CN106598717B - A kind of method for scheduling task based on time slice - Google Patents
A kind of method for scheduling task based on time slice Download PDFInfo
- Publication number
- CN106598717B CN106598717B CN201611113867.3A CN201611113867A CN106598717B CN 106598717 B CN106598717 B CN 106598717B CN 201611113867 A CN201611113867 A CN 201611113867A CN 106598717 B CN106598717 B CN 106598717B
- Authority
- CN
- China
- Prior art keywords
- task
- processor
- time
- coefficient
- execution
- 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.)
- Active
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
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/48—Indexing scheme relating to G06F9/48
- G06F2209/483—Multiproc
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/48—Indexing scheme relating to G06F9/48
- G06F2209/486—Scheduler internals
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multi Processors (AREA)
Abstract
The invention discloses a kind of method for scheduling task based on time slice.Its implementation is: system detection reaches or at the end of having task to there is new task, the task parameters of all tasks are reinitialized immediately, then each task is split into and executes the reasonable subtask of time slice length, and it determines the execution sequence of these subtasks and executes the processor of these subtasks, final updating schedule of tasks, it is executed according to new schedule of tasks scheduler task, and timely feedback the ending message that task execution is completed to task parameters initialization section above-mentioned, promote it to generate new schedule of tasks.The present invention can be used for isomorphism multicomputer system.
Description
Technical field
The invention belongs to operating system field more particularly to a kind of method for scheduling task based on time slice, can use
In isomorphism multicomputer system.
Background technique
Method for scheduling task is the critical problem that must be taken into consideration when designing high performance multicomputer system.With complete
How the aggravation of ball energy shortages reduces processor energy consumption and has caused more and more to pay close attention to.Therefore, a good task tune
Degree method can make system while guaranteeing service quality, moreover it is possible to run under alap rated power, to reach section
It can purpose.
Existing background technique generallys use open loop dispatching method as shown in Figure 1.Multiple tasks enter multiprocessor system
When system, system can be that each task estimates an execution time according to the size of each task.When a task is in some processing
When executing on device, this processor will be occupied always in estimation executes the time.It is all based in view of the estimation execution time the worst
What situation provided, so actually usually the compared estimate execution time will fulfil ahead of schedule task, in remaining this period later, i.e.,
Make running on processor there is no task, running frequency operation when task is but still executed with front, at this moment, is not only remained
Remaining this period is wasted, but also can generate unnecessary energy consumption.
The estimation execution time of task differs bigger with actual execution time, illustrates that the uncertainty of system environments is bigger,
The energy consumption of system wastes is also bigger.It is then desired to a kind of method for scheduling task be found, to overcome the uncertainty of system environments
Bring energy waste.
Summary of the invention
The present invention proposes a kind of method for scheduling task based on time slice, using the feedback information for the task that is completed come real
When adjust system schedule of tasks, so that system is constantly in optimum Working, thus achieve the purpose that reduce total energy consumption.
To achieve the above object, technical solution of the present invention includes:
A kind of method for scheduling task based on time slice, comprising the following steps:
Step 1: whenever system monitoring to the task ginseng for thering is new task to reach or at the end of having task execution, in system
Number is initialized；
Step 2: splitting task, and determines task execution time segment；
Step 3: it determines the sequence that executes of task, and exports schedule of tasks；
Step 4: system real-time update schedule of tasks successively controls each processor according to the sequencing of time slice
Execution task.
Further according to the method for scheduling task based on time slice, whenever system monitoring is to there is new post in step 1
At the end of business reaches or has task execution, the task parameters in system are initialized；
Total task number in system is N, the task in system is numbered with 1,2 ..., N, the set of number composition
It is expressed as I, I={ 1,2 ..., N }, then the set of all tasks can be expressed as T, T={ T in systemi|i∈I}；biIt indicates to appoint
The arrival time of business, xiThe estimation of expression task at least executes time, diThe execution time limit of expression task, then any in set T
Be engaged in TiUse Ti=(bi,xi,di) indicate；If task TiIn biMoment reaches, then it must be in next diIt is completed in time, i.e.,
Cut-off time is bi+di；
xiRequired time when executing the task in the worst cases for processor with maximum frequency, so the reality of task
It is minimum to execute time yiThan xiIt is small, if γiFor actual execution time estimation coefficient, then there is yi=γixi, 0 < γi≤ 1, γiMore
Small, the uncertainty of task parameters is bigger in system.
Further according to the method for scheduling task based on time slice, task is split in step 2, and determine and appoint
The time slice that business executes；
To all tasks in task-set T, successively make to map as follows to n-th task since first task: the 1st
Task T1Arrival time b1It is mapped to time instant τ0, cut-off time b1+d1It is mapped to time instant τ1, meet τ1-τ0=d1；2nd task
T2Arrival time b2It is mapped to time instant τ1, cut-off time b2+d2It is mapped to time instant τ2, meet τ2-τ1=d2；…；N-th task
TNArrival time bNIt is mapped to time instant τN-1, cut-off time bN+dNIt is mapped to time instant τN, meet τN-τN-1=dN；t1,t2Meet t1
=τ0< τ1< ... < τN=t2, then time interval [t1,t2] just by τ0,τ1,...,τNIt is divided into the timeslice of many segments
Section；Above-mentioned { τ is identified with μ0,τ1,...,τNIn element, μ ∈ U, U={ 0,1 ..., N-1 }, [t1,t2] in one when
Between segment can be used [τμ,τμ+1] indicate, it is ensured that at least need 1 processor can be in [t1,t2] time interior completion task-set T
In all task；
Include m identical processors, the running frequency f of each processor in system1,f2,...,fmax(f1< f2
< ... < fmax), it is indicated with running frequency class set Q={ 1,2 ..., l }, wherein 1 corresponding the smallest running frequency f1,
L corresponds to maximum running frequency fmax；The voltage/frequency of the processor adjusts DVFS technology by dynamic voltage frequency respectively
It is independently adjusted, multiple processors can execute different tasks respectively simultaneously with different speed grades；
Individual task in the task-set T can split into multiple subtasks to execute step by step, then determine that subtask will
In which execution time slice, and execute duration and running frequency grade；
Indicate a running frequency grade of processor with q, q ∈ Q,Referred to as load distribution coefficient, expression processing
Device will be in time interval [τμ,τμ+1] on it is a certain when it is a length ofTime interval in held with running frequency grade q
Row task Ti,Task T1Load distribution coefficient sets be
Indicate that processor will could execute the task totally 3 times on 2 time slices；In time slice [τ0,τ1] on, once with fortune
Line frequency grade 1 executes [τ0,τ1] in wherein 0.4 (τ1-τ0) second, [τ is once executed with running frequency grade 20,τ1] in
Wherein 0.2 (τ1-τ0) second, and this is executed twice must be ordered into execution, i.e., before equal it is primary execute after the completion of could execute under
It once, cannot be to execute while carrying out twice；In time slice [τ1,τ2] on, [τ is executed with running frequency grade 11,τ2] in
Wherein 0.3 (τ2-τ1) second, it is current execute complete after, then illustrate entire task T1It completes；
There is following constraint condition:
(1) task can only once be assigned on a processor and be performed, and then have
μ∈U；
(2) each processor is in [τμ,τμ+1] in the maximum load that can handle be 1, i.e., in [τμ,τμ+1] in the processor
On there is task executing always, and system includes m processor, therefore system is in [τμ,τμ+1] in the maximum load that can handle
As m then has
Relationship between the standardization operating rate s and running frequency f of processor is s=f/fmax,f∈[f1,fmax],s
∈ [0,1], corresponding power is by P (s at this timeq) provide, then in time interval [t1,t2] interior system total energy consumption
Method is equivalent to a purpose variable and constrains the controllable finite dimensional linear planning of conditional number amount in the step 2
Problem finds out load distribution coefficient of all tasks in each time sliceSo that system is in entire time interval
[t1,t2] in total energy consumption E (t1,t2) reach minimum；
To obtain load distribution coefficient setsIndicate that a certain processor will be in the time
It is spaced [τμ,τμ+1] in it is a certain when it is a length ofTime interval in running frequency grade q execute task Ti。
Further according to the method for scheduling task based on time slice, task is determined in step 3 executes sequence,
And export schedule of tasks；
All processors are numbered with 1,2 ..., m, the set expression of number composition is K, K={ 1,2 ..., m }；
I ∈ I indicates mission number；K ∈ K indicates processor number；Q ∈ Q indicates frequency Operation class；In time interval [t1,t2]
Interior sometime segmentIn, use starting coefficientsIndicate that processor k existsMoment
Start q in mode and runs task Ti, with cut-off coefficientIt indicatesMoment terminates the task
This is executed, and is specifically executed as follows:
(3-1) sets 0 to starting coefficients of all tasks in all time slices and cut-off coefficient, to system application one
A processor is used to execute task, claims the processor to be in and enlivens state, and other processors not being applied temporarily are known as Idle state,
Being now in and enlivening the processor maximum number of state is k=1；
The sequence that (3-2) is incremented by according to mission number i, since the 1st task, each task for being followed successively by task-set T
Determine starting coefficients and cut-off coefficient；To the 1st task, end coefficientIt is assigned a value ofSince the 2nd task
Until n-th task, using the cut-off coefficient of previous task as the starting coefficients of this task；
It is k that (3-3), which is now in and enlivens the processor maximum number of state, if the cut-off coefficient of previous task is plus this
The sum of load distribution coefficient of business is not more than processor maximum number k, then to all execution frequency level q, this task is being located
The starting coefficients that cut-off coefficient when on reason device k with speed grade q execution is set as this task add the sum of load distribution coefficient,
This task is assigned, then return step 3-2, starts next task distribution；
It is numbered if the cut-off coefficient of previous task is greater than processor maximum plus the sum of load distribution coefficient of this task
K then sets 1 for cut-off coefficient of this task when on processor k with speed grade q execution, illustrates allocated processor
Full load is had reached, therefore again to one processor of system application；Rule for application are as follows: processor its number of the 1st application is
The processor number of 1, the 2nd application is 2 ... ..., and the processor number of the m times application is m；If the number of application is more than task
The maximum quantity of processor, then illustrate system fault in execution module, this task schedule terminates, and is now in the place for enlivening state
Managing device maximum number is k=k+1；The load distribution coefficient of this task is subtracted into this task on processor k-1 with speed grade q
The difference of cut-off coefficient and starting coefficients when execution, as cut-off system of this task when on processor k with speed grade q execution
Number；Return step 3-2 starts next task distribution；
After (3-4) all tasks are assigned, output includes the starting coefficients of all tasksWith cut-off coefficientOne setThe collection is combined into schedule of tasks.
Further according to the method for scheduling task based on time slice, system real-time update task schedule in step 4
Table, according to time slice such as [τ0,τ1], [τ1,τ2] ... ..., [τN-1,τN] sequencing, successively control each processor and execute and appoint
Business；In time slice [τμ,τμ+1] in, according toWithValue, task execution module control processor k inMoment starts execution task T with running frequency grade qi, and in
Moment terminates the execution of this subtask；
Each task TiAt the end of, the ending message of the task is timely fed back to the input started by task execution module
End, when task executions all in system finish, this task schedule terminates.
Compared with the prior art, the invention has the following advantages:
1. in the present invention, being accelerated using the method for feedback by the information Real-time Feedback for the task that is completed to input terminal
Effective update of the system to schedule of tasks, the uncertain bring that can greatly reduce system environments influence, and system makes
The processing speed that the utilization rate and system of processor are total is promoted.
2. in the present invention, since task fulfils the remaining time ahead of schedule, can be recycled, to execute other
Business, this is all advantageous the total energy consumption for improving system performance and reduction system.
Detailed description of the invention
Fig. 1 is the realization frame diagram of the method for the invention；
Fig. 2 is the implementation flow chart of the method for the invention；
Fig. 3 is the exemplary diagram that individual task is split into multiple subtasks；
Fig. 4 is the exemplary diagram being mapped to multiple tasks by the task execution time limit on time shaft.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, below in conjunction with attached drawing, to of the present invention
Scheme and effect are described in further detail.
A kind of method for scheduling task based on time slice of the present invention, as depicted in figs. 1 and 2, specific implementation step
It is as follows:
Step 1: whenever system monitoring to the task ginseng for thering is new task to reach or at the end of having task execution, in system
Number is initialized.
Total task number in system is N, the task in system is numbered with 1,2 ..., N, the set of number composition
It is expressed as I, I={ 1,2 ..., N }, then the set of all tasks can be expressed as T, T={ T in systemi|i∈I}.Use biIt indicates
The arrival time of task, xiThe estimation of expression task at least executes time, diIt is the execution time limit of expression task, then any in set T
Task TiUse Ti=(bi,xi,di) indicate.If task TiIn biMoment reaches, then it must be in next diIt is completed in time,
I.e. cut-off time is bi+di。
xiRequired time when executing the task in the worst cases for processor with maximum frequency, so the reality of task
It is minimum to execute time yiThan xiIt is small, if γiFor actual execution time estimation coefficient, then there is yi=γixi, 0 < γi≤1。γiMore
Small, the uncertainty of task parameters is bigger in system.
Step 2: splitting task, and determines the time slice of task execution.
To all tasks in task-set T, successively make to map as follows to n-th task since first task: the 1st
Task T1Arrival time b1It is mapped to time instant τ0, cut-off time b1+d1It is mapped to time instant τ1, meet τ1-τ0=d1；2nd task
T2Arrival time b2It is mapped to time instant τ1, cut-off time b2+d2It is mapped to time instant τ2, meet τ2-τ1=d2；…；N-th task
TNArrival time bNIt is mapped to time instant τN-1, cut-off time bN+dNIt is mapped to time instant τN, meet τN-τN-1=dN.As shown in figure 3,
If t1,t2Meet t1=τ0< τ1< ... < τN=t2, then time interval [t1,t2] just by τ0,τ1,...,τNIt is divided into many
The time slice of segment.Above-mentioned { τ is identified with μ0,τ1,...,τNIn element, μ ∈ U, U={ 0,1 ..., N-1 }.[t1,
t2] in a time slice can be used [τμ, τμ+1] indicate, it is ensured that at least need 1 processor can be in [t1,t2] time
It is interior to complete task all in task-set T.
Include m identical processors, the running frequency f of each processor in system1,f2,...,fmax(f1< f2
< ... < fmax), it is indicated with running frequency class set Q={ 1,2 ..., l }, wherein 1 corresponding the smallest running frequency f1,
L corresponds to maximum running frequency fmax.It is independent that the voltage/frequency of processor adjusts DVFS technology by dynamic voltage frequency respectively
Ground is adjusted, and multiple processors can execute different tasks respectively simultaneously with different speed grades.
As shown in figure 4, the individual task in task-set T can split into multiple subtasks to execute step by step, then determine son
Which task execute time slice at, and executes duration and running frequency grade.
Indicate a running frequency grade of processor with q, q ∈ Q,Referred to as load distribution coefficient, expression processing
Device will be in time interval [τμ,τμ+1] on it is a certain when it is a length ofTime interval in running frequency grade q
Execution task Ti,Task T1Load distribution coefficient sets beIndicate that processor will could execute this totally 3 times on 2 time slices
Business.In time slice [τ0,τ1] on, [τ is once executed with running frequency grade 10,τ1] in wherein 0.4 (τ1-τ0) second, one
It is secondary that [τ is executed with running frequency grade 20,τ1] in wherein 0.2 (τ1-τ0) second, and this executes to be ordered into twice and hold
Row could execute next time after the completion of primary execution before waiting, and cannot execute while carrying out twice.In time slice [τ1,
τ2] on, [τ is executed with running frequency grade 11,τ2] in wherein 0.3 (τ2-τ1) second, it is current execute complete after, then say
Bright entire task T1It completes.
And it is rightThere is following constraint condition:
(1) task can only once be assigned on a processor and be performed, and then have
μ∈U。
(2) each processor is in [τμ,τμ+1] in the maximum load that can handle be 1, i.e., in [τμ,τμ+1] in the processor
On there is task executing always, and system includes m processor, therefore system is in [τμ,τμ+1] in the maximum load that can handle
As m then has
Relationship between the standardization operating rate s and running frequency f of processor is s=f/fmax,f∈[f1,fmax],s
∈ [0,1], corresponding power is by P (s at this timeq) provide, then in time interval [t1,t2] interior system total energy consumption
Method is equivalent to a purpose variable and constrains the controllable finite dimensional linear planning of conditional number amount in the step 2
Problem seeks load distribution coefficient of all tasks in each time sliceSo that system is in entire time interval [t1,
t2] in total energy consumption E (t1,t2) reach minimum.
And then obtain load distribution coefficient setsIndicate that a certain processor will be in the time
It is spaced [τμ,τμ+1] in it is a certain when it is a length ofTime interval in running frequency grade q execute task Ti。
Step 3: it determines the sequence that executes of task, and exports schedule of tasks.
All processors are numbered with 1,2 ..., m, the set expression of number composition is K, K={ 1,2 ..., m }.
I ∈ I indicates mission number；K ∈ K indicates processor number；Q ∈ Q indicates frequency Operation class.In time interval [t1,t2]
Interior sometime segment [τμ,τμ+1] in, use starting coefficientsIndicate that processor k existsMoment
Start q in mode and runs task Ti, with cut-off coefficientIt indicatesMoment terminates the task
This is executed.It executes as follows:
(3-1) sets 0 to starting coefficients of all tasks in all time slices and cut-off coefficient, to system application one
A processor claims the processor to be in and enlivens state, other processors not being applied temporarily are known as Idle state for executing task.
Being now in and enlivening the processor maximum number of state is k=1.
The sequence that (3-2) is incremented by according to mission number i, since the 1st task, each task for being followed successively by task-set T
Determine starting coefficients and cut-off coefficient.To the 1st task, end coefficientIt is assigned a value ofSince the 2nd task
Until n-th task, using the cut-off coefficient of previous task as the starting coefficients of this task.
It is k that (3-3), which is now in and enlivens the processor maximum number of state, if the cut-off coefficient of previous task is plus this
The sum of load distribution coefficient of business is not more than processor maximum number k, then to all execution frequency level q, this task is being located
The starting coefficients that cut-off coefficient when on reason device k with speed grade q execution is set as this task add the sum of load distribution coefficient.
This task is assigned, return step 3-2, is started as the distribution of next task；
It is numbered if the cut-off coefficient of previous task is greater than processor maximum plus the sum of load distribution coefficient of this task
K then sets 1 for cut-off coefficient of this task when on processor k with speed grade q execution, illustrates allocated processor
Full load is had reached, therefore again to one processor of system application.Rule for application are as follows: processor its number of the 1st application is
The processor number of 1, the 2nd application is 2 ... ..., and the processor number of the m times application is m.If the number of application is more than task
The maximum quantity of processor, then illustrate system fault in execution module, this task schedule terminates, and is now in the place for enlivening state
Managing device maximum number is k=k+1.The load distribution coefficient of this task is subtracted into this task on processor k-1 with speed grade q
The difference of cut-off coefficient and starting coefficients when execution, as cut-off system of this task when on processor k with speed grade q execution
Number.Return step 3-2 starts as the distribution of next task.
After (3-4) all tasks are assigned, output includes the starting coefficients of all tasksWith cut-off coefficientOne setFor schedule of tasks.
Step 4: system real-time update schedule of tasks, according to time slice such as [τ0,τ1], [τ1,τ2] ..., [τN-1,τN]
Sequencing, successively control each processor and execute task.In time sliceIt is interior, according toWith's
Value, task execution module control processor k inMoment starts to execute with running frequency grade q appoints
Be engaged in Ti, and inMoment terminates the execution of this subtask.
Each task TiAt the end of, the ending message of the task is timely fed back to the input started by task execution module
End.When task executions all in system finish, this task schedule terminates.
Above-mentioned is only to clearly illustrate example of the present invention, not limit the embodiments.For affiliated
For the those of ordinary skill in field, other various forms of variations or variation can also be made on the basis of the above description.
There is no necessity and possibility to exhaust all the enbodiments, and thus changes and variations that derived from are still
Among protection scope of the present invention.
Claims (2)
1. a kind of method for scheduling task based on time slice, which comprises the following steps:
Step 1: whenever system monitoring is to having new task to reach or at the end of having task execution, the task parameters in system into
Row initialization；
Step 2: splitting task, and determines task execution time segment；
Step 3: it determines the sequence that executes of task, and exports schedule of tasks；
Step 4: system real-time update schedule of tasks successively controls each processor and executes according to the sequencing of time slice
Task；
Wherein, whenever system monitoring is to there is new task to reach or task at the end of having task execution, in system in step 1
Parameter is initialized；Total task number in system is N, and the task in system is numbered with 1,2 ..., N, number composition
Set expression be I, I={ 1,2 ..., N }, then the set of all tasks can be expressed as T, T={ T in systemi|i∈I}；bi
The arrival time of expression task, xiThe estimation of expression task at least executes time, diThe execution time limit of expression task, then in set T
Any task TiUse Ti=(bi,xi,di) indicate；If task TiIn biMoment reaches, then it must be in next diIt is complete in time
At i.e. cut-off time is bi+di；
xiRequired time when executing the task in the worst cases for processor with maximum frequency, so task is practical minimum
Execute time yiThan xiIt is small, if γiFor actual execution time estimation coefficient, then there is yi=γixi, 0 < γi≤ 1, γiIt is smaller,
The uncertainty of task parameters is bigger in system；
Task is split in step 2, and determines the time slice of task execution；To all tasks in task-set T, from first
A task starts to n-th task successively to make to map as follows: the 1st task T1Arrival time b1It is mapped to time instant τ0, when cut-off
Carve b1+d1It is mapped to time instant τ1, meet τ1-τ0=d1；2nd task T2Arrival time b2It is mapped to time instant τ1, cut-off time b2
+d2It is mapped to time instant τ2, meet τ2-τ1=d2；…；N-th task TNArrival time bNIt is mapped to time instant τN-1, cut-off time bN
+dNIt is mapped to time instant τN, meet τN-τN-1=dN；t1,t2Meet t1=τ0< τ1< ... < τN=t2, then time interval [t1,
t2] just by τ0,τ1,...,τNIt is divided into the time slice of many segments；{ τ is identified with μ0,τ1,...,τNIn element, μ ∈
U, U={ 0,1 ..., N-1 }, [t1,t2] in a time slice can be used [τμ,τμ+1] indicate, it is ensured that it at least needs at 1
Managing device can be in [t1,t2] task all in task-set T is completed in the time；
Include m identical processors, the running frequency f of each processor in system1,f2,...,fmax(f1< f2< ...
< fmax), it is indicated with running frequency class set Q={ 1,2 ..., l }, wherein 1 corresponding the smallest running frequency f1, l correspondence
Maximum running frequency fmax；The voltage/frequency of the processor adjusts DVFS technology independently by dynamic voltage frequency respectively
It is adjusted, multiple processors can execute different tasks respectively simultaneously with different running frequency grades；
Individual task in the task-set T can split into multiple subtasks to execute step by step, then determine subtask will be at which
A execution time slice, and execute duration and running frequency grade；
Indicate a running frequency grade of processor with q, q ∈ Q,Referred to as load distribution coefficient indicates that processor will
In time interval [τμ,τμ+1] on it is a certain when it is a length ofTime interval in running frequency grade q execution
Task Ti,
There is following constraint condition:
(1) task can only once be assigned on a processor and be performed, and then have
(2) each processor is in [τμ,τμ+1] in the maximum load that can handle be 1, i.e., in [τμ,τμ+1] in one on the processor
Directly there is task executing, and system includes m processor, therefore system is in [τμ,τμ+1] in the maximum load that can handle be
M then has
Relationship between the standardization operating rate s and running frequency f of processor is s=f/fmax,f∈[f1,fmax],s∈[0,
1], at this time corresponding power by P (sq) provide, then in time interval [t1,t2] interior system total energy consumption
Method is equivalent to a purpose variable and constrains the controllable finite dimensional linear planning problem of conditional number amount in the step 2,
Find out load distribution coefficient of all tasks in each time sliceSo that system is in entire time interval [t1,t2]
Interior total energy consumption E (t1,t2) reach minimum；
To obtain load distribution coefficient setsIndicate that a certain processor will be in time interval
[τμ,τμ+1] in it is a certain when it is a length ofTime interval in running frequency grade q execute task Ti；
Task is determined in step 3 executes sequence, and exports schedule of tasks；With 1,2 ..., m compiles all processors
Number, the set expression of number composition is K, K={ 1,2 ..., m }；I ∈ I indicates mission number；K ∈ K indicates that processor is compiled
Number；Q ∈ Q indicates running frequency grade；In time interval [t1,t2] in sometime segmentIn, it is with starting
NumberIndicate that processor k existsMoment starts q in mode and runs task Ti, with cut-off coefficientIt indicatesMoment terminates this execution of the task, specifically executes as follows:
(3-1) sets 0 to starting coefficients of all tasks in all time slices and cut-off coefficient, at system application one
Reason device claims the processor to be in and enlivens state, other processors not being applied temporarily are known as Idle state, at this time for executing task
Numbering in the processor maximum for enlivening state is k=1；
The sequence that (3-2) is incremented by according to mission number i, since the 1st task, each task being followed successively by task-set T is determined
Starting coefficients and cut-off coefficient；To the 1st task, end coefficientIt is assigned a value ofSince the 2nd task always
To n-th task, using the cut-off coefficient of previous task as the starting coefficients of this task；
It is k that (3-3), which is now in and enlivens the processor maximum number of state, if the cut-off coefficient of previous task is plus this task
The sum of load distribution coefficient is not more than processor maximum number k, then to all running frequency grade q, this task is in processor k
On with running frequency grade q execution when cut-off coefficient be set as the starting coefficients of this task plus the sum of load distribution coefficient,
This task is assigned, then return step 3-2, starts next task distribution；
If the cut-off coefficient of previous task is greater than processor maximum number k plus the sum of load distribution coefficient of this task,
1 is set by cut-off coefficient of this task when on processor k with running frequency grade q execution, illustrates allocated processor
Full load is had reached, therefore again to one processor of system application；Rule for application are as follows: processor its number of the 1st application is
The processor number of 1, the 2nd application is 2 ... ..., and the processor number of the m times application is m；If the number of application is more than task
The maximum quantity of processor, then illustrate system fault in execution module, this task schedule terminates, and is now in the place for enlivening state
Managing device maximum number is k=k+1；The load distribution coefficient of this task is subtracted into this task on processor k-1 with running frequency
The difference of cut-off coefficient and starting coefficients when grade q is executed, as this task with running frequency grade q execution on processor k
When cut-off coefficient；Return step 3-2 starts next task distribution；
After (3-4) all tasks are assigned, output includes the starting coefficients of all tasksWith cut-off coefficient's
One setThe collection is combined into schedule of tasks.
2. according to claim 1 based on the method for scheduling task of time slice, which is characterized in that system is real-time in step 4
Schedule of tasks is updated, according to time slice such as [τ0,τ1], [τ1,τ2] ... ..., [τN-1,τN] sequencing, successively control is each
Processor executes task；In time slice [τμ,τμ+1] in, according toWithValue, task execution module control processing
Device k inMoment starts execution task T with running frequency grade qi, and inMoment terminates the execution of this subtask；
Each task TiAt the end of, the ending message of the task is timely fed back to the input terminal started by task execution module, when being
When all task executions finish in system, this task schedule terminates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611113867.3A CN106598717B (en) | 2016-12-07 | 2016-12-07 | A kind of method for scheduling task based on time slice |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611113867.3A CN106598717B (en) | 2016-12-07 | 2016-12-07 | A kind of method for scheduling task based on time slice |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106598717A CN106598717A (en) | 2017-04-26 |
| CN106598717B true CN106598717B (en) | 2019-06-11 |
Family
ID=58595325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611113867.3A Active CN106598717B (en) | 2016-12-07 | 2016-12-07 | A kind of method for scheduling task based on time slice |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106598717B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090153B (en) * | 2017-12-11 | 2020-04-14 | 深圳云天励飞技术有限公司 | Searching method, searching device, electronic equipment and storage medium |
| CN109002377A (en) * | 2018-07-26 | 2018-12-14 | 郑州云海信息技术有限公司 | A kind of processor detection method, processor detection device and computer equipment |
| WO2020073938A1 (en) * | 2018-10-10 | 2020-04-16 | 上海寒武纪信息科技有限公司 | Task scheduler, task processing system, and task processing method |
| CN109871394B (en) * | 2019-01-17 | 2022-11-11 | 苏宁易购集团股份有限公司 | Full-distributed high-concurrency calculation method and device |
| CN110213075B (en) * | 2019-04-04 | 2020-01-31 | 特斯联（北京）科技有限公司 | multi-network node configuration method, device, terminal equipment and medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1910553A (en) * | 2004-01-08 | 2007-02-07 | 皇家飞利浦电子股份有限公司 | Method and apparatus for scheduling task in multi-processor system based on memory requirements |
| CN101075694A (en) * | 2007-06-08 | 2007-11-21 | 清华大学 | Method for allocating task and elongating cell usage time to multi-state monoprocessor |
| CN101226421A (en) * | 2008-01-16 | 2008-07-23 | 浙江大学 | MSR method for real time embedded system EDF low-power consumption scheduling |
| CN103944997A (en) * | 2014-04-29 | 2014-07-23 | 上海交通大学 | Load balancing method with combination of random sampling and virtualization technology |
-
2016
- 2016-12-07 CN CN201611113867.3A patent/CN106598717B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1910553A (en) * | 2004-01-08 | 2007-02-07 | 皇家飞利浦电子股份有限公司 | Method and apparatus for scheduling task in multi-processor system based on memory requirements |
| CN101075694A (en) * | 2007-06-08 | 2007-11-21 | 清华大学 | Method for allocating task and elongating cell usage time to multi-state monoprocessor |
| CN101226421A (en) * | 2008-01-16 | 2008-07-23 | 浙江大学 | MSR method for real time embedded system EDF low-power consumption scheduling |
| CN103944997A (en) * | 2014-04-29 | 2014-07-23 | 上海交通大学 | Load balancing method with combination of random sampling and virtualization technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106598717A (en) | 2017-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106598717B (en) | A kind of method for scheduling task based on time slice | |
| Van Tilborg et al. | Foundations of real-time computing: Scheduling and resource management | |
| CN101620550B (en) | Embedded real-time scheduling method based on fuzzy multiple features of task | |
| CN101923487A (en) | Comprehensive embedded type real-time period task scheduling method | |
| CN106773711B (en) | A kind of the hybrid tasks scheduling method and model of railway locomotive operation steerable system | |
| CN108154317B (en) | Workflow group scheduling method based on example self-adaptive distribution integration in multi-cloud environment | |
| CN106775932B (en) | Real-time workflow scheduling method triggered by random event in cloud computing system | |
| Ströhle et al. | Online mechanism design for scheduling non-preemptive jobs under uncertain supply and demand | |
| CN111738446A (en) | Scheduling method, device, equipment and medium of deep learning inference engine | |
| CN111861412B (en) | Completion time optimization-oriented scientific workflow scheduling method and system | |
| CN111026519B (en) | Distributed task priority scheduling method and system and storage medium | |
| CN103745225A (en) | Method and system for training distributed CTR (Click To Rate) prediction model | |
| CN101237469A (en) | Method for optimizing multi-QoS grid workflow based on ant group algorithm | |
| CN108536530B (en) | Multithreading task scheduling method and device | |
| CN104915253A (en) | Work scheduling method and work processor | |
| CN113138860B (en) | Message queue management method and device | |
| Mittal et al. | Integrated dynamic scheduling of hard and QoS degradable real-time tasks in multiprocessor systems | |
| Anand et al. | Meeting deadlines: How much speed suffices? | |
| CN101996105A (en) | Static software/hardware task dividing and dispatching method for reconfigurable computing platform | |
| CN103314357B (en) | Perform for using multiple stage in the microprocessor and the method and apparatus dispatched is issued to instruction | |
| CN109308216A (en) | A kind of monokaryon system real time task dispatching method for imprecise computation | |
| CN113076181A (en) | Data processing flow optimization method, system and storage medium | |
| Liu et al. | On-line scheduling of real-time services with profit and penalty | |
| CN104484160A (en) | Instruction scheduling and register allocation method on optimized clustered VLIW (Very Long Instruction Word) processor | |
| Minh et al. | Modeling green-aware job scheduling problem in decentralized data centers |
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 |