CN102143526B - Method for selecting sensor resource nodes based on energy balance and quality of service (QoS) constraints - Google Patents

Abstract

The invention discloses a method for selecting sensor resource nodes based on energy balance and quality of service (QoS) constraints. The method comprises the following steps of: classifying the QoS constraints of a task in a wireless sensor grid environment according to characteristics of limited sensor node energy and resource scheduling in the wireless sensor grid environment; differentiating influence degree of different QoS constraints levels on resource scheduling; preferably scheduling the task with higher QoS constraint level for the tasks with different QoS constraints levels under the condition of according with QoS constraints; for the task with the same QoS constraint level, evaluating a consumption value of resource scheduling in each time when the resources are selected by using an evaluation function; and selecting the task and the resources with the minimum consumption value as a mapping result according to the balance between energy consumption of the sensor nodes and power consumption of each node. In the method, the influence of energy consumption and the QoS constraints in the wireless sensor grid on resource scheduling when resources scheduling is performed is taken into comprehensive consideration.

Description

Sensor resource node selecting method based on balancing energy and QoS constraint

Technical field

The present invention relates to the technical field that grid computing and wireless sensor network combine, be specifically related to a kind of sensor resource node selecting method based on balancing energy and QoS constraint.

Background technology

In this year, along with the high speed development of wireless technology and sensor technology, the progress of the technology such as microelectronics, embedded system is promoting the fast development of wireless sensor network technology.Wireless sensor network consists of a series of sensor node, and each node has environment sensing, data are processed and wireless communication ability.Sensor node has powered battery, calculates the characteristics that storage capacity is limited, communication bandwidth is low, and this is restricted it when the disposal and utilization the data obtained.Grid computing by geographically distributing, the various resources of isomery link together, and form a virtual high-performance supercomputer, for the user provides, be available anywhere and computing capability reliably.Now, the grid that has high-speed computational capability, flood tide storage capacity and high-speed communication bandwidth characteristics has become solved a standard mode large-scale distributed, that heterogeneous resource is shared in the dynamic virtual community.

Along with going deep into of sensor technology and grid research, sensor node more and more is used as a kind of resource and has been incorporated in grid, thereby grid and sensor technology can be combined, utilize sensor node Real-time Obtaining resource, be shared by grid platform; The mass data that the computational resource that utilizes grid to have and storage resources are collected sensor node by technology such as data mining, data fusion, distributed data bases is processed, is analyzed and stored.

At present, the grid researcher puts forth effort on research grid resource scheduling (the selection problem of resource node), for a gridding task, selects suitable gridding resource node to be shone upon, and makes its maximizing the benefits in the process of scheduling.In recent years, the grid researcher has proposed a lot of Grid Resource Schedule Algorithms, but these dispatching algorithms are mainly designed for conventional mesh, and for the conventional mesh in different field, the goal in research of its Grid Resource Schedule Algorithms may the emphasis difference.As resource had to the high-performance calculation of requirement, the matter of utmost importance that resource scheduling algorithm need to be considered is QoS; The equipment limited along with various energy reserves more and more joins grid, an energy-optimised aspect that has also become Grid Resource Schedule Algorithms to pay close attention to.

With scheduling of resource in conventional mesh, compare, sensor node energy limited and the sensor grid task qos requirement to resource node in the wireless senser grid, must consider the impact of energy consumption and QoS constraint on scheduling of resource in the wireless senser grid while making scheduling of resource.Therefore, be necessary to provide the sensor resource node selecting method under a kind of applicable wireless sensor network lattice ring border.

Summary of the invention

The purpose of this invention is to provide the sensor resource node selecting method under a kind of applicable wireless sensor network lattice ring border, consider the impact of energy consumption and QoS constraint on scheduling of resource in the wireless senser grid.

To achieve these goals, the invention provides a kind of sensor resource node selecting method based on balancing energy and QoS constraint, comprise the steps:

(1) the QoS service according to the qos requirement of job invocation, all available resources provided is tested, and obtains the QoS constraint rank of each task, and described QoS constraint rank comprises rigid QoS constraint, soft level QoS constraint, level QoS constraint as possible;

(2) the QoS service according to the qos requirement of job invocation, all available resources provided is tested, and judges whether each task can be carried out on each resource, wherein can carry out and mean that resource meets the qos requirement of task;

(3) obtain the quantity of efficient resource corresponding to approximately intrafascicular each task of each rank QoS;

(4) first dispatch the task of rigid QoS constraint, dispatch again the task of soft level QoS constraint, finally scheduling is the task of level QoS constraint as possible, task for same rank QoS constraint, the task of efficient resource minimum number in this rank of priority scheduling QoS constraint task, the identical task for efficient resource quantity, consider the energy consumption of node and the balancing energy between each node simultaneously, adopt valuation functions to assess each task scheduling and try the consumption figures of carrying out on each resource, use the Min-min algorithm to find the task with minimum minimal consumption value, the described task with minimum minimal consumption value is assigned on the resource that obtains this minimum minimal consumption value.

Compared with prior art, the sensor resource node selecting method that the present invention is based on balancing energy and QoS constraint has following advantage:

(1) the QoS constraint of task has been carried out to classification, be respectively rigid level QoS constraint, soft level QoS constraint and level QoS constraint as possible, distinguish the priority of task according to the influence degree of different QoS constraint rank degree of exchanging, the priority of task scheduling that QoS constraint rank is high;

(2) on the basis of the QoS constraint that meets task, the energy consumption of consideration task in scheduling process, the aspect using the energy consumption of node as valuation functions is assessed scheduling consumption each time, the energy consumption of reduce resource node;

(3) when consideration is energy-optimised, the equilibrium that each node energy is consumed is also as a factor, select the more resource node of those dump energies in the process of scheduling as far as possible, thereby reach the equalization of whole energy consumption of sensor network nodes, improve reliability and the life cycle of sensor network.

By following description also by reference to the accompanying drawings, it is more clear that the present invention will become, and these accompanying drawings are for explaining embodiments of the invention.

The accompanying drawing explanation

Fig. 1 is the flow chart that the present invention is based on the sensor resource node selecting method of balancing energy and QoS constraint.

Embodiment

With reference now to accompanying drawing, describe embodiments of the invention, in accompanying drawing, similar element numbers represents similar element.

The present invention is based on the sensor resource node selecting method of balancing energy and QoS constraint in explanation before, first illustrate that lower sensor resource scheduling problem, the energy under the sensor grid environment of multi-QoS constraint that the method relates to consumes the relevant parameter in equalization problem, the description of sensor grid scheduling of resource environment, dispatching method.

Sensor resource scheduling problem under the multi-QoS constraint

Under the sensor grid scheduling of resource environment of multi-QoS constraint, task is many-sided to the qos requirement of resource, only considers that the QoS constraint of one dimension or apteryx is obviously the scheduling requirement that does not meet task.So, in the process that sensor resource is dispatched, must take into full account the problem of multi-QoS constraint.

To in sensor grid, resource there being the high-performance calculation of requirement, the submission task that is the user has more qos requirement to sensor node, therefore taking into full account the QoS(Quality of Service of task and resource, service quality) requirement is particularly important in dispatching algorithm.Under the sensor grid environment, according to task, the qos requirement difference of resource is carried out to classification to the QoS constraint of task::

(1) rigid level QoS: the QoS constraint to resource system platform, CPU and bandwidth etc. all belongs to rigid level QoS constraint;

(2) soft level QoS: expense, reliability etc. all belongs to soft level QoS constraint;

(3) a level QoS as possible: the QoS except rigid level QoS constraint and soft level QoS constraint retrains and all belongs to grade QoS as possible and retrain.

From top principle of classification, rigid level QoS is the highest to resource requirement in all rank QoS, scheduling of resource is played to conclusive effect, and the scheduling that only meets rigid level QoS is only effectively, and the task with this rigid level QoS could mapped execution.: if task meets soft level QoS, and scheduling is effective, and can make maximizing the benefits; If task does not meet soft level QoS, scheduling is effectively, but benefit reduces.Level QoS constraint is less on the impact of scheduling of resource as possible, and the QoS of the level as possible constraint of task will realize as far as possible and meet.

According to the classification of top three kinds of QoS constraint, consider the validity of resource to task, the sensor grid resource scheduling under the multi-QoS constraint is summed up as to the selection problem of task, i.e. the issue of priority of task choosing in the process of scheduling

Energy under the sensor grid environment consumes equalization problem

Because the sensor resource node in sensor grid is arranged in special occasions usually, use powered battery, power supply is non-exchange, so the power saving of sensor node becomes extremely important; Approach exhaustion due to some sensor node energy, may have influence on the service efficiency of other sensor node in whole zone, therefore the sensor node energy consumes balanced problem and also must consider (like this in the situation that do not affect the service efficiency of sensor node, at completing user, submit on the basis of task, extend as far as possible the life cycle of grid, before the energy of sensor resource node runs out of, make its maximization of utility).

Owing to needing the energy-conservation and energy of considering sensor node to consume balanced in the process of carrying out resource selection, therefore the sensor grid resource scheduling can be converted into to the selection problem of resource, for a gridding task, the resource of How to choose optimum is shone upon and is made sensor node reach balance on energy-conservation and balancing energy.

Sensor grid scheduling of resource environment is described

Consider that the sensor resource node has distributivity, isomerism, the characteristics such as wireless, sensor grid scheduling of resource environment is as follows:

(1) each gridding task to be scheduled is all independent task, countless according to relying on or communication between task;

(2) resource nodes can only be carried out a task at synchronization, until this task completes, could carry out other tasks, and task is monopolized a resource until complete;

(3) energy of each sensor resource node is limited, the energy initial value difference of different sensors resource node;

(4) the sensor resource node does not have energy consumption when idle condition;

(5) the energy consumption of sensor resource node is only limited to the consumption of tasks carrying energy;

(6) resource is to describe its available QoS method of service issue, and the mode that task is served with the QoS that describes its needs is submitted to.

Relevant parameter in dispatching method

retrain relevant parameter-definition with QoS:

(1) set T={t ₁, t ₂..., t _mthe individual independently task t of expression m _i(i=1,2 ..., set m);

(2) set R={r ₁, r ₂..., r _nmean n heterogeneous resource node r under grid environment _j(j=1,2 ..., set n);

(3) the expection time of implementation matrix of task

element et wherein _i,jwhen the QoS service that expression provides all available resources according to the qos requirement of job invocation is tested, testing out of task t _iat resource r _jon time of implementation, if resource meets the qos requirement of task, record its expection time of implementation, if resource r _jdo not meet task t _iqos requirement, et _i,jvalue be defined as ∞, resource can not meet the qos requirement of task, task t _ican not be at resource r _jupper execution;

(4) matrix carried out of task

element c wherein _i,jexpression task t _iat resource r _jon whether can carry out (be resource r _jwhether meet task t _iqos requirement), if can carry out c _i,jvalue be 1, mean resource r _jmeet appointed task t _iqoS constraint, the QoS constraint of this task may be rigid level QoS, soft level QoS or any in level QoS as possible; Can not carry out its value is 0, mean that this task is that rigid QoS constraint and corresponding resource can't meet the QoS constraint of task (Matrix C the carried out E reflection resource of task is for the validity situation of task, can draw according to expection time of implementation matrix ETC, each element in Ergodic Matrices ETC successively, if its value is numeral, can carries out element corresponding in Matrix C E and be set to 1; If the element value in ETC is ∞, can carry out element corresponding in Matrix C E and be set to 0, when the QoS service that the Matrix C the carried out E of task also can provide all available resources according to the qos requirement of job invocation is certainly tested, judge whether each task can be carried out and draw on each resource);

(5) QoS of task constraint rank array QoL=(m ₁, m ₂..., m _m) mean that the QoS of the individual independently task of m retrains rank, the QoS service all available resources provided according to the qos requirement of job invocation is tested, obtain the QoS constraint rank of each task, obtain the QoS constraint rank array QoL of task, its intermediate value is to be expressed as rigid QoS constraint at 2 o'clock, be the soft QoS constraint of 3 expression, be 4 and mean level QoS constraint as possible;

(6) task priority vector NoR1=(n ₁, n ₂..., n _n) mean efficient resource quantity corresponding to task that QoS constraint rank is rigid QoS, as n _ithe quantity of the efficient resource that the task i that expression QoS constraint rank is rigid QoS is corresponding is n _iif the QoS rank of corresponding task is for rigid QoS its value is defined as ∞, i.e. the maximum of system definition; Task priority vector NoR2=(n in like manner ₁, n ₂..., n _n) mean that QoS retrains efficient resource quantity corresponding to task that rank is soft QoS, task priority vector NoR3=(n ₁, n ₂..., n _n) mean that QoS retrains efficient resource quantity corresponding to task of rank for the level QoS that does the best, each component NoR of NoR vector _i(i=1,2,3) can be drawn by the Matrix C the carried out E of task, each component of NoR vector equal to carry out Matrix C E row and.

consume relevant parameter-definition with energy:

(1) energy parameter matrix $E={\left\{e_{i,j}\right\}}_{m\&times;n}=\left\{\begin{array}{cc}{e}_{\mathrm{1,1}}& {\mathrm{<figure-callout\; id="2"\; label="e"\; filenames="HDA0000055779930000011.png"\; state="\{\{state\}\}">e</figure-callout>}}_{\mathrm{1,2}}\\ .& .\\ .& .\\ .& .\\ e_{n,1}& e_{n,2}\end{array}\right\},$ This matrix is a n * 2 matrixes, wherein e _{i, 1}the residual energy value that means resource i, e _{i, 2}the execution energy expenditure rate (in the unit interval resource i execute the task consumed energy value) that means resource i;

(2) the execution energy of task consumes EoC _i,j.The execution energy of task consumes EoC _i,jfor task t _iat resource r _jon time of implementation et _i,jexecution energy expenditure rate e with resource j _{j, 2}product, as shown in the formula:

EoC _i,j=et _i,j*e _j,2 (4-1)

Resource r _jexecute task t _iafter dump energy be resource r _jresidual energy value e _{j, 1}consume EoC with the execution energy of task _i,jdifference, use difference e _{j, 1}-EoC _i,jreplace e in energy parameter matrix E _{j, 1}value.Task choosing during different resource node, the dump energy of sensor resource node will be as the reference foundation, preferentially selects resource that dump energy is maximum as mapping result.If there is the dump energy of a plurality of sensor resource nodes when identical, preferentially select energy to consume minimum resource as mapping result, thereby reach energy-conservation purpose.

the parameter-definition relevant with optimum span:

(1) expection deadline matrix

element ct wherein _i,jexpression task t _iat resource r _jon the expection deadline, from resource r _jresource r starts to execute the task _jexecute task t _ithe time spent;

(2) the maximum execution time Makespan of first resource scheduling.The maximum execution time Makespan of first resource scheduling equals the difference between the zero-time of maximum deadline of task and scheduling, from first task, starts to executing last task institute's time spent.

Suppose task t _iat resource r _jon Starting Executing Time be d _j, the expection time of implementation of task is et _i,j, task t _iat resource node r _jon deadline ct _i,jfor:

ct _i,j=d _j+et _i,j (4-2)

Can draw expection deadline matrix ECT according to formula (4-2).If task t _idistributed to resource r _jcarry out, resource r _jt finishes the work _itime of implementation ct _i=et _i,j.If zero-time is 0, the maximum execution time Makespan=max (ct that first resource is dispatched _i), t _i∈ T.

Consider the time loss that meets task QoS constraint in the once scheduling of energy consumption and formula (4-2) reflection in the once scheduling of formula (4-1) reflection, propose valuation functions TF (Tradeoff Function):

TF(i，j，qol)=1／qol×ct _i,j＋1／qol×EoC _i,j－(1－2／qol)×e _j,1

Wherein, i represents task number, and j means resource number, and qol means the Qos constraint rank of task i, and TF (i, j, qol) means task t _ibe assigned to resource r _jenergy consumption when upper and the integrated value of time loss, estimate the reasonability of scheduling of resource by this integrated value.

The QoS constraint rank of task is higher, and the proportion of optimum span (Optimal Makespan) in once dispatching is larger; The Qos constraint rank of task is lower, and it is larger that the node of sensor resource node and energy consume balanced proportion in once dispatching.For any other task of QoS confinement level, the optimum span ct of task _i,jconsume EoC with energy _i,jproportion in valuation functions TF is identical; And other descends along with the QoS confinement level, the dump energy e of sensor resource node _{j, 1}proportion in valuation functions TF increases, thereby the equilibrium that between outstanding sensor node, energy consumes preferentially selects the more sensor node of dump energy as mapping result.

Explanation now the present invention is based on the sensor resource system of selection of balancing energy and QoS constraint.With reference to figure 1, described method comprises the steps:

Step S1, the QoS service all available resources provided according to the qos requirement of job invocation is tested, if resource r _jmeet task t _iqos requirement, record its expection time of implementation, if resource r _jdo not meet task t _iqos requirement, expect that the time of implementation is recorded as ∞, the expection time of implementation of record is recorded in the expection time of implementation matrix ETC of task;

Step S2, each element of the expection time of implementation matrix ETC in traversal step S1 successively, its value is set to 1 by the element of the correspondence position of the Matrix C carried out of task E for numerical value, otherwise the element of correspondence position is set to 0;

Step S3, when the QoS service according to the qos requirement of job invocation, all available resources provided in step S1 is tested, obtain the QoS constraint rank of each task, and the QoS of each task constraint rank is recorded in QoS constraint rank array QoL, described QoS constraint rank comprises rigid QoS constraint, soft level QoS constraint, level QoS constraint as possible;

Step S4, determine the quantity of the efficient resource that approximately intrafascicular each task of each rank QoS is corresponding according to the Matrix C the carried out E of task, by approximately intrafascicular each task of rigid QoS, the quantity of corresponding efficient resource is recorded in task priority vector NoR1, by approximately intrafascicular each task of soft level QoS, the quantity of corresponding efficient resource is recorded to task priority vector NoR2, and the quantity of efficient resource corresponding to approximately intrafascicular each task of level QoS of doing the best is recorded in task priority vector NoR3;

Step S5, carry out non-descending to task priority vector NoR1, NoR2, NoR3;

Step S6, task in scheduler task priority vector NoR1, during scheduling, in priority scheduling task priority vector NoR1, (component value is exactly the some values in a vector to the component value minimum, by step, S4 obtains) task, the task that a plurality of component values are identical if having adopts valuation functions to assess each task scheduling and try the consumption figures of carrying out on each resource, use the Min-min algorithm to find and there is the task of minimum minimal consumption value (to each task, find out the minimal consumption value of this task in the consumption figures of its corresponding all resources, choose minimum task corresponding to minimal consumption value, Here it is Min-min algorithm),

Step S7, be assigned to the described task with minimum minimal consumption value on the resource that obtains this minimum minimal consumption value;

Step S8 deletes the described task with minimum minimal consumption value from task priority vector NoR1, upgrades task priority vector NoR1;

Step S9, judge whether task in task priority vector NoR1 all completes scheduling (be task all delete) in task priority vector NoR1, if not, goes to step S6, if so, and continuation step S10;

Step S10, task in scheduler task priority vector NoR2, during scheduling, the task of component value minimum in priority scheduling task priority vector NoR2, adopt valuation functions to assess each task scheduling and try the consumption figures of carrying out on each resource, use the Min-min algorithm to find the task with minimum minimal consumption value;

Step S11, be assigned to the described task with minimum minimal consumption value on the resource that obtains this minimum minimal consumption value;

Step S12 deletes the described task with minimum minimal consumption value from task priority vector NoR2, upgrades task priority vector NoR2;

Step S13, judge whether task in task priority vector NoR2 all completes scheduling (be task all delete) in task priority vector NoR2, if not, goes to step S10, if so, and continuation step S14;

Step S14, task in scheduler task priority vector NoR3, during scheduling, the task of component value minimum in priority scheduling task priority vector NoR3, adopt valuation functions to assess each task scheduling and try the consumption figures of carrying out on each resource, use the Min-min algorithm to find the task with minimum minimal consumption value;

Step S15, be assigned to the described task with minimum minimal consumption value on the resource that obtains this minimum minimal consumption value;

Step S16 deletes the described task with minimum minimal consumption value from task priority vector NoR3, upgrades task priority vector NoR3;

Step S17, judge whether task in task priority vector NoR3 all completes scheduling (be task all delete) in task priority vector NoR3, if not, goes to step S14, if so, and end.

As shown from the above technical solution, the sensor resource system of selection that the present invention is based on balancing energy and QoS constraint has following features:

1, consider the multi-QoS constraint of task, the energy-optimised and balancing energy of node, be applicable to the scheduling of resource under the sensor grid environment;

2, when being dispatched, the qos requirement of task is carried out to classification, the QoS of task constraint is divided into to rigid level QoS, soft level QoS and the level QoS that does the best, task to different stage QoS constraint, in the process of scheduling of resource, the influence degree according to QoS constraint to scheduling of resource, the higher task of QoS constraint rank is priority scheduling more, that is: to thering is the task of rigid QoS constraint, preferentially select resource preferentially to shine upon; To thering is the task of soft level QoS constraint, time preferential mapping of inferior preferential selection resource; To thering is the task of level QoS constraint as possible, finally select resource finally to shine upon;

3, retraining for same rank QoS of task, in the process of scheduling of resource, select the fewer priority of task scheduling of those available resources nodes, and consider energy consumption and the balancing energy of sensor node, assess the consumption figures of first resource scheduling by valuation functions, comprise optimum span, energy consumption and balancing energy, using the min-min algorithm to find the task with minimum minimal consumption value is dispatched, can guarantee in the situation that meet QoS constraint, the life cycle of maximized prolongation sensor grid;

4, considered the QoS constraint of energy consumption equalization and the task of sensor node: under the condition that meets the QoS constraint, consider the energy consumption problem of resource node in scheduling process, reduce the energy consumption in the resource node scheduling process as far as possible, use the more sensor resource node of dump energy in the process of scheduling as far as possible, make the energy of sensor node consume equalization, thereby in the situation that meet the service time that task scheduling increases each sensor node, the life cycle that extends sensor grid.

Above in conjunction with most preferred embodiment, invention has been described, but the present invention is not limited to the embodiment of above announcement, and should contain the various modifications of carrying out according to essence of the present invention, equivalent combinations.

Claims (1)

1. the sensor resource node selecting method based on balancing energy and QoS constraint, comprise the steps:

Step S1, the QoS service all available resources provided according to the qos requirement of job invocation is tested, if resource r _jmeet task t _iqos requirement, record its expection time of implementation, if resource r _jdo not meet task t _iqos requirement, expect that the time of implementation is recorded as ∞, the expection time of implementation of record is recorded in the expection time of implementation matrix ETC of task, wherein, t _ibe i task, i=1,2 ..., m; r _jbe j resource node, j=1,2 ..., n, the expection time of implementation matrix of task

Element et in formula _i,jwhen the QoS service that expression provides all available resources according to the qos requirement of job invocation is tested, testing out of task t _iat resource r _jon time of implementation, if resource meets the qos requirement of task, record its expection time of implementation, if resource r _jdo not meet task t _iqos requirement, et _i,jvalue be defined as ∞, resource can not meet the qos requirement of task, task t _ican not be at resource r _jupper execution;

Step S2, each element of expection time of implementation matrix ETC in traversal step S1 successively, its value is set to 1 by the element of the correspondence position of the Matrix C carried out of task E for numerical value, otherwise the element of correspondence position is set to 0, wherein, the matrix carried out of described task

Element c in formula _i,jexpression task t _iat resource r _jon whether can carry out, if can carry out c _i,jvalue be 1, mean resource r _jmeet appointed task t _iqoS constraint, the QoS constraint of this task is rigid level QoS, soft level QoS or any in level QoS as possible; Can not carry out its value is 0, means that this task is rigid QoS constraint and corresponding resource can't meet the QoS constraint of task;

Step S3, when the QoS service according to the qos requirement of job invocation, all available resources provided in step S1 is tested, obtain the QoS constraint rank of each task, and the QoS of each task constraint rank is recorded in QoS constraint rank array QoL, described QoS constraint rank comprises rigid QoS constraint, soft level QoS constraint, level QoS constraint as possible, the QoS constraint rank array QoL=(m of described task ₁, m ₂..., m _m) mean that the QoS of the individual independently task of m retrains rank, the QoS service all available resources provided according to the qos requirement of job invocation is tested, obtain the QoS constraint rank of each task, obtain the QoS constraint rank array QoL of task, its intermediate value is to be expressed as rigid QoS constraint at 2 o'clock, be the soft QoS constraint of 3 expression, be 4 and mean level QoS constraint as possible;

Step S4, determine the quantity of the efficient resource that approximately intrafascicular each task of each rank QoS is corresponding according to the Matrix C the carried out E of task, by approximately intrafascicular each task of rigid QoS, the quantity of corresponding efficient resource is recorded in task priority vector NoR1, by approximately intrafascicular each task of soft level QoS, the quantity of corresponding efficient resource is recorded to task priority vector NoR2, the quantity of efficient resource corresponding to approximately intrafascicular each task of level QoS of doing the best is recorded in task priority vector NoR3 to described task priority vector NoR1=(n ₁, n ₂..., n _n) mean that QoS retrains efficient resource quantity corresponding to task that rank is rigid QoS, n _ithe quantity of the efficient resource that the task i that expression QoS constraint rank is rigid QoS is corresponding is n _iif the QoS rank of corresponding task is for rigid QoS its value is defined as ∞, i.e. the maximum of system definition; Described task priority vector NoR2=(n ₁, n ₂..., n _n) mean that QoS retrains efficient resource quantity corresponding to task that rank is soft QoS; Described task priority vector NoR3=(n ₁, n ₂..., n _n) mean that QoS retrains efficient resource quantity corresponding to task of rank for the level QoS that does the best, each component NoR of NoR vector _i(i=1,2,3) are drawn by the Matrix C the carried out E of task, each component of NoR vector equal to carry out Matrix C E row and;

Step S5, carry out non-descending to task priority vector NoR1, NoR2, NoR3;

Step S6, the task in scheduler task priority vector NoR1, during scheduling, the task of component value minimum in priority scheduling task priority vector NoR1, wherein component value is exactly the some values in a vector, and by step, S4 obtains; The task that a plurality of component values are identical if having adopts valuation functions to assess each task scheduling and try the consumption figures of carrying out on each resource, use the Min-min algorithm to find the task with minimum minimal consumption value, described valuation functions TF (i, j, qol)=1/qol * ct _i,j+ 1/qol * EoC _i,j-(1-2/qol) * e _{j, 1}

Wherein, i represents task number, and j means resource number, and qol means the Qos constraint rank of task i, and TF (i, j, qol) means task t _ibe assigned to resource r _jenergy consumption when upper and the integrated value of time loss, estimate the reasonability of scheduling of resource by this integrated value;

Step S7, be assigned to the described task with minimum minimal consumption value on the resource that obtains this minimum minimal consumption value;

Step S8 deletes the described task with minimum minimal consumption value from task priority vector NoR1, upgrades task priority vector NoR1;

Step S9, judge whether the task in task priority vector NoR1 all completes scheduling, if not, goes to step S6, if so, continues step S10;

Step S10, task in scheduler task priority vector NoR2, during scheduling, the task of component value minimum in priority scheduling task priority vector NoR2, adopt valuation functions to assess each task scheduling and try the consumption figures of carrying out on each resource, use the Min-min algorithm to find the task with minimum minimal consumption value;

Step S11, be assigned to the described task with minimum minimal consumption value on the resource that obtains this minimum minimal consumption value;

Step S12 deletes the described task with minimum minimal consumption value from task priority vector NoR2, upgrades task priority vector NoR2;

Step S13, judge whether the task in task priority vector NoR2 all completes scheduling, if not, goes to step S10, if so, continues step S14;

Step S14, task in scheduler task priority vector NoR3, during scheduling, the task of component value minimum in priority scheduling task priority vector NoR3, adopt valuation functions to assess each task scheduling and try the consumption figures of carrying out on each resource, use the Min-min algorithm to find the task with minimum minimal consumption value;

Step S15, be assigned to the described task with minimum minimal consumption value on the resource that obtains this minimum minimal consumption value;

Step S16 deletes the described task with minimum minimal consumption value from task priority vector NoR3, upgrades task priority vector NoR3;

Step S17, judge whether the task in task priority vector NoR3 all completes scheduling, if not, goes to step S14, if so, finishes.

Images