CN105978720A - Service selection method satisfying end-to-end QoS constraint - Google Patents

Abstract

The invention provides a service selection method satisfying the end-to-end QoS constraint. The method comprises the following steps that (1) a global QoS constraint demand of a user is obtained, and the QoS value is normalized according to a normalization method; (2) historical QoS information of each individual server in combined services is collected, and the QoS value of each individual service in next time is predicted; (3) aimed at the basis structure of the combined services and different QoS characteristics and QoS prediction values, the global QoS constraint is decomposed to obtain the QoS constraint value of each individual service; and (4) and the service is selected by taking the QoS constraint values of the individual services as constraint. The time complexity, satisfying the end-to-end QoS constraint, of service combination is reduced, and combined services that satisfy the user requirement can be combined within linear time.

Description

A kind of method for service selection meeting end-to-end QoS constraint

Technical field

The invention belongs to Services Composition and self adaptation field, be specifically related to a kind of services selection meeting end-to-end QoS constraint Method.

Background technology

Services Composition has serviced to be formed new service by combination is multiple, thus meets what single service cannot meet User's request.When multiple services provide identical function, services selection is the problem that Services Composition must solve. The generally foundation of services selection is the non-functional attribute of service, the QoS attribute i.e. serviced.The QoS of service wraps Containing many aspects (such as: price, reliability, response time etc.), between these aspects, how choosing comprehensively is also service group The aspect closed and select to need consideration.Different QoS preference based on user, service is carried out selection is QoS The key problem of the Services Composition of perception.

But in the Services Composition demand of actual QoS perception, particularly at the Services Composition of " customer-centric " In, user only can propose the QoS constraint of the overall situation to composite services, and is not concerned with the QoS binding occurrence of individual services.And Services selection is then typically with the QoS binding occurrence of individual services as foundation.Therefore, the of overall importance and service choosing of user's request Wide gap is there is between the locality selected.That is, it would be desirable to select each on the premise of given overall situation QoS constraint Individual service is to complete services selection, thus realizes Services Composition.In order to solve this problem, research worker proposes such as The multiple distinct methods solved based on Multidimensional Knapsack Problems.But solving of these methods is np problem, it is difficult to limited Time in obtain optimal solution.In the case, how to complete QoS by the time complexity of reduction problem solving to feel The major issue of the Services Composition known.

Summary of the invention

In order to overcome above-mentioned the deficiencies in the prior art, the present invention provides a kind of services selection side meeting end-to-end QoS constraint Method.Present invention reduces the time complexity meeting the lower Services Composition of end-to-end QoS constraint, it is possible to group in linear session Close the composite services meeting user's request.

In order to realize foregoing invention purpose, the present invention adopts the following technical scheme that:

A kind of method for service selection meeting end-to-end QoS constraint, described method comprises the steps:

(1) the overall QoS obtaining user retrains demand, and according to method for normalizing, qos value is made normalized；

(2) collect the QoS historical information of each individual services in composite services, and predict that each individual services is next The qos value in moment；

(3) QoS characteristics for the basic structure of described composite services and different and QoS predictive value, by the overall situation QoS constraint is implemented to decompose, and obtains the QoS binding occurrence of each individual services；

(4) with the QoS binding occurrence of described individual services for constraint, service implementation selects.

Preferably, in described step (1), the described formula that qos value does normalized is as follows:

$\stackrel{\‾}{q_i}=\left|\frac{q_i-q_{i_{min}}}{q_{i_{max}}-q_{i_{min}}}\right|$

In formula, q_iFor the actual value of parameter,It is respectively optimum and the worst-case value of this qos parameter,For QoS Value after parameter normalization, after normalized, the codomain unified definition of all QoS attributes is interval interior in [0,1], And it is the best to be worth the biggest explanation performance.

Preferably, described step (2) comprises the steps:

Step 2-1, according to described step 1 method to service qos value make normalized；

Step 2-2, different autocorrelation performances based on QoS characteristic and according to service history QoS information, use based on The QoS of the service QoS forecasting mechanism prediction service subsequent time of extreme value.

Preferably, described step (3) comprises the steps:

Step 3-1, respectively will parallel according to order from the inside to surface by the structure of composite services, sequentially, select and circulate four Plant structure to convert, obtain described composite services outermost Services Composition structure；

Step 3-2, structure according to described composite services, be polymerized the QoS predictive value of service according to aggregating algorithm；

Step 3-3, basic structure according to described Services Composition, decompose the overall QoS constraint of service for basic structure QoS for individual services；

Step 3-4, iterative step 3-3, until all of overall situation QoS constraint all decomposes individual services；

Step 3-5, introduce coefficient of relaxation R, the QoS binding occurrence of individual services is relaxed.

Preferably, in described step 3-2, described aggregating algorithm comprises the steps:

Step 3-2-1, assume to have one group of service WS_i(i=1 ..., n), wherein each service WS_iAll there is one group of QoS attribute Q_i=(q_ij)_1*m；

Step 3-2-2, in the Services Composition of sequential organization, it is assumed that comprise service WS_i(i=1 ..., n), for cost q_ik, After combination, its cost is expressed asFor probabilistic QoS index q_ik, QoS index after combination It is expressed asIn kth component, wherein R_1nFor the mapping relations between QoS attribute, when QoS attribute one During cause, R_1nIt is 1；

In the Services Composition of parallel organization, it is assumed that comprise service WS_i(i=1 ..., n), for cost q_ik, after combination, Its cost is expressed asFor probabilistic QoS index q_ik, the QoS index after combination represents

In the Services Composition of choice structure, it is assumed that comprise service WS_i(i=1 ..., n), for cost QoS index q_ik, After combination, QoS index is expressed as Max{ (q_ikR_1n) | i=1 ... n}；For probabilistic QoS index q_ik, after combination QoS index is expressed as Min{ (q_ikR_1n)_k| i=1 ... n}；

In the Services Composition of loop structure, the special case of sequential organization, the therefore polymerization point of its cost are regarded in the polymerization of QoS as It is not expressed as Kq_ikWithWherein K is the cycle-index of loop structure.

Preferably, in described step 3-3, the described QoS that the overall QoS constraint of service is decomposed into individual services is profit QoS predictive value with serviceThe overall QoS constraint carrying out guide service is decomposed, and for serial structure, QoS divides Solve threshold valueIt is expressed as follows:

For parallel organization, QoS constraint exploded representation is as follows:

In formula, QoS_GFor overall situation QoS constraint.

Preferably, in described step 3-5, after introducing corresponding coefficient of relaxation, individual services S_iQoS threshold Revise as follows:

Wherein R (R ＞ 0).

Preferably, described step (4) comprises the steps:

Step 4-1, with the QoS binding occurrence of described individual services for according to all candidate service are screened, only servicing All QoS index be above service QoS binding occurrence service could retain as candidate service, further for user Select；

Step 4-2, with the preference Pre (U of user₁,U₂,...U_n) QoS property value (Q to all services₁,Q₂,...,Q_n) enter Row weighting, as follows:

$\underset{i=1}{\overset{n}{\Σ}}{U}_i*Q_i$

According to the result of weighted sum, candidate service is ranked up, thus selected user is the most satisfied, and satisfied end arrives The service of end QoS constraint；

Step 4-3, perform according to the result of services selection whether QoS polymerization meets with checking end-to-end QoS constraint, if It is unsatisfactory for, then according to the result of step 4-2, corresponding service is adjusted, until meeting end-to-end QoS constraint.

Compared with prior art, the beneficial effects of the present invention is:

The method of the present invention sets its binding occurrence without user for each Component service in composite services and realizes end-to-end QoS constraint Services Composition；The time complexity meeting the lower Services Composition of end-to-end QoS constraint can be reduced, it is possible to In linear session, combination meets the composite services of user's request；" customer-centric " characteristic, user can be in service QoS participates in its distribution according to its constraint preference in decomposing；The history QoS information serviced by utilization predicts service QoS Information, and with this QoS distribution carrying out guide service, there is more preferable real-time, and can reflect that a Component service is at combination clothes Importance in business.

Accompanying drawing explanation

Fig. 1 is a kind of method for service selection flow chart meeting end-to-end QoS constraint that the present invention provides,

Fig. 2 is the QoS time of the various different autocorrelation performances that the present invention provides to change schematic diagram,

Fig. 3 is the composite services structural representation meeting end-to-end QoS constraint that the present invention provides

Detailed description of the invention

Below in conjunction with the accompanying drawings the present invention is described in further detail.

Idea of the invention is that to retrain the overall QoS of user and decompose the constraint being decomposed into each service according to certain algorithm, Thus the services selection for QoS perception provides selection gist, and finally realize meeting the Services Composition of end-to-end QoS constraint.

It is the same with QoS polymerization that overall situation QoS constraint decomposition can regard one Reverse Problem of QoS polymerization in Services Composition as, Decomposition result is all had a major impact by the combinative structure between QoS and the service of single service.The principle that QoS decomposes is Decomposition result can react the QoS of each Component service importance in composite services QoS in QoS polymerization as far as possible.In order to the greatest extent Amount allows the QoS of service distribute rationally, and this method make use of the record to service history QoS information, and instructs based on this The decomposition of overall situation QoS.

As it is shown in figure 1, the present invention provides a kind of method for service selection meeting end-to-end QoS constraint, specifically comprise the following steps that

Step 1, the overall QoS of acquisition user retrain demand, and according to method for normalizing, qos value are made normalized；

According to application demand, obtaining the QoS demand of user, the QoS demand of user generally comprises multiple different QoS Characteristic, such as: time delay, expense, reliability etc., can represent with vector Q: Q=(q₁,q₂,q₃,...,q_n)；Then Use following method that QoS does normalized:

$\stackrel{\→}{q}=\left|\frac{q_i-q_{i_{min}}}{q_{i_{max}}-q_{i_{min}}}\right|$

Wherein q_iFor the actual value of parameter,It is respectively optimum and the worst-case value of this qos parameter, can be by system Meter draws, it is possibility to have association area expert or ISP are given,For the value after qos parameter normalization.Logical Crossing normalized, the codomain unified definition of all QoS attributes is in [0,1] is interval, and it is the best to be worth the biggest explanation performance. If response time is 0.7 will be better than response time 0.5, reliability 0.6 is also good than reliability 0.5.So QoS The comparison model of parameter is the most relatively easy.

Step 2, collect the QoS historical information of each individual services in composite services, and predict each individual services next The qos value in individual moment；

Step 2-1, according to step 1 method to service qos value make normalized；

Step 2-2, QoS for different autocorrelation performances predict its qos value:

With the autocorrelation performance of qos value as foundation, the QoS Changing Pattern of service substantially can be divided into periodically QoS, becomes Gesture qos value and the qos value etc. of randomness, as shown in Figure 2.

Assuming that need to predict current t_nThe qos value in moment, and the qos value in known k moment before this is sequence t_n-1,t_n-2,......,t_n-k, for different autocorrelation performances, point different situations are come pre-by QoS Forecasting Methodology based on extreme value Survey the QoS threshold of service.

1) periodically QoS

QoS time dependent relation such as Fig. 2 a of given service) shown in, the QoS threshold of service predict will with one time Between cycle T be the predetermined period as QoS, analyze t_nThe time cycle region at place (t, t+T], wherein t≤t_n≤ t+T, And utilize region K data before this to be intercepted, to ensure t≤t_n-i≤ t+T, obtains time series t_n-i,......,t_n-2,t_n-1, wherein i≤k.Carry out curve matching based on method of least square based on this sequence, obtain function F:t → (0,1], to obtain t_nF (the t in moment_n).Additionally, according to the historical information of QoS monitoring, this QoS can be obtained Maximum QoS in cycle_max.T is predicted based on this_nThe QoS threshold in moment is as follows:

2) tendency QoS

QoS time dependent relation such as Fig. 2 b of given service) shown in, i.e. the value of QoS is that the trend of stabilisation is sent out Exhibition.In the case, the QoS of given k time point before this, carry out curve matching based on method of least square accordingly, Obtain function f:t → (0,1], to obtain t_nF (the t in moment_n).For the QoS of tendency, at t_nThe service QoS in moment Threshold value is:

3) randomness QoS

QoS time dependent relation such as Fig. 2 c of given service) shown in, i.e. qos value change at random, does not has regularity. The most in all senses, therefore, service is at t in the QoS matching serviced in the case_nThe QoS threshold in moment can be defined into The maximum of QoS in section preset time, it may be assumed that

Step 3, basic structure for described composite services, and different QoS characteristics and QoS predictive value, will be complete Office's QoS constraint is implemented to decompose, and obtains the QoS binding occurrence of each individual services；

Step 3-1, by combinative structure as shown in Figure 3 according to order from the inside to surface respectively will order, parallel, circulation and Select four kinds of structures to convert, obtain composite services outermost Services Composition structure, and be polymerized it according to aggregating algorithm Qos value；

Step 3-1-1, according to following composite services QoS aggregating algorithm, it is thus achieved that the qos value of each composite services.

Assume there is one group of service WS_i(i=1 ..., n), wherein each service WS_iAll there is one group of QoS attribute vector Q_i=(q_ij)_1*m。

Services Composition for sequential organization, it is assumed that comprise service WS_i(i=1 ..., n), for cost q_ik(such as time, flower Take), after combination, its cost can be expressed asFor probabilistic QoS index q_ik(as reliably Property, availability, the QoS index after combination is represented byIn kth component；Wherein R_1nFor QoS Mapping relations between attribute, when QoS attribute is consistent, R_1nIt is 1.

In the Services Composition of parallel organization, it is assumed that comprise service WS_i(i=1 ..., n), for cost q_ik(such as time, flower Take), after combination, its cost can be expressed asFor probabilistic QoS index q_ik, after combination QoS index can represent

In the Services Composition of choice structure, it is assumed that comprise service WS_i(i=1 ..., n), for cost QoS index q_ik, After combination, QoS index can be expressed as Max{ (q_ikR_1n) | i=1 ... n}；For probabilistic QoS index q_ik, combination After QoS index can be expressed as Min{ (q_ikR_1n)_k| i=1 ... n}.In choice structure, it is the worst with every attribute One Virtual Service of value is as our candidate service.Composite services actual perform time, due to binding any one Candidate service is all than the QoS superior performance of this service, as long as therefore can ensure that the QoS polymerization of Services Composition can meet User's QoS demand end to end, then during service execution, QoS can be guaranteed certainly end to end.

In the Services Composition of loop structure, the polymerization of QoS can regard the special case of sequential organization as, therefore for QoS Index q_ik, the polymerization of its cost can be expressed as Kq_ikWithWherein K is the cycle-index of loop structure.

Step 3-1-2, replace as basic service with composite services, repeat 3-1-1 step, the most final composite services Only single structure position；

Step 3-2, according in step 3-1 calculate obtained by service QoS prediction, and as shown in Figure 3 combination clothes Business structure, is decomposed into the QoS binding occurrence of each atomic service by the overall QoS constraint of service.Base in view of Services Composition Between this structure can convertibility and composability, the most basic structure in Services Composition is serial structure and parallel organization.

Step 3-2-1, basic structure according to Services Composition, retrain QoS for basic structure by the overall QoS of service_GPoint Solve as individual services S_iQoS threshold

When considering the threshold decomposition of service, utilize the predictive value of service QoSCarry out the overall QoS of guide service about Bundle decomposes.For serial structure, the decomposition threshold of QoSCan be as follows:

For parallel organization, QoS constraint is decomposed the most as follows:

Wherein QoS_GFor overall situation QoS constraint.

Step 3-2-2, iterative step 3-2-1, until all of constraint all decomposes individual services position.

Step 3-2-3, due to services selection time, reach the less likely of services selection lower limit simultaneously, this algorithm also introduces Coefficient of relaxation R, suitably relaxes the QoS binding occurrence of individual services on the basis of predictive value, so that ensure can be more preferable Ground selects service.

Therefore, after introducing corresponding coefficient of relaxation, individual services S_iQoS thresholdCan revise as follows:

Wherein R (R ＞ 0).

Step 4, with the QoS binding occurrence of described individual services for constraint, service implementation selects.

Step 4-1, with the QoS binding occurrence of individual services obtained by calculating in step 3 for according to all candidate service are carried out Screening, all QoS index only serviced are above servicing the service of QoS binding occurrence and could retain as candidate service, With for further selection by the user；

Step 4-2, with the preference Pre (U of user₁,U₂,...U_n) QoS property value (Q to all services₁,Q₂,...,Q_n) enter Row weighting, as follows:

$\underset{i=1}{\overset{n}{\Σ}}{U}_i*Q_i$

According to the result of weighted sum, candidate service is ranked up, such that it is able to selected user is the most satisfied, and meets The service of end-to-end QoS constraint；

Step 4-3, perform according to the result of services selection whether QoS polymerization meets with checking end-to-end QoS constraint, as Fruit is unsatisfactory for, and is adjusted corresponding service according to the result of step 4-2, until meeting end-to-end QoS constraint.

Finally should be noted that: above example only in order to illustrate that technical scheme is not intended to limit, although Being described in detail the present invention with reference to above-described embodiment, those of ordinary skill in the field are it is understood that still The detailed description of the invention of the present invention can be modified or equivalent, and appointing without departing from spirit and scope of the invention What amendment or equivalent, it all should be contained in the middle of scope of the presently claimed invention.

Claims (8)

1. the method for service selection meeting end-to-end QoS constraint, it is characterised in that described method comprises the steps:

(1) the overall QoS obtaining user retrains demand, and according to method for normalizing, qos value is made normalized；

(2) collect the QoS historical information of each individual services in composite services, and predict that each individual services is next The qos value in moment；

(3) QoS characteristics for the basic structure of described composite services and different and QoS predictive value, by the overall situation QoS constraint is implemented to decompose, and obtains the QoS binding occurrence of each individual services；

(4) with the QoS binding occurrence of described individual services for constraint, service implementation selects.

Method for service selection the most according to claim 1, it is characterised in that in described step (1), described by QoS The formula that value does normalized is as follows:

$\stackrel{\‾}{q_i}=\left|\frac{q_i-q_{i_{\min }}}{q_{i_{max}}-q_{i_{min}}}\right|$

In formula, q_iFor the actual value of parameter,It is respectively optimum and the worst-case value of this qos parameter,For QoS Value after parameter normalization, after normalized, the codomain unified definition of all QoS attributes is interval interior in [0,1], And it is the best to be worth the biggest explanation performance.

Method for service selection the most according to claim 1, it is characterised in that described step (2) comprises the steps:

Step 2-1, according to described step 1 method to service qos value make normalized；

Step 2-2, different autocorrelation performances based on QoS characteristic and according to service history QoS information, use based on The QoS of the service QoS forecasting mechanism prediction service subsequent time of extreme value.

Method for service selection the most according to claim 1, it is characterised in that described step (3) comprises the steps:

Step 3-1, respectively will parallel according to order from the inside to surface by the structure of composite services, sequentially, select and circulate four Plant structure to convert, obtain described composite services outermost Services Composition structure；

Step 3-2, structure according to described composite services, be polymerized the QoS predictive value of service according to aggregating algorithm；

Step 3-3, basic structure according to described Services Composition, decompose the overall QoS constraint of service for basic structure QoS for individual services；

Step 3-4, iterative step 3-3, until all of overall situation QoS constraint all decomposes individual services；

Step 3-5, introduce coefficient of relaxation R, the QoS binding occurrence of individual services is relaxed.

The most according to claim 4, method for service selection, it is characterised in that in described step 3-2, described polymerization is calculated Method comprises the steps:

Step 3-2-1, assume to have one group of service WS_i(i=1 ..., n), wherein each service WS_iAll there is one group of QoS attribute Q_i=(q_ij)_1*m；

Step 3-2-2, in the Services Composition of sequential organization, it is assumed that comprise service WS_i(i=1 ..., n), for cost q_ik, After combination, its cost is expressed asFor probabilistic QoS index q_ik, QoS index after combination It is expressed asIn kth component, wherein R_1nFor the mapping relations between QoS attribute, when QoS attribute one During cause, R_1nIt is 1；

In the Services Composition of parallel organization, it is assumed that comprise service WS_i(i=1 ..., n), for cost q_ik, after combination, Its cost is expressed asFor probabilistic QoS index q_ik, the QoS index after combination represents

In the Services Composition of choice structure, it is assumed that comprise service WS_i(i=1 ..., n), for cost QoS index q_ik, After combination, QoS index is expressed as Max{ (q_ikR_1n) | i=1 ... n}；For probabilistic QoS index q_ik, after combination QoS index is expressed as Min{ (q_ikR_1n)_k| i=1 ... n}；

In the Services Composition of loop structure, the special case of sequential organization, the therefore polymerization point of its cost are regarded in the polymerization of QoS as It is not expressed as Kq_ikWithWherein K is the cycle-index of loop structure.

Method for service selection the most according to claim 4, it is characterised in that in described step 3-3, described by service Overall QoS constraint to be decomposed into the QoS of individual services be the QoS predictive value utilizing serviceCarry out guide service Overall situation QoS constraint is decomposed, for serial structure, the decomposition threshold of QoSIt is expressed as follows:

For parallel organization, QoS constraint exploded representation is as follows:

In formula, QoS_GFor overall situation QoS constraint.

Method for service selection the most according to claim 4, it is characterised in that in described step 3-5, corresponding introducing Coefficient of relaxation after, individual services S_iQoS thresholdRevise as follows:

Wherein R (R ＞ 0).

Method for service selection the most according to claim 1, it is characterised in that described step (4) comprises the steps:

Step 4-1, with the QoS binding occurrence of described individual services for according to all candidate service are screened, only servicing All QoS index be above service QoS binding occurrence service could retain as candidate service, further for user Select；

Step 4-2, with the preference Pre (U of user₁,U₂,...U_n) QoS property value (Q to all services₁,Q₂,...,Q_n) enter Row weighting, as follows:

$\underset{i=1}{\overset{n}{\Σ}}{U}_i*Q_i$

According to the result of weighted sum, candidate service is ranked up, thus selected user is the most satisfied, and satisfied end arrives The service of end QoS constraint；

Step 4-3, perform according to the result of services selection whether QoS polymerization meets with checking end-to-end QoS constraint, if It is unsatisfactory for, then according to the result of step 4-2, corresponding service is adjusted, until meeting end-to-end QoS constraint.