CN109582448A - A kind of edge calculations method for scheduling task towards criticality and timeliness - Google Patents
A kind of edge calculations method for scheduling task towards criticality and timeliness Download PDFInfo
- Publication number
- CN109582448A CN109582448A CN201811208098.4A CN201811208098A CN109582448A CN 109582448 A CN109582448 A CN 109582448A CN 201811208098 A CN201811208098 A CN 201811208098A CN 109582448 A CN109582448 A CN 109582448A
- Authority
- CN
- China
- Prior art keywords
- task
- value
- criticality
- timeliness
- resource
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- 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
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mobile Radio Communication Systems (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of edge calculations method for scheduling task towards criticality and timeliness, utilize network flow optimization method, and comprehensively consider the criticality and actual effect of different task in marginal information processing, propose the edge calculations method for scheduling task towards criticality and timeliness, pass through the criticality of the information handling task of user's input and the time restriction of task execution, the consumption of calculating task process resource, establish task key degree and effective consumption graph model, using network flow optimization method, different tasks is scheduled and resource allocation.This method guarantees under the premise of resource fairness distributes to can not only carry out fair resource allocation to task, and criticality is high, effective high task can be assigned to more resources and be calculated, and faster completes.This method is primarily adapted for use in the task schedule at edge.
Description
Technical field
The invention belongs to task scheduling technique fields more particularly to a kind of edge calculations towards criticality and timeliness to appoint
Business dispatching method.
Background technique
In limbic system, often there are a large amount of different types of service operations in different calculate nodes, inhomogeneity
The different types of task of the service processing of type, it is also different for the demand of computing resource, if each task schedule cannot be arrived
Suitable node enables different types of task to possess relatively reasonable resource and is calculated, then will lead to the wave of resource
Take, i.e., is still likely to occur the free time of resource in the case where there is considerable task to overstock, causes task response-time too long.
Currently, the designer of task scheduling system appoints with the appearance of various distributed task dispatching technologies in order to improve
Business dispatching efficiency, reduces system response time, need to solve the problems, such as follows: 1) scheduler cannot be distinguished and treat longtime running service
Operation and batch quantity analysis operation;2) scheduler handle different type task needs to increase different types of scheduling strategy function, increases
The complexity of service logic and the complexity of configuration, so that mistake easily occur in deployment and O&M;3) scheduler will appear queue and prolong
It waits late and task can not rollback.In view of the above-mentioned problems, researcher proposes the dispatching method based on network flow.
However it is suitable for the calculating service environment of cloud center resources abundance and right currently based on the dispatching algorithm of network flow more
In the application scenarios that the execution time limit of service does not require, it is not suitable for limbic system, limbic system is in high confrontation, battlefield ring
The changeable environment in border, is needed important computations task in finite time, is completed using limited resources.If conventional scheduling method
It is used in limbic system, will cause the result is that treating on an equal basis to all tasks, the high task of key high-timeliness can not divide
Enough resources are fitted on, cause to wait in line, the phenomenon that low task of the low timeliness of criticality largely occupies limited resource.It can not
Guarantee that important edge task is completed in finite time.The particular problem of the existing dispatching algorithm based on network flow is as follows:
1) the consumption value calculating method of method does not account for task key degree and timeliness;
2) resource allocation methods of method do not account for task key degree and timeliness;
3) cost function of method does not account for task key degree and timeliness.
Summary of the invention
It is a primary object of the present invention to be in high confrontation, the limbic system of changeable environment provides a kind of task based access control pass
The edge method for scheduling task of key degree and timeliness, so that important needs to give the edge calculations task responded in finite time
Enough computing resources are obtained, preferentially ensure its completion.The method specifically includes following steps:
Step 1, the task key degree timeliness value of calculating task;
Step 2, according to task key degree timeliness value calculating task consumption value;
Step 3, flow network figure is modified according to runtime data, task consumption value and scheduling strategy.
Step 4, reasonable distribution is carried out to the resource that task obtains according to criticality and timeliness.
Step 1 includes the following steps:
Step 1-1, if task has n different criticalities, the criticality of i-th of task of note is Xi, the key of task
Degree is defined by the user, and i value is 1~n, calculates the opposite criticality pi of i-th of task according to the following formula:
Step 1-2 needs deadline Di to determine the timeliness value di of i-th of task according to i-th of task:
Step 1-3 calculates the criticality timeliness value wi of i-th of task according to the following formula:
Wi=pi*di;
Step 1-4 carries out minimum, maximum specification by criticality timeliness value of the following formula to all tasks:
Wherein, maxa indicates the maximum value in the criticality timeliness value of all tasks, and mina indicates the pass of all tasks
Minimum value in key degree timeliness value, newmaxa、newminaIt respectively indicates and needs to be mapped to the maximum value and minimum after standardization
Wi value, i.e., be mapped to [new by valuemina,newmaxa] in section, if newmina,newmaxaRespectively 0 and 1, that is, it is mapped to 0 and arrives
In 1 section.
Step 2 includes the following steps:
The consumption value wcosti of i-th of task is defined as:
Wherein, W1~Wn represents every dimension (cpu busy percentage dimension, memory usage dimension etc., can be according to the not people having the same aspiration and interest
Degree strategy increase dimensional information) weight, Wn indicate the n-th dimension weight, dimension include cpu occupancy, memory usage etc.,
Cpu occupancy and memory usage are respectively indicated with cpu, memory in formula (5), dn indicates remaining resource in its computer
Occupancy, user can be customized in their own needs, for example, if need network bandwidth occupancy minimum can another dn be
Collected Task Network bandwidth usage;The value range specification of every dimension turns to [0, c], and c is a constant value, to figure
Task associated cost is assigned a value of cost (u, v) when being constructed, and task associated cost cost (u, v) indicates task by calculate node
U is dispatched to the consumption executed after v, cpu, memory and network bandwidth comprising occupancy etc..
Step 3 includes the following steps:
Task schedule flow network is indicated that wherein V is node set with digraph G=(V, E, U, C), section by step 3-1
Point indicates resource supply and demand entity, including mission requirements side and resource provision side, and node is for indicating task, aggregation node, rack
With resource used in the scheduling such as machine.E is the set on side, indicates to be calculated in task schedule to corresponding machine;U indicates side
Capacity, in order to complete the stock number that particular task machine can be supplied at most;After C indicates task schedule to specific machine,
Running the task needs resource consumption to be used;
Step 3-2 finds all directed chains by starting point sv in digraph G to terminal tv, successively marks every chain
Maximum capacity simultaneously calculates corresponding minimal consumption using the minimal consumption function towards criticality and timeliness, if maximum hold
When amount is 0, then consumption is infinity, and the side that maximum capacity is 0 is labeled as | |, go to step 3-3;Otherwise, when all units
When expense is all 0, at this point, no longer there is the directed chain by sv to tv, augmentation cannot be carried out again, be at this time least cost maximum
Stream terminates scheduling;
Step 3-3 initializes digraph G, makes initial feasible flow 0, chooses the smallest directed chain of unit consumption,
The augmentation of maximum capacity is carried out to this directed chain cannot then carry out augmentation when no longer there is the directed chain by s v to t v again,
Terminate scheduling, is at this time minimal consumption max-flow.
In step 3-2, the minimal consumption function towards criticality and timeliness is as follows:
min∑(w,v)∈EWcost (w, v) f (w, v),
Wherein wcost (w, v), which represents in task schedule flow network task and is dispatched to calculate node v from calculate node w, executes
Consumption, f (w, v) indicate w to v task quantity.
Step 4 includes:
Set of tasks is set as T, set of physical resources M, Mission Scheduling is mapped as finding out from set T to set M
Side set so that consumption on side is minimum.
Task service set 1,2 ... is set, n has a resource requirement rs1, rs2 ..., rsn respectively, setting rs1≤rs2 <
=...≤rsn enables server have ability c.Validity when comprehensively considering resource requirement and mission critical, the present invention are taken the post as
The resource ASj calculation method of business j is as follows:
ASj is initially distributed to each task according to resource requirement by scheme from small to large.If after distribution, having and appointing
The distribution resource of business is more than required resource, then spare resources is continued to distribute by above-mentioned formula.
In step 4, in order to determine the side of T to M, the present invention is following triple task schedule constraint specification:
Schedule constraints=<T, R, B>
Wherein T indicates task, and R indicates the resource of task restriction, and it is obtained in resource that B indicates that task is scheduled for
Benefit.The present invention describes schedule constraints using benefit function, it may be assumed that
max∑B (7)
The present invention, which is converted to schedule constraints, decides whether the problem of establishing side between the different nodes of figure.The present invention is in T
A line T → R is established between R, each edge can be endowed a cost value.Cost is the value opposite with benefit B.Maximum return
The corresponding minimal consumption function of function is as follows
min∑cost (8)
The technical solution for realizing the object of the invention includes the following contents:
(1) task key degree and timeliness calculation method include the opposite criticality calculation method of task, timeliness calculating
Method, criticality timeliness method for normalizing.Task is calculated with respect to criticality calculation method according to the task key degree of artificial settings
Ratio shared by each criticality is to calculate opposite criticality.Timeliness is calculated according to the service response time specification of task, is made
It is higher to obtain the fewer timeliness of time specification.The mode that criticality timeliness combines comprehensively considers the influence of the two.
(2) task based access control is key and the fair resource share distribution method of timeliness according to criticality and timeliness to appointing
The resource that business obtains carries out reasonable distribution;When facing to group task distribution scarce resource, task is enjoyed right of equal value and is come
Resource is obtained, but the stronger task of criticality more high-timeliness should obtain more resources to guarantee the efficient fortune of key task
Row.At this moment the relatively low task of criticality actually only needs the resource fewer than other users.The present invention program is solved at this
How resource is distributed in the case of kind.The present invention is based on a kind of Resource Sharing Technologies widely used in practical applications, and " maximum is most
Small fair share " algorithm proposes to share algorithm with the minimax resource fairness of timeliness based on criticality.The algorithm is in justice
It is then preferential by key and timeliness on the basis of distributing to the desired satisfiable minimum essential requirement of each task
The resource allocation not used is given the needing large resource of the task by principle.
(3) the key consumption calculation method with timeliness of task based access control passes through task key degree and timeliness measure
Calculate the synthesis criticality and actual effect of each task, criticality height and the high task of timeliness can be in task call flow graphs
Obtain higher consumption;
(4) the task based access control criticality and the most minimal consumption max-flow optimization method of timeliness are the tune of dispatching algorithm
The core optimization process for spending stream continues to optimize Scheduling Flow according to the key cost function with timeliness of task based access control, so that
Criticality and the high task of timeliness can be assigned to relatively large number of consumption.Optimization method finds one every time and reaches from source point
The path of meeting point, increase flow, and the paths meet so that increased flow least cost, until can not find one from
Source point reaches the path of meeting point, and algorithm terminates, and flow can reach the maximum stream flow of network, while mission critical and timeliness at this time
Property cost function be that the high task of key high-timeliness distributes more various flow;Due to being increased the smallest flower every time
Take, i.e., consumption minimum value when current the least cost is all arrival present flow rates, therefore last wastage in bulk or weight is minimum.
Compared with the prior art, the present invention has the following advantages:
(1) for the dispatching algorithm currently based on network flow be suitable for more the calculating service environment of cloud center resources abundance with
For the application scenarios that the execution time limit of service does not require, it is not suitable for limbic system and adapts to high confrontation, the changeable ring in battlefield
Border needs important computations task in finite time, the problem of completion using limited resources, patent using task significance and when
The measure and resource allocation methods of effect property, so that the more resources of task acquisition that importance high-timeliness is high, reduce etc.
To.
(2) different degree height is preferentially ensured in scheduling process, the high task of timeliness enables these tasks in resource
More resources are obtained under conditions of fair allocat, faster complete task.
Detailed description of the invention
The present invention is done with reference to the accompanying drawings and detailed description and is further illustrated, it is of the invention above-mentioned or
Otherwise advantage will become apparent.
Fig. 1 is the task schedule based on network flow towards criticality and timeliness.
Fig. 2 is flow network modeling figure.
Fig. 3 is that flow network figure models abstract concept.
Fig. 4 is the distribution of fair resource share.
Fig. 5 is task consumption value example.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
Mission Scheduling is converted into such as Fig. 1 present invention the flow network for a task based access control criticality and timeliness
Carry out minimal consumption max-flow optimization problem.By combining task key degree and timeliness solving to minimal consumption max-flow
To optimal task schedule position.I.e. being executed in task schedule to most suitable calculate node.So that criticality is high
The high task of timeliness most can be used reasonably by different types of task under conditions of the supply of override Support Resource
Resource.Specific step is as follows:
1. according to the criticality and timeliness of mission bit stream calculating task.
Mission bit stream includes the timeliness of the particular content of task and the criticality of task and task.The timeliness of task
Property be meant that task needed to be completed within the time how long from that time being submitted, if this time is shorter,
Then illustrate that the more urgent timeliness of task is high, it is on the contrary then lower.
If task has n different criticalities, remember that the criticality of each task is Xi, value is higher to illustrate criticality more
It is high.Firstly, determining the opposite criticality of each task according to the value of Xi, formula (1) is seen:
Deadline Di is needed to determine the timeliness value of each safe task according to task:
Task key degree timeliness value is determined according to pi, di
Wi=pi*di (3)
In order to be calculated under same standard, the present invention carries out minimum-most after the wi value for calculating all tasks
Big specification, minimum-maximum specification carry out linear transformation to original value.Assuming that maximum value maxa and minimum value mina.Minimum-
Maximum specification passes through calculating:
Wherein, maxa indicates the maximum value in the criticality timeliness value of all tasks, and mina indicates the pass of all tasks
Minimum value in key degree timeliness value, newmaxa、newminaIt respectively indicates and needs to be mapped to the maximum value and minimum after standardization
Wi value, i.e., be mapped to [new by valuemina,newmaxa] in section, if newmina,newmaxaRespectively 0 and 1, that is, it is mapped to 0 and arrives
In 1 section.
2. according to task key degree timeliness value calculating task consumption value.
The monitor value when present invention is run according to cpu, memory usage when task key degree timeliness value and task run etc.
Calculating task consumption value wcost.The consumption value wcosti of i-th of task is defined as:
Wherein, w1~wn represents every dimension (cpu busy percentage dimension, memory usage dimension etc., can be according to the not people having the same aspiration and interest
Degree strategy increases dimensional information) weight, the value range specification of every dimension turns to [0, c], and c is a constant value.To figure
Task associated cost is assigned a value of cost (u, v) when being constructed.
3. modifying flow network figure according to runtime data, the consumption value of calculating and scheduling strategy.
The present invention program indicates that wherein V is node set task schedule flow network, with digraph G=(V, E, U, C),
Indicate resource supply and demand entity, including mission requirements side and resource provision, node is for indicating task, aggregation node, rack and machine
Resource used in the scheduling such as device.E is the set on side, indicates to be calculated in task schedule to corresponding machine.The appearance on U expression side
Amount, in order to complete the stock number that particular task machine can be supplied at most.After C indicates task schedule to specific machine, operation
The task needs resource consumption to be used.
For modeling result as shown in Figure 1, left side is set of tasks T, right side is set of physical resources M, M1 to M4, U1, U2 table
Show task.Side T0,2 → M1 indicates that T0,2 can be dispatched to M1 machine and execute up.T2,1 → U0 indicates that resource T2,1 does not have temporarily
It is scheduled for executing on machine.The capacity of each edge indicates that can task be dispatched on respective physical machine (capacity), consumes table
After showing in task call to machine, the task could be executed by needing to expend more upper resources.Modeling abstract concept meaning of the invention
See the table below with shown in Fig. 2, Fig. 3.
Table 1 models abstract concept and explains
The present invention is mapped as Mission Scheduling to find out the set from set T to the side of set M, so that the consumption on side
It is minimum.Under the premise of guaranteeing that the high task of importance high-timeliness preferentially executes, makes the more flexible property of scheduling of resource, mention simultaneously
High task schedule efficiency obtains high task schedule quality, high machine utilization rate and low dispatch delay.
4. determining the task of the task and scheduling that wait and determining deployment position
Using the minimal consumption max-flow constraint solver based on criticality and timeliness according to modified flow network,
The optimal solution of scheduling, the i.e. determination of the determination of optimal scheduling position and waiting and dispatching for task are determined in flow network, and are generated
New scheduling strategy is scheduled.
In order to find the set from set T to the side of set M, the present invention proposes that a kind of task based access control is key and timeliness
Fair resource share distribution method, for determining the side of T to U first, i.e. which task dispatching waits for.
Fairness in the present invention refers to that operation can take into account the crucial journey of task on the basis of Fairshare resource share
Degree and timeliness.For task j, it includes number of tasks be NT, calculating its fair share by fair algorithm is AS, if adjusted
The resource share that degree algorithm distributes to task j is AS, then the dispatching algorithm meets fairness.Fairness in the present invention passes through figure
Edge capacity construct to reach.As shown in Figure 4.
When facing to group task distribution scarce resource, task enjoys right of equal value to obtain resource, but criticality
The stronger task of more high-timeliness should obtain more resources to guarantee the efficient operation of key task.At this moment criticality is relatively
Low task actually only needs the resource fewer than other users.How the present invention program solution distributes money in this case
Source.The present invention is based on a kind of Resource Sharing Technology widely used in practical applications, " minimax fair share " algorithm is mentioned
Algorithm is shared with the minimax resource fairness of timeliness based on criticality out.The algorithm distributes to each task in fair share
On the basis of desired satisfiable minimum essential requirement, then by the key money that will not used with timeliness preferential principle
Distribute to the needing large resource of the task in source.
Consider task service set 1 ..., n has resource requirement rs1, rs2 respectively ..., rsn.It is assumed that rs1≤rs2≤...
≤ rsn enables server have ability c.Validity when comprehensively considering resource requirement and mission critical, the present invention are gone out on missions j's
Resource ASj calculation method is as follows:
ASj is initially distributed to each task according to resource requirement by scheme from small to large.If certain after distribution
It is more than required resource that task, which distributes resource, then spare resources is continued to distribute by above-mentioned formula.After the process, Mei Geren
Business obtains more without requiring than oneself, moreover, obtained resource will not be than other use if its demand is not being met
Most resources that family obtains are also few.The task that demand is not being met will be in waiting list.The program maximises money
The resource for user's smallest allocation that source is not being met.Example is as follows:
There is user group G, there are 4 users in the group, resource requirement is respectively 2,4,4,10, and weight is respectively 4,2.5,1,
0.5 total resources is 16.
(1) weight is standardized first, sets 1 for minimal weight, then weight becomes 8,5,2,1, summation 16.
Total resources are divided into 16 equal parts, four users respectively obtain 8,5,2,1.
(2) user more than 1 obtains 6 parts of resources, and user more than 2 obtains 1 part of resource, and user 3,4 resources are unsatisfactory for, therefore,
By extra 7 parts of resources according still further to weight distribution to user 3,4, user 3, and 4 obtain 7* (2/3), 7* (1/3) part money again respectively
Source;
(3) so far, user 3 obtains 6.666 parts of resources, and user 4 obtains 3.333, the resource that user has more is divided again
Provisioned user 4 completes distribution.
In order to determine the side of T to M, the present invention is following triple task schedule constraint specification:
Schedule constraints=<T, R, B>
Wherein T indicates task, and R indicates the resource of task restriction, and it is obtained in resource that B indicates that task is scheduled for
Benefit.The present invention describes schedule constraints using benefit function, it may be assumed that
max∑B (7)
The present invention, which is converted to schedule constraints, decides whether the problem of establishing side between the different nodes of figure.The present invention is in T
A line T → R is established between R, each edge can be endowed a cost value.Cost is the value opposite with benefit B.Maximum return
The corresponding minimal consumption function of function is as follows
min∑cost (8)
It now lifts simple examples and illustrates influence of the cost function for scheduling.If Fig. 5 task T1 is image processing tasks, need
It is dispatched on the server with image processor and is executed, existing two machines M1 and M2, it is now desired to which selection is image
It handles in task schedule to M1 or M2.M1 is configured with image processor, and M2 does not configure no image processor.T1 has scheduling to M1
Constraint, obtaining benefit is 1.T1 is to M2 without schedule constraints, benefit 0.By calculating minimal consumption function, i.e. greatest benefit letter
Number, T1 can be scheduled for M1 execution.
The present invention proposes that the minimal consumption function towards criticality and timeliness is as follows:
min∑(w,v)∈Ewcost(w,v)f(w,v) (9)
Wherein wcost (w, v) represents the consumption that task in network flow diagrams is dispatched to v execution from w, and f (w, v) indicates w to v's
Task quantity.
Its core concept of maximal flows at lowest cost derivation algorithm is maintenance max-flow, and continuous iteration finds least cost,
Solution procedure is as follows:
Step1 finds all directed chains by starting point sv in figure to terminal tv
(1) all directed chains by sv to tv are found;
(2) it successively marks the maximum capacity of every chain and is calculated using towards criticality and the minimal consumption function of timeliness
Corresponding minimal consumption out;
(3) if maximum capacity is 0, consumption is infinity, and by maximum capacity be 0 side labeled as " | | ",
Turn step2;Otherwise, when all unit costs are all 0, at this point, no longer there is the directed chain by sv to tv, it cannot
Again
Augmentation is carried out, at this point, being maximal flows at lowest cost, is terminated;
Step2 carries out augmentation to initial feasible flow
(1) former flow network figure is initialized, makes initial feasible flow 0;
(2) the smallest directed chain of unit consumption is chosen, the augmentation of maximum capacity is carried out to this directed chain;
(3) when no longer there is the directed chain by s v to t v, then augmentation cannot be carried out again, is terminated, at this point, being minimal consumption
Max-flow.
The present invention provides a kind of edge calculations method for scheduling task towards criticality and timeliness, implements the skill
There are many method and approach of art scheme, the above is only a preferred embodiment of the present invention, it is noted that this technology is led
For the those of ordinary skill in domain, various improvements and modifications may be made without departing from the principle of the present invention, these
Improvements and modifications also should be regarded as protection scope of the present invention.The available prior art of each component part being not known in the present embodiment
It is realized.
Claims (7)
1. a kind of edge calculations method for scheduling task towards criticality and timeliness, which comprises the steps of:
Step 1, the task key degree timeliness value of calculating task;
Step 2, according to task key degree timeliness value calculating task consumption value;
Step 3, flow network figure is modified according to runtime data, task consumption value and scheduling strategy;
Step 4, reasonable distribution is carried out to the resource that task obtains according to criticality and timeliness.
2. the method as described in claim 1, which is characterized in that step 1 includes the following steps:
Step 1-1, if task has n different criticalities, the criticality of i-th of task of note is Xi, and i value is 1~n, root
The opposite criticality pi of i-th of task is calculated according to following formula:
Step 1-2 needs deadline Di to determine the timeliness value di of i-th of task according to i-th of task:
Step 1-3 calculates the criticality timeliness value wi of i-th of task according to the following formula:
Wi=pi*di;
Step 1-4 carries out minimum, maximum specification by criticality timeliness value of the following formula to all tasks:
Wherein, maxa indicates the maximum value in the criticality timeliness value of all tasks, and mina indicates the criticality of all tasks
Minimum value in timeliness value, newmaxa、newminaIt respectively indicates and needs to be mapped to the maximum value and minimum value after standardization, i.e.,
Wi value is mapped to [newmina,newmaxa] in section, if newmina,newmaxaRespectively 0 and 1, that is, it is mapped to 0 to 1 area
Between in.
3. method according to claim 2, which is characterized in that step 2 includes the following steps:
The consumption value wcosti of i-th of task is defined as:
Wherein, W1~Wn represents the weight of every dimension, and Wn indicates the weight of the n-th dimension, divided in formula (5) with cpu, memory
Not Biao Shi cpu occupancy and memory usage, dn indicates the occupancy of remaining resource in its computer;The value model of every dimension
It encloses specification to turn to [0, c], c is a constant value, and task associated cost is assigned a value of cost (u, v), task when constructing figure
Associated cost cost (u, v) expression task is dispatched to the consumption executed after v by calculate node u.
4. method as claimed in claim 3, which is characterized in that step 3 includes the following steps:
Task schedule flow network is indicated that wherein V is node set, node table with digraph G=(V, E, U, C) by step 3-1
Show resource supply and demand entity, including mission requirements side and resource provision side;E is the set on side, indicates task schedule to corresponding machine
On calculated;U indicates the capacity on side, in order to complete the stock number that particular task machine can be supplied at most;C indicates task
After being dispatched to specific machine, running the task needs resource consumption to be used;
Step 3-2 finds all directed chains by starting point sv in digraph G to terminal tv, successively marks the maximum of every chain
Capacity simultaneously calculates corresponding minimal consumption using the minimal consumption function towards criticality and timeliness, if maximum capacity is
When 0, then consumption is infinity, and the side that maximum capacity is 0 is labeled as | |, go to step 3-3;Otherwise, when all unit costs all
When being 0, at this point, no longer there is the directed chain by sv to tv, augmentation cannot be carried out again, is at this time maximal flows at lowest cost, terminated
Scheduling;
Step 3-3 initializes digraph G, makes initial feasible flow 0, the smallest directed chain of unit consumption is chosen, to this
Directed chain carries out the augmentation of maximum capacity cannot then carry out augmentation when no longer there is the directed chain by s v to t v again, end
Scheduling, is at this time minimal consumption max-flow.
5. method as claimed in claim 4, which is characterized in that in step 3-2, the minimum towards criticality and timeliness
Cost function is as follows:
min∑(w,v)∈EWcost (w, v) f (w, v),
What wherein wcost (w, v) represented that task in task schedule flow network is dispatched to that calculate node v executes from calculate node w disappears
Consumption, f (w, v) indicate the task quantity of w to v.
6. method as claimed in claim 5, which is characterized in that step 4 includes:
Set of tasks is set as T, set of physical resources M, Mission Scheduling is mapped as finding out the side from set T to set M
Set so that consumption on side is minimum;
Task service set 1,2 ... is set, n has a resource requirement rs1, rs2 ..., rsn respectively, setting rs1≤rs2≤...≤
Rsn enables server have ability c, and the resource ASj calculation method of task j is as follows:
ASj is distributed to each task according to resource requirement from small to large, if having the distribution resource of task after distribution
More than required resource, then spare resources are continued to distribute by above-mentioned formula.
7. method as claimed in claim 6, which is characterized in that in step 4, in order to determine the side of T to M, about task schedule
Beam is described as following triple:
Schedule constraints=<T, R, B>,
Wherein T indicates task, and R indicates the resource of task restriction, and B indicates that task is scheduled for effect obtained in resource
Benefit describes schedule constraints using benefit function, it may be assumed that
max∑B (7)
Schedule constraints are converted to and decide whether the problem of establishing side between the different nodes of figure, a line is established between T and R
T → R, each edge can be endowed a cost value, and cost is the value opposite with benefit B, and the corresponding minimum of maximum return function disappears
It is as follows to consume function:
min∑cost (8)。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811208098.4A CN109582448B (en) | 2018-10-17 | 2018-10-17 | Criticality and timeliness oriented edge calculation task scheduling method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811208098.4A CN109582448B (en) | 2018-10-17 | 2018-10-17 | Criticality and timeliness oriented edge calculation task scheduling method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109582448A true CN109582448A (en) | 2019-04-05 |
CN109582448B CN109582448B (en) | 2021-03-16 |
Family
ID=65920628
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811208098.4A Active CN109582448B (en) | 2018-10-17 | 2018-10-17 | Criticality and timeliness oriented edge calculation task scheduling method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109582448B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110099384A (en) * | 2019-04-25 | 2019-08-06 | 南京邮电大学 | Resource regulating method is unloaded based on side-end collaboration more MEC tasks of multi-user |
CN110198339A (en) * | 2019-04-17 | 2019-09-03 | 浙江大学 | A kind of edge calculations method for scheduling task based on QoE perception |
CN110515716A (en) * | 2019-08-29 | 2019-11-29 | 中国科学院软件研究所 | It is a kind of to support priority and anti-affine cloud Optimization Scheduling and system |
CN110532078A (en) * | 2019-08-29 | 2019-12-03 | 中国科学院软件研究所 | A kind of edge calculations method for optimizing scheduling and system |
CN110647398A (en) * | 2019-09-18 | 2020-01-03 | 北京航空航天大学 | Intersection control task scheduling method facing edge calculation and based on task criticality and timeliness |
CN110995855A (en) * | 2019-12-13 | 2020-04-10 | 深圳先进技术研究院 | Microservice cluster scheduling method, scheduling device and computer readable storage medium |
CN111858051A (en) * | 2020-07-20 | 2020-10-30 | 国网四川省电力公司电力科学研究院 | Real-time dynamic scheduling method, system and medium suitable for edge computing environment |
CN112291335A (en) * | 2020-10-27 | 2021-01-29 | 上海交通大学 | Optimized task scheduling method in mobile edge calculation |
CN112466115A (en) * | 2019-12-30 | 2021-03-09 | 北京航空航天大学 | Bus intersection priority passing control system and method based on edge calculation |
WO2024114378A1 (en) * | 2022-12-01 | 2024-06-06 | 马上消费金融股份有限公司 | Resource allocation method, apparatus, and electronic device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103324525A (en) * | 2013-07-03 | 2013-09-25 | 东南大学 | Task scheduling method in cloud computing environment |
CN103838621A (en) * | 2012-11-27 | 2014-06-04 | 中国电信股份有限公司 | Method and system for scheduling routine work and scheduling nodes |
CN105677461A (en) * | 2015-12-30 | 2016-06-15 | 西安工业大学 | Mixed-criticality tasks scheduling method based on criticality |
CN105868190A (en) * | 2015-01-19 | 2016-08-17 | 中国移动通信集团河北有限公司 | Method and system for optimizing task processing in ETL |
-
2018
- 2018-10-17 CN CN201811208098.4A patent/CN109582448B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103838621A (en) * | 2012-11-27 | 2014-06-04 | 中国电信股份有限公司 | Method and system for scheduling routine work and scheduling nodes |
CN103324525A (en) * | 2013-07-03 | 2013-09-25 | 东南大学 | Task scheduling method in cloud computing environment |
CN105868190A (en) * | 2015-01-19 | 2016-08-17 | 中国移动通信集团河北有限公司 | Method and system for optimizing task processing in ETL |
CN105677461A (en) * | 2015-12-30 | 2016-06-15 | 西安工业大学 | Mixed-criticality tasks scheduling method based on criticality |
Non-Patent Citations (1)
Title |
---|
郭成昊 等: "基于协同效益的任务与资源匹配优化方法", 《指挥信息系统与技术》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110198339A (en) * | 2019-04-17 | 2019-09-03 | 浙江大学 | A kind of edge calculations method for scheduling task based on QoE perception |
CN110198339B (en) * | 2019-04-17 | 2020-08-21 | 浙江大学 | QoE (quality of experience) perception-based edge computing task scheduling method |
CN110099384A (en) * | 2019-04-25 | 2019-08-06 | 南京邮电大学 | Resource regulating method is unloaded based on side-end collaboration more MEC tasks of multi-user |
CN110099384B (en) * | 2019-04-25 | 2022-07-29 | 南京邮电大学 | Multi-user multi-MEC task unloading resource scheduling method based on edge-end cooperation |
WO2020216135A1 (en) * | 2019-04-25 | 2020-10-29 | 南京邮电大学 | Multi-user multi-mec task unloading resource scheduling method based on edge-end collaboration |
CN110515716B (en) * | 2019-08-29 | 2021-11-30 | 中国科学院软件研究所 | Cloud optimization scheduling method and system supporting priority and inverse affinity |
CN110515716A (en) * | 2019-08-29 | 2019-11-29 | 中国科学院软件研究所 | It is a kind of to support priority and anti-affine cloud Optimization Scheduling and system |
CN110532078A (en) * | 2019-08-29 | 2019-12-03 | 中国科学院软件研究所 | A kind of edge calculations method for optimizing scheduling and system |
CN110647398A (en) * | 2019-09-18 | 2020-01-03 | 北京航空航天大学 | Intersection control task scheduling method facing edge calculation and based on task criticality and timeliness |
CN110995855B (en) * | 2019-12-13 | 2022-06-21 | 深圳先进技术研究院 | Microservice cluster scheduling method, scheduling device and computer readable storage medium |
CN110995855A (en) * | 2019-12-13 | 2020-04-10 | 深圳先进技术研究院 | Microservice cluster scheduling method, scheduling device and computer readable storage medium |
CN112466115A (en) * | 2019-12-30 | 2021-03-09 | 北京航空航天大学 | Bus intersection priority passing control system and method based on edge calculation |
CN111858051A (en) * | 2020-07-20 | 2020-10-30 | 国网四川省电力公司电力科学研究院 | Real-time dynamic scheduling method, system and medium suitable for edge computing environment |
CN111858051B (en) * | 2020-07-20 | 2023-09-05 | 国网四川省电力公司电力科学研究院 | Real-time dynamic scheduling method, system and medium suitable for edge computing environment |
CN112291335A (en) * | 2020-10-27 | 2021-01-29 | 上海交通大学 | Optimized task scheduling method in mobile edge calculation |
CN112291335B (en) * | 2020-10-27 | 2021-11-02 | 上海交通大学 | Optimized task scheduling method in mobile edge calculation |
WO2024114378A1 (en) * | 2022-12-01 | 2024-06-06 | 马上消费金融股份有限公司 | Resource allocation method, apparatus, and electronic device |
Also Published As
Publication number | Publication date |
---|---|
CN109582448B (en) | 2021-03-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109582448A (en) | A kind of edge calculations method for scheduling task towards criticality and timeliness | |
EP2695325B1 (en) | Traffic scheduling device | |
CN107015856A (en) | Task scheduling approach generation method and device under cloud environment in scientific workflow | |
Nabi et al. | DRALBA: Dynamic and resource aware load balanced scheduling approach for cloud computing | |
CN105373426B (en) | A kind of car networking memory aware real time job dispatching method based on Hadoop | |
Bansal et al. | Dynamic task-scheduling in grid computing using prioritized round robin algorithm | |
CN104268018A (en) | Job scheduling method in Hadoop cluster and job scheduler | |
CN103401939A (en) | Load balancing method adopting mixing scheduling strategy | |
CN110187956B (en) | Layered real-time task scheduling method and system for multi-agent platform | |
CN106201681B (en) | Method for scheduling task based on pre-release the Resources list under Hadoop platform | |
CN109032769A (en) | A kind of continuous integrating CI task processing method and device based on container | |
CN108737462A (en) | A kind of cloud computation data center method for scheduling task based on graph theory | |
Li et al. | Endpoint-flexible coflow scheduling across geo-distributed datacenters | |
Biswas et al. | Multi-level queue for task scheduling in heterogeneous distributed computing system | |
CN111159859B (en) | Cloud container cluster deployment method and system | |
CN106407007B (en) | Cloud resource configuration optimization method for elastic analysis process | |
Singh et al. | Market-inspired dynamic resource allocation in many-core high performance computing systems | |
CN105138391B (en) | The multitasking virtual machine distribution method of cloud system justice is distributed towards wide area | |
Chen et al. | Deadline-constrained MapReduce scheduling based on graph modelling | |
Tahir et al. | UDRF: Multi-resource fairness for complex jobs with placement constraints | |
Li et al. | On scheduling of high-throughput scientific workflows under budget constraints in multi-cloud environments | |
Yang et al. | An offloading strategy based on cloud and edge computing for industrial Internet | |
CN108958919A (en) | More DAG task schedule expense fairness assessment models of limited constraint in a kind of cloud computing | |
CN108563497B (en) | Efficient multi-dimensional algorithm scheduling method and task server | |
CN104834571B (en) | A kind of data prefetching method applied to cloud workflow schedule |
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 |