In addition, for task
Figure 960494DEST_PATH_IMAGE007
, the task computation cycle for certain value, irrelevant with processor voltage.E is message set,
Figure 167802DEST_PATH_IMAGE017
expression task
Figure 485651DEST_PATH_IMAGE007
to task the message of transmitting,
Figure 173170DEST_PATH_IMAGE019
represent pass-along message
Figure 482928DEST_PATH_IMAGE020
the call duration time of the expense of sending out.In addition, we use respectively
Figure 911242DEST_PATH_IMAGE021
with
Figure 964649DEST_PATH_IMAGE022
expression task
Figure 804429DEST_PATH_IMAGE007
subsequent tasks set and predecessor task set.
System resource: a heterogeneous distributed system is comprised of many processor nodes can with different disposal ability, is designated as
Figure 222772DEST_PATH_IMAGE023
, these nodes pass through express network
Figure 780792DEST_PATH_IMAGE024
be connected with each other.All processors in system all support dynamic electric voltage to regulate (DVS) technology, and each processor all has h the magnitude of voltage of arranging by descending order, are designated as
Figure 559261DEST_PATH_IMAGE025
, corresponding frequency values is designated as
Figure 519127DEST_PATH_IMAGE026
, wherein
Figure 170688DEST_PATH_IMAGE027
.We can represent that with the two-dimensional matrix X of a n * m n duty mapping is to the scheduling process of m heterogeneous processor.If task
Figure 888109DEST_PATH_IMAGE007
be dispatched to processor
Figure 283318DEST_PATH_IMAGE008
,
Figure 786106DEST_PATH_IMAGE028
=1, otherwise =0, wherein
Figure 813284DEST_PATH_IMAGE028
Figure DEST_PATH_IMAGE029
x.
Embodiment is carried out according to following steps.
One, read parallel task directed acyclic graph DAG file.
Two, obtain initiating task scheduling sequence.For parallel task collection V, total n=|V| task, from export task
Figure 12185DEST_PATH_IMAGE001
start, calculate the priority of each task until beginning task
Figure 805697DEST_PATH_IMAGE002
finish.Then according to task priority size ascending order, arrange, obtain initiating task scheduling sequence.Wherein, task
Figure 799061DEST_PATH_IMAGE007
priority
Figure 233017DEST_PATH_IMAGE030
definition as shown in Equation (1).
Figure 235608DEST_PATH_IMAGE031
（1）
Three, obtain optimal threshold.First, calculated threshold span.From first task of initiating task scheduling sequence, start to travel through all tasks, calculate and copy current task
Figure 696676DEST_PATH_IMAGE003
best forerunner (
Figure 860941DEST_PATH_IMAGE032
) the energy consumption moreenergy(that increases equals
Figure 102567DEST_PATH_IMAGE032
calculating energy consumption deduct it with
Figure 830220DEST_PATH_IMAGE007
between communication energy consumption),
Figure 208112DEST_PATH_IMAGE033
(use here krepresent
Figure 746541DEST_PATH_IMAGE032
), get the minimum value of its moreenergy and maximal value respectively as minimum threshold min_threshold and max-thresholds max_threshold.Best forerunner wherein
Figure 209883DEST_PATH_IMAGE032
definition is as shown in formula (2):
Figure 242692DEST_PATH_IMAGE034
（2）
Task
Figure 475090DEST_PATH_IMAGE007
earliest start time on each processor (EST) is recursive calculation from top to bottom, and wherein the earliest start time of entrance task is 0.Computing formula is:
（3）
Here, task
Figure 135059DEST_PATH_IMAGE007
actual earliest finish time can use task
Figure 204515DEST_PATH_IMAGE007
its optimum processor (
Figure 556999DEST_PATH_IMAGE036
) on earliest finish time represent:
Figure 437230DEST_PATH_IMAGE037
（4）
Task
Figure 875165DEST_PATH_IMAGE007
earliest finish time (ECT) equal task
Figure 184530DEST_PATH_IMAGE007
actual earliest start time add task
Figure 391520DEST_PATH_IMAGE007
execution time in its optimum processor, as shown in formula (5):
Figure 691921DEST_PATH_IMAGE038
（5）
Task
Figure 617151DEST_PATH_IMAGE007
optimum processor can calculate according to formula (6):
Figure 841459DEST_PATH_IMAGE039
（6）
For task
Figure 106219DEST_PATH_IMAGE007
, it is at processor on calculating for power consumption formula (7) represent, wherein
Figure 225932DEST_PATH_IMAGE040
it is processor
Figure 253931DEST_PATH_IMAGE008
at voltage
Figure 373197DEST_PATH_IMAGE014
under power consumption.
（7）
Like this, the activity energy consumption of all tasks can use formula (8) to represent.Wherein,
Figure 649643DEST_PATH_IMAGE042
,
Figure 215754DEST_PATH_IMAGE043
, be respectively task
Figure 81259DEST_PATH_IMAGE007
the maximum power dissipation of place processor, maximum frequency and maximum voltage.In addition, all processor free time can be consumed formula (9) and represent, makespanit is the scheduling length of the parallel task after task scheduling finishes.
（8）
Figure 572600DEST_PATH_IMAGE046
（9）
Message transfer
Figure 420120DEST_PATH_IMAGE047
the energy consuming can represent with formula (10).Like this, the network service energy consumption in whole system eLcan use formula (11) to represent.Wherein, PL is network service power consumption.
Figure 217175DEST_PATH_IMAGE048
（10）
（11）
Secondly, calculate scheduling length span.Task scheduling length while utilizing dispatching method that the application proposes to ask respectively threshold value for max_threshold and min_threshold – 1 is minimum scheduling length and maximum scheduling length.The calculating of scheduling length comprises following three steps: the first step, and under current threshold value, utilize the grouping strategy in step 4 to divide into groups to task; Second step, utilize scheduling strategy in step 5 by duty mapping to corresponding processor; The 3rd step, obtains the task scheduling length under current threshold value according to task scheduling result.Details can be referring to step 4 and step 5.
Finally, specify one can meet the scheduling length that user performance requires, from min_threshold – 1 to max_threshold, travel through all threshold values, utilize the dispatching method that this patent proposes to solve the task scheduling length under current threshold value, until the scheduling length of trying to achieve is less than or equal to, specify scheduling length just to finish traversal, current threshold value is now optimal threshold.
Four, task grouping.From first task of initiating task scheduling sequence, carry out depth-first search task is divided into groups.In task search procedure, if current task best forerunner (
Figure 682288DEST_PATH_IMAGE032
) not grouping, be assigned to
Figure 837195DEST_PATH_IMAGE003
the grouping at place, and be labeled as and distribute, otherwise to whether copying
Figure 933327DEST_PATH_IMAGE032
weigh.If copied
Figure 644931DEST_PATH_IMAGE032
, can increase replication task
Figure 447802DEST_PATH_IMAGE032
calculating energy consumption, reduce task scheduling time span and
Figure 275075DEST_PATH_IMAGE032
with
Figure 858503DEST_PATH_IMAGE003
between communication energy consumption, the time of minimizing is designated as lesstime, the energy consumption of increase is designated as moreenergy.If lesstime>0, and moreenergy is less than or equal to optimal threshold, by replication task
Figure 373798DEST_PATH_IMAGE032
add current task to
Figure 31175DEST_PATH_IMAGE003
the grouping at place, otherwise exit current group and choose first in initiating task scheduling sequence and do not divide group task to carry out next one to divide into groups.Wherein, the minimizing time , increase energy consumption
Figure 598609DEST_PATH_IMAGE033
.
The deadline of permission the latest (LACT) of export task equals its earliest finish time, and the deadline of permission the latest of other tasks is recursive calculation from top to bottom, as shown in Equation (12):
Figure 652015DEST_PATH_IMAGE051
（12）
Task
Figure 491795DEST_PATH_IMAGE003
the start time of permission the latest (LAST) equal it and allow the latest the deadline to deduct its execution time, LAST (
Figure 910138DEST_PATH_IMAGE003
) define as formula (13):
Figure 468159DEST_PATH_IMAGE052
（13）
Five, duty mapping.From first grouping, calculate and be respectively grouped in the energy consumption on all unappropriated processors, current group is mapped on the processor of energy consumption minimum, and this processor of mark is occupied, so circulation is gone down until the grouping of all tasks is all assigned with away.
Six, processor voltage regulates.Utilize the dependence exploration task free time between task, the execution time under each voltage on dispatch processor that sets the tasks makes task computation energy consumption minimum, then dynamic adjustments processor voltage in task implementation.
For task
Figure 745163DEST_PATH_IMAGE003
, its computation period and execution time, the relation of carrying out between frequency are:
Figure 439449DEST_PATH_IMAGE053
.If task
Figure 356590DEST_PATH_IMAGE003
after being finished, also there is free time under processor highest frequency/voltage, can save calculating energy consumption by the voltage/frequency reducing on its place processor.At this moment, computation period can be expressed as:
（14）
Here,
Figure 469219DEST_PATH_IMAGE055
(15)
According to the dependence between task, we can obtain duty mapping to processor actual Starting Executing Time (AST) and maximum allowable execution time (MAET) afterwards, and account form is as shown in formula (16) and formula (17).Wherein,
Figure 470542DEST_PATH_IMAGE056
expression task
Figure 293005DEST_PATH_IMAGE003
with
Figure 497721DEST_PATH_IMAGE057
in same grouping.
Figure 696621DEST_PATH_IMAGE058
（16）
（17）
Therefore, task execution time meet following relation:
Figure 739292DEST_PATH_IMAGE060
（18）
In the present invention, adopting dynamic electric voltage to regulate the object of (DVS) technology is that the execution time under each voltage of processor minimizes task computation energy consumption by exploration task.Here, we had both considered processor activity energy consumption, had considered again the idle energy consumption of processor.Therefore, the target equation of each task can be expressed as:
（19）
Above optimization problem is an integral linear programming problem, it take formula (14), (15), (18) is equation of condition, take formula (19) as target equation, can utilize integral linear programming to solve the integer optimum solution of instrument LP_solve execution time of trying to achieve task under each voltage of processor.

Claims (1)

1. the energy-saving scheduling method based on adaptive threshold in heterogeneous distributed system, is characterized in that: comprise the following steps:
One, first read parallel task directed acyclic graph DAG file;
Two, obtain initiating task scheduling sequence; For parallel task collection V, total n=|V| task, from export task
Figure 969973DEST_PATH_IMAGE001
start, calculate the priority of each task until beginning task
Figure 903294DEST_PATH_IMAGE002
finish; Then according to task priority size ascending order, arrange, obtain initiating task scheduling sequence;
Three, obtain optimal threshold; First, from first task of initiating task scheduling sequence, start to travel through all tasks, calculate and copy current task
Figure 509856DEST_PATH_IMAGE003
best forerunner (
Figure 768799DEST_PATH_IMAGE004
) the energy consumption moreenergy that increases, the energy consumption moreenergy that wherein increased equals calculating energy consumption deduct it with
Figure 951485DEST_PATH_IMAGE003
between communication energy consumption, get minimum value wherein and maximal value respectively as minimum threshold min_threshold and max-thresholds max_threshold; Then, task scheduling length while utilizing the computing method of scheduling length to ask threshold value to be respectively max_threshold and min_threshold – 1 is minimum scheduling length and maximum scheduling length, the computing method of described scheduling length comprise following three steps: the first step, and under current threshold value, utilize the grouping strategy in step 4 to divide into groups to task; Second step, utilize scheduling strategy in step 5 by duty mapping to corresponding processor; The 3rd step, obtains the task scheduling length under current threshold value according to task scheduling result; Finally, specify one can meet the scheduling length that user performance requires, from min_threshold – 1, start to travel through all threshold values, and utilize dispatching method to ask task scheduling length under current threshold value, until the scheduling length of trying to achieve is less than, specify scheduling length just to finish traversal, current threshold value is optimal threshold;
Four, task grouping; From first task of initiating task scheduling sequence, carry out depth-first search until beginning task finishes; In task search procedure, if current task
Figure 412554DEST_PATH_IMAGE003
best forerunner (
Figure 763770DEST_PATH_IMAGE004
) grouping, distributed to the grouping at current task place, and be labeled as and distribute, otherwise to whether copying
Figure 5395DEST_PATH_IMAGE004
carry out; If copied
Figure 483781DEST_PATH_IMAGE004
, can increase replication task calculating energy consumption, reduce task scheduling length and
Figure 150834DEST_PATH_IMAGE004
with
Figure 614176DEST_PATH_IMAGE003
between communication energy consumption; Suppose that the time reducing is lesstime, the time wherein reducing equals the scheduling length reducing, and the final energy consumption increasing is that moreenergy(equals
Figure 958570DEST_PATH_IMAGE004
calculating energy consumption deduct it with
Figure 394230DEST_PATH_IMAGE003
between communication energy consumption), if lesstime>0, and moreenergy is less than or equal to optimal threshold, by replication task
Figure 900298DEST_PATH_IMAGE004
add current task to
Figure 926635DEST_PATH_IMAGE003
the grouping at place, otherwise exit current group and choose first in initiating task scheduling sequence and do not divide group task to carry out next one to divide into groups;
Five, duty mapping; From first grouping, calculate and be respectively grouped in energy consumption on all unappropriated processors, current group is mapped on the processor of energy consumption minimum, and this processor of mark is occupied, so circulation is gone down until all groupings are all assigned with away;
Six, processor voltage regulates; The free time of utilizing the dependence between task to produce, sets the tasks the execution time under each voltage on dispatch processor, makes the calculating energy consumption of generation minimum, then in task implementation, dynamically adjusts processor voltage.
CN201110312108.0A 2011-10-14 2011-10-14 Self-adaptive threshold-based energy-saving scheduling method in heterogeneous distributed system Expired - Fee Related CN102360246B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201110312108.0A CN102360246B (en) 2011-10-14 2011-10-14 Self-adaptive threshold-based energy-saving scheduling method in heterogeneous distributed system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201110312108.0A CN102360246B (en) 2011-10-14 2011-10-14 Self-adaptive threshold-based energy-saving scheduling method in heterogeneous distributed system

Publications (2)

Publication Number Publication Date
CN102360246A CN102360246A (en) 2012-02-22
CN102360246B true CN102360246B (en) 2014-04-09

Family

ID=45585581

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201110312108.0A Expired - Fee Related CN102360246B (en) 2011-10-14 2011-10-14 Self-adaptive threshold-based energy-saving scheduling method in heterogeneous distributed system

Country Status (1)

Country Link
CN (1) CN102360246B (en)

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102650957A (en) * 2012-04-09 2012-08-29 武汉理工大学 Self-adaptive energy-saving dispatching method in isomorphic cluster system based on dynamic voltage regulation technology
CN102749987B (en) * 2012-06-20 2014-12-03 武汉理工大学 High energy efficiency resource allocating method for isomorphic cluster system of computer
CN103235640B (en) * 2013-01-08 2016-01-13 北京邮电大学 A kind of large-scale parallel task energy-saving scheduling method based on DVFS technology
CN103116526B (en) * 2013-02-22 2016-08-03 中国人民解放军国防科学技术大学 The maximum power dissipation control method of high-performance heterogeneous Computing machine
CN103176850A (en) * 2013-04-10 2013-06-26 国家电网公司 Electric system network cluster task allocation method based on load balancing
CN104102532B (en) * 2013-04-15 2017-09-26 同济大学 Research-on-research stream scheduling method based on low energy consumption in a kind of isomeric group
CN103399626B (en) * 2013-07-18 2016-01-20 国家电网公司 Towards Parallel application dispatching system and the method for the power-aware of hybrid compute environment
CN104239137B (en) * 2014-08-21 2017-12-08 东软集团股份有限公司 Multi-model Method of Scheduling Parallel and device based on DAG node optimal paths
CN104331326A (en) * 2014-11-25 2015-02-04 华南师范大学 Scheduling method and system for cloud computing
CN106339252B (en) * 2015-07-08 2020-06-23 阿里巴巴集团控股有限公司 Self-adaptive optimization method and device for distributed DAG system
CN106055409B (en) 2016-05-31 2017-11-14 广东欧珀移动通信有限公司 The distribution method and mobile terminal of a kind of processor resource
CN107302562B (en) * 2017-05-23 2019-12-03 中国科学院计算技术研究所 A kind of the adaptive command processing system and method for internet-of-things terminal equipment
CN107943561B (en) * 2017-12-14 2019-06-11 长春工程学院 A kind of scientific workflow method for scheduling task towards cloud computing platform
CN110674078B (en) * 2019-10-08 2020-11-10 北京航空航天大学 Digital twin system complex task heterogeneous multi-core parallel efficient solving method and system
CN111240818B (en) * 2020-01-09 2023-08-08 黔南民族师范学院 Task scheduling energy-saving method in heterogeneous GPU heterogeneous system environment
CN112631749B (en) * 2020-12-19 2024-03-26 北京化工大学 Scheduling method for parallel application programs in heterogeneous distributed system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101135927A (en) * 2006-10-12 2008-03-05 浙江大学 Simplifying method facing to embedded system low-power consumption real time task scheduling
CN101387952A (en) * 2008-09-24 2009-03-18 上海大学 Single-chip multi-processor task scheduling and managing method
CN101957780A (en) * 2010-08-17 2011-01-26 中国电子科技集团公司第二十八研究所 Resource state information-based grid task scheduling processor and grid task scheduling processing method
CN102207769A (en) * 2011-05-24 2011-10-05 东北大学 Static voltage scheduling-based energy optimization method of DVS (Dynamic Voltage Scaling) system

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2003083693A1 (en) * 2002-04-03 2003-10-09 Fujitsu Limited Task scheduler in distributed processing system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101135927A (en) * 2006-10-12 2008-03-05 浙江大学 Simplifying method facing to embedded system low-power consumption real time task scheduling
CN101387952A (en) * 2008-09-24 2009-03-18 上海大学 Single-chip multi-processor task scheduling and managing method
CN101957780A (en) * 2010-08-17 2011-01-26 中国电子科技集团公司第二十八研究所 Resource state information-based grid task scheduling processor and grid task scheduling processing method
CN102207769A (en) * 2011-05-24 2011-10-05 东北大学 Static voltage scheduling-based energy optimization method of DVS (Dynamic Voltage Scaling) system

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
《Optimal Scheduling Algorithm for Distributed-Memory Machines》;Sekhar Darbha等;《IEEE Transactions on Parallel and Distributed Systems》;19980131;第9卷(第1期);第87页至第95页 *
Sekhar Darbha等.《Optimal Scheduling Algorithm for Distributed-Memory Machines》.《IEEE Transactions on Parallel and Distributed Systems》.1998,第9卷(第1期),第87页至第95页.
Ziliang Zong等.《EAD and PEBD：Two energy-aware duplication scheduling algorithms for parallel tasks on homogeneous clusters》.《IEEE Transactions on computers》.2011,第60卷(第3期),第360页至第374页. *

Also Published As

Publication number Publication date
CN102360246A (en) 2012-02-22

Similar Documents

Publication Publication Date Title
CN102360246B (en) Self-adaptive threshold-based energy-saving scheduling method in heterogeneous distributed system
KR101629155B1 (en) Power-aware thread scheduling and dynamic use of processors
CN102508714A (en) Green-computer-based virtual machine scheduling method for cloud computing
CN109582448B (en) Criticality and timeliness oriented edge calculation task scheduling method
Changtian et al. Energy-aware genetic algorithms for task scheduling in cloud computing
CN102819460B (en) Budget power guidance-based high-energy-efficiency GPU (Graphics Processing Unit) cluster system scheduling algorithm
CN101169731A (en) Multiple-path multiple-core server and its CPU virtualization processing method
CN104199736A (en) Method for saving energy of data center under cloud environment
Jejurikar Energy aware non-preemptive scheduling for hard real-time systems
CN102650957A (en) Self-adaptive energy-saving dispatching method in isomorphic cluster system based on dynamic voltage regulation technology
CN103914346A (en) Group-based dual-priority task scheduling and energy saving method for real-time operating system
Kessaci et al. Parallel evolutionary algorithms for energy aware scheduling
Terzopoulos et al. Bag-of-task scheduling on power-aware clusters using a dvfs-based mechanism
CN103257900A (en) Real-time task set resource reservation method on multiprocessor for reducing CPU occupation
CN104102532B (en) Research-on-research stream scheduling method based on low energy consumption in a kind of isomeric group
CN103049326B (en) Method and system for managing job program of job management and scheduling system
Wo et al. Overbooking-based resource allocation in virtualized data center
Huang et al. Dynamic allocation/reallocation of dark cores in many-core systems for improved system performance
CN111562837A (en) Power consumption control method for multi-CPU/GPU heterogeneous server
Rajabi et al. Communication-aware and energy-efficient resource provisioning for real-time cloud services
Meng et al. Improvement of the dynamic priority scheduling algorithm based on a heapsort
Liu et al. An energy efficient clustering-based scheduling algorithm for parallel tasks on homogeneous DVS-enabled clusters
Ansari et al. Power-aware scheduling of fixed priority tasks in soft real-time multicore systems
CN115562812A (en) Distributed virtual machine scheduling method, device and system for machine learning training
Ghonoodi Green Energy-aware task scheduling using the DVFS technique in Cloud Computing

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20140409

Termination date: 20181014