CN103606042B - Services Composition instance migration availability deciding method based on dynamic dependency graph - Google Patents

Abstract

The present invention provides a kind of Services Composition instance migration availability deciding method based on executed path dynamic dependency graph, with develop before and after Services Composition, example to be migrated executed path for input, to migrate availability deciding result for output, its realization includes: analyze the dependence between activity in the executed path of example to be migrated, generate executed path dynamic dependency graph, and obtain arrival executed path terminate to locate all set up to variable-definition；By Dynamic Slicing method backward, yojan dynamic dependency graph；Obtain the controlling stream graph of the Services Composition after developing；Based on the controlling stream graph of Services Composition after developing and the dynamic dependency graph after yojan, it may be judged whether exist a topological sorting be develop after an active sequences of Services Composition, be met if there is the validity then migrated, otherwise be unsatisfactory for.The method utilizing the present invention can avoid the issuable false negative of conventional method, preferably supports the dynamic evolution of Services Composition.

Description

Services Composition instance migration availability deciding method based on dynamic dependency graph

Technical field

The present invention relates to service computing technique field, the especially judgement of Services Composition instance migration validity, in particular to A kind of Services Composition instance migration availability deciding method based on executed path dynamic dependency graph, is adapted to determine that before evolution Can the lower example that generate, that have not carried out end of Services Composition process definition, safely and effectively move to the service group after developing Continue executing with under the particular state definition of conjunction process.

Background technology

Along with distributed object technology and the development of XML technology, Web service technology occurs therewith and tends to ripe.Web service It is the effective mechanism of Data and Information integration on Web, by using the standards based on XML such as WSDL, UDDI and SOAP And agreement, solve the problem such as code reuse, heterogeneous distributed calculating, have that interoperability is strong, loose coupling, the spy such as cross-platform Property.In order to ensure the reusability of single service, single Web service granularity is unsuitable excessive, thus when single simple services can not When meeting user's request, be by the service of plural number function is effectively integrated by certain service describing and constraint, it is achieved Yong Huding The Services Composition of justice, to provide value-added service.Web service combination has level, extensibility, dynamically and adaptivity etc. Feature, by reusing existing service, automatically generates new service or system, greatly improves the production efficiency of software.

Owing to the market demand, technology innovation, regulation change etc. reason causes the external environment condition of enterprise to open changeable, business procedure Model also the most constantly to develop.Kernel-based methods model, the respective service example combinations that is currently running can therefore suffer from impact, Need by terminating, restart and the means such as migration dynamically processing.Compared to the method for other poor efficiency high consumption, example dynamic Migration has more flexibility, will one develop before service composition workflow example from original Work flow model current state, Move under the Work flow model dbjective state after developing, it is ensured that the maximization worked.

In order can continue executing with according to the Services Composition process after developing after ensureing instance migration, (its subsequent execution is developing with one The lower example behavior generated of rear Services Composition definition is consistent), it would be desirable to the Services Composition after evolution finds one with to be migrated The corresponding dbjective state that example current state is consistent.If finding so one corresponding dbjective state, just can be after instance migration Continue executing with according to the Services Composition definition after developing under this state.But, if (real by example state considers data mode The current value of example correlated variables), the transportable property of example is undecidable.The undecidability of this problem makes us to look for Migration validity is verified to a computable sufficient and necessary condition.Therefore, we can only find the checking of adequate condition and move Move validity.

Existing method mostly by abstract for the data message of example fall, thus the control stream mode being based only on example carries out instance migration The judgement of validity.The representative work of this kind of method have that Aalst et al. proposes based on process model behavior inheritance pass before and after developing The workflow instance transportable sex determination method of system, but the method can not process the Work flow model before and after evolution and be unsatisfactory for inheriting The situation of relation, does not the most have generality.Aalst also proposed a kind of workflow migration side not limiting Workflow logs type Method, the method, by comparing the Work flow model before and after change, finds out all regions that there occurs change in Work flow model, as Really workflow instance is not at region of variation, then under example allows the corresponding states of the Work flow model after moving to evolution.Due to this Class method does not accounts for the data mode of example, is therefore not appropriate for the Services Composition dynamic evolution scene of data perception.

Though having certain methods to consider data mode in example state, but existing method being the most conservative, seldom Instance migration is considered effective.Such as: Casati et al. utilizes " track reproduction " technology to judge that the migration of Services Composition example has Effect property, if the executed active sequences of an i.e. example can reappear under the process model developed, then this instance migration is to have Effect.The concept of " track reproduction " has been incorporated in the method for entitled ADEPT by Rinderle and Reichert et al., the method Executed active sequences has done some process, after deleting evolution during the activity of disappearance only remaining in loop structure After the once activity arrived performed by circulation, thus obtain the track after a yojan (i.e. active sequences), if after this yojan Active sequences can reappear under the Services Composition process after evolution, then the migration of respective instance is effective.Although the method Initial " track reproduction " technology has been done some and loosened, but do not accounted for the data independence between activity, and therefore the method depends on The most conservative.

Summary of the invention

The defect existed for prior art or deficiency, it is desirable to provide a kind of Services Composition example based on dynamic dependency graph Migrate availability deciding method, be adapted to determine that reality that generate under the definition of the Services Composition process before evolution, that have not carried out end Can example, safely and effectively move to continue executing with under the particular state definition of the Services Composition process after developing, be avoided that tradition The more issuable false negatives of method (false negative), such that it is able to allow more example migrate, preferably supports The dynamic evolution of Services Composition.

For reaching above-mentioned purpose, the technical solution adopted in the present invention is as follows:

A kind of Services Composition instance migration availability deciding method based on dynamic dependency graph, with develop before and after Services Composition, with And the executed path of example to be migrated is input, with the migration availability deciding result of example to be migrated for output, including following Step:

Step 1, analyze the dependence between activity in the executed path of example to be migrated, automatically generate this executed path Dynamic dependency graph, the dynamic dependency graph of example the most to be migrated, and obtain and arrive executed path and terminate all fixed up to variable of place The set of justice (reaching definitions of variables)；

Step 2, yojan dynamic dependency graph: by the method for Dynamic Slicing backward, delete and do not have influential activity on instance migration And corresponding dependence edge；

Step 3, to develop after Services Composition use control flow analysis method obtain controlling stream graph；And

Dynamic dependency graph after step 4, controlling stream graph based on the Services Composition after developing and example yojan to be migrated, it is judged that about Whether there is a topological sorting (i.e. active sequences) under dynamic dependency graph after letter to reappear under the Services Composition after evolution, I.e. judge whether topological sorting be develop after an active sequences of Services Composition, if it is present migrate is effective Property is met, otherwise is unsatisfactory for.

Further, in the implementing of described step 1, the executed path representation of example to be migrated is σ={ σ₁,σ₂... }, σ₁, σ₂... expression activity, movable σ_iAll input variables be inPut (σ_i), all output variables are outPut (σ_i)；Between activity Dependence is expressed as R=(σ_i,σ_j), σ_iAnd σ_jRepresenting the activity in aforementioned execution route, it comprises control dependence, different Step dependence and data dependence relation, and represented in dynamic dependency graph DDG by the form of directed edge；To be migrated The dynamic dependency graph of example is expressed as DDG=(N_DDG, E_DDG), N_DDGRepresent the active junction point collection of dynamic dependency graph, by event All unduplicated movable compositions, E in daily record_DDGRepresent the dependence edge collection between dynamic dependency graph node, by control dependence collection, Asynchronous call Dependency Set and data dependence relation collection composition；The realization of this step 1 comprises the following steps:

1.1 initialize set ReachingDefs (P₁, σ) and set Uses beSet Uses represents the variables collection used, Set ReachingDefs (P₁, σ) and it is all defined variable set dynamically terminating place up to definition set, i.e. execution route σ, its Middle P₁For the Services Composition before developing, before traversal evolution, all active junction points in the execution route of example to be migrated, add to dynamic The active junction point collection N of state dependency graph_DDGIn, analyzing can control dependence collection S between all active junction points_CDAnd asynchronous call Dependence collection S_AD；

1.2 traversal active junction point collection N_DDGIf, movable σ_iWith movable σ_jBetween exist control rely on or asynchronous call rely on, namely: Movable σ_iWith movable σ_jBetween binary crelation belong to control dependence collection S_CDWith asynchronous call dependence collection S_AD, then by phase The control dependence edge answered or asynchronous call dependence edge add the dependence edge set E of dynamic dependency graph to_DDGIn；

1.3 use dataflow analysis method, are added by all true data dependence edges in dynamic dependency graph DDG, specifically, and traversal Movable σ in execution route σ_jAll input variable var ∈ inPut (σ_j), if movable σ_jEmploy movable σ_kThe variable of definition, I.e. movable σ_kVariable belong to described set ReachingDefs (P₁, σ), then will be with movable σ_jFor starting point, with movable σ_kFor terminal True data dependence edge add in dynamic dependency graph DDG, simultaneously by activity σ_jInput variable add described set Uses to In；

1.4 use dataflow analysis method, join in dynamic dependency graph DDG by all anti-data dependence limits, specifically, time Go through movable σ in execution route σ_jAll output variable var ∈ outPut (σ_j), if movable σ_jDefine movable σ_kThe change used Amount, i.e. movable σ_kVariable belong to described set Uses, then will be with movable σ_kFor starting point, with movable σ_jAnti-data for terminal Dependence edge adds in dynamic dependency graph DDG, simultaneously by activity σ_jOutput variable add movable σ to_jDefined all changes Set Def (the σ of amount_jIn)；

1.5 use dataflow analysis method, join in dynamic dependency graph DDG by all output data dependence limits, specifically, Movable σ in traversal execution route σ_jAll output variable var ∈ outPut (σ_j), if this output variable is by movable σ_iDefinition, I.e. movable σ_iVariable belong to described set ReachingDefs (P₁, σ), and movable σ_jRedefined this output variable, then will be with Movable σ_iFor starting point, with movable σ_jOutput data dependence limit for terminal is added in DDG, simultaneously by activity σ_jOutput become Amount adds described set Def (σ to_jIn), movable σ_iDefined variable add to set Kill (σ_j), this set Kill (σ_j) it is to be migrated Example along execution route σ perform time, all arrival activity σ_jAnd by movable σ_jThe variable-definition being capped after redefining is constituted Set；

All elements in described set Def (σ j) is added to described set ReachingDefs (P by 1.6₁, σ) in, by described Kill (σ_j) In all elements from described set ReachingDefs (P₁, σ) and middle deletion；

1.7 repeat the above steps 1.2 to 1.6, until all activities in complete execution route σ of traversal；And

1.8 establish the activity of dynamic dependency graph DDG according to active junction points all in execution route and the dependence between them Nodal set N_DDGWith dependence edge collection E_DDG。

Further, in described step 2, the dynamic dependency graph after yojan is expressed as PDDG=(N_PDDG,E_PDDG), N_PDDGRepresent The active junction point collection of dynamic dependency graph, E_PDDGRepresent the dependence edge collection between node；The realization of this step 2 comprises the following steps:

2.1 by Services Composition P before described evolution₁Project on all variablees under current state Services Composition after evolution, it is thus achieved that Projection ReachingDefs (the P of all variablees₁,σ)∩V(P₂), P₂For the Services Composition after developing, V (P₂) represent P₂In all Movable；

2.2 traversal projection ReachingDefs (P₁,σ)∩V(P₂All variablees in), i.e. find out in described dynamic dependency graph DDG The active junction point of defined variable；

2.3 active junction points obtained based on step 2.2, take Dynamic Slicing method backward to obtain directly or indirectly variation value Movable section S；And

2.4 deletion dynamic dependency graph DDG are not belonging to the activity of the most movable section S and dependence edge corresponding with this activity, Thus obtain dynamic dependency graph PDDG after yojan.

Further, the realization of described step 2.4 comprises the following steps:

2.4.1 active junction point collection N is initialized_PDDGWith dependence edge collection E_PDDGFor sky, traversal projection ReachingDefs (P₁,σ)∩V(P₂) In all variablees, if there is variable by movable σ_iDefinition, and movable σ_iThe not active junction point collection of dynamic dependency graph after yojan N_PDDGIn, then by this activity σ_iAdd active junction point collection N to_PDDGNeutralize in a storehouse traversalQueue；

If 2.4.2 storehouse traversalQueue is not empty, then in traversalQueue first activity is popped, if with This activity is that the dependence edge of terminal belongs to dependence edge collection E_DDGBut it is not belonging to dependence edge collection E_PDDG, add this dependence edge to dependence Limit collection E_PDDGIn；If the starting point of this dependence edge is not at active junction point collection N_PDDGIn, add this node to active junction point collection N_PDDGIn, and add in storehouse traversalQueue；

2.4.3 repeat the above steps 2.4.1-2.4.2, until having traveled through projection ReachingDefs (P₁,σ)∩V(P₂All variablees in)； And

2.4.4 the active junction point collection N obtained based on step 2.4.1 to 2.4.4_PDDGWith dependence edge collection E_PDDG, after building yojan Dynamic dependency graph PDDG.

Further, in described step 4, migrate the concrete of validity and judge to comprise the following steps:

The active junction point collection N of dynamic dependency graph PDDG after 4.1 traversal yojan_PDDGIn all activities, if after yojan The dependence edge collection E of dynamic dependency graph PDDG_PDDGIn, often there is dependence edge with node n as terminal, then node n In-degree indegree [n] adds 1, and wherein this indegree [n] represents the in-degree of node n, and initial value is 0；

If 4.2 node n in-degrees are 0, then add in set zeroIndegreeNodeSet, this set ZeroIndegreeNodeSet represents the node set that in-degree is 0, and initial value is

4.3 repeat above-mentioned 4.1 and 4.2, until having traveled through the active junction point collection N of dynamic dependency graph PDDG_PDDGIn all activities；

If 4.4 set zeroIndegreeNodeSet are notDetect whether to there is an in-degree indegree [n_i] be 0 activity n_iMeet (P₂, S_C)[n_i>(P₂,S_C'), i.e. can be at the Services Composition P after evolution₂Current state (P₂, S_COccur under), if Do not exist, then the migration validity of Services Composition example is unsatisfactory for, wherein (P₂,S_C)[n_i>(P₂,S_C') represent n_iCan be in process P2 S_CPerform and along with n under state_iExecution P₂Can be by state S_CEnter new state S_C’；

4.5 if there is movable n_i, then by node n_iDelete from set zeroIndegreeNodeSet and add topological sorting ρ to In, update P simultaneously₂Current state；

The dependence edge collection E of 4.6 traversal dynamic dependency graph PDDG_PDDG, by all and node n_iAdjacent node in-degree subtracts 1；

4.7 repeat above-mentioned 4.4 to 4.6, until set zeroIndegreeNodeSet isNow if there is an active sequences Topological sorting ρ can be in Services Composition process P after evolution₂Middle playback, then the migration validity of Services Composition example meets.

From the above technical solution of the present invention shows that, the method that the present invention proposes does not requires the executed active sequences of example to be migrated Reappear according under the Services Composition definition after evolution of both definite sequences, thus compared to traditional track replay method, institute of the present invention Extracting method is the most flexible, is avoided that the tradition more issuable false negatives of track replay method (false negative), thus can To allow more example migrate, preferably support the dynamic evolution of Services Composition.

By by way of example embodiments of the present invention being described in detail below in conjunction with accompanying drawing, other features of the present invention, Feature and advantage will become apparent from.

Accompanying drawing explanation

Fig. 1 is the principle explanatory diagram of Services Composition instance migration.

Fig. 2 is the overall reality that Services Composition instance migration availability deciding method based on executed path dynamic dependency graph is exemplary Existing flow chart.

Fig. 3 is the flow chart migrating availability deciding in Fig. 2 exemplary process diagram.

Fig. 4 is an example of dynamic dependency graph DDG of the front Services Composition execution route of evolution.

Fig. 5 is an example of dynamic dependency graph PDDG obtained after the dynamic dependency graph to Fig. 4 carries out yojan.

Fig. 6 a-6b be develop after the example of controlling stream graph of Services Composition.

Fig. 7 a-7b is an example of the controlling stream graph of the front Services Composition of evolution.

Detailed description of the invention

As shown in Figure 1 and Figure 2, according to the preferred embodiment of the present invention, Services Composition instance migration based on dynamic dependency graph has Effect sex determination method, with the Services Composition before and after developing and the executed path (i.e. executed active sequences) of example to be migrated For inputting, with the migration availability deciding result of example to be migrated for output, as in figure 2 it is shown, its concrete implementation includes following Step:

Step 1, analyze the dependence between activity in the executed path of example to be migrated, automatically generate this executed path Dynamic dependency graph, the dynamic dependency graph of example the most to be migrated, and obtain and arrive executed path and terminate all fixed up to variable of place The set of justice (reaching definitions of variables)；

Step 2, yojan dynamic dependency graph: by the method for Dynamic Slicing backward, delete and do not have influential activity on instance migration And corresponding dependence edge；

Step 3, to develop after Services Composition use control flow analysis method obtain controlling stream graph；And

Dynamic dependency graph after step 4, controlling stream graph based on the Services Composition after developing and example yojan to be migrated, it is judged that about Whether there is a topological sorting (i.e. active sequences) under dynamic dependency graph after letter to reappear under the Services Composition after evolution, I.e. judge whether topological sorting be develop after an active sequences of Services Composition, if it is present migrate is effective Property is met, otherwise is unsatisfactory for.

Below with reference to Fig. 2 and combine shown in Fig. 3-6, describe the present embodiment in detail realizes process.

For the sake of for convenience of description, as the Services Composition inputted with BPEL (Business Process Execution in the present embodiment Language), as a example by process, system, OWL-S (Ontology Web Language for Services) certainly it are not limited thereto The Services Composition of other data perception such as process is also applied for the present embodiment.

For meeting the BPEL process of WS-BPEL 2.0 specification, after the present embodiment can judge the evolution given accurately and effectively Whether Services Composition example can replace the lower example that generate, that have not carried out end of process definition before developing, thus moves to mesh Mark state, to continue to provide current service.

As in figure 2 it is shown, the Services Composition instance migration validity based on executed path dynamic dependency graph that the present embodiment provides is sentenced Determining method, its concrete implementation comprises the following steps:

Step 1, analyze dependence R between activity in the executed path σ of example to be migrated, automatically generate executed path Dynamic dependency graph DDG, the dynamic dependency graph of example the most to be migrated, and obtain arrive executed path σ terminate place all up to Set ReachingDefs (the P of variable-definition (reaching definitions of variables)₁,σ)。

Wherein: the executed path representation of example to be migrated is σ={ σ₁,σ₂... }, σ₁, σ₂... expression activity, movable σ_iInstitute Having input variable is inPut (σ i), and all output variables are outPut (σ_i)；Dependence between activity is expressed as R=(σ_i,σ_j), σ_i And σ_jRepresenting the activity in aforementioned execution route, it comprises control dependence, asynchronous call dependence and data dependence relation, And represented in dynamic dependency graph DDG by the form of directed edge；The dynamic dependency graph of example to be migrated is expressed as DDG =(N_DDG, E_DDG), N_DDGRepresent the active junction point collection of dynamic dependency graph, be made up of unduplicated activities all in event log, E_DDGRepresent the dependence edge collection between dynamic dependency graph node, closed by controlling dependence collection, asynchronous call Dependency Set and data dependence Assembly forms.

As preferably embodiment, this step implements and comprises the following steps:

1.1 initialize set ReachingDefs (P₁, σ) and set Uses beSet Uses represents the variables collection used, Set ReachingDefs (P₁, σ) and it is all defined variable set dynamically terminating place up to definition set, i.e. execution route σ, its Middle P₁For the Services Composition before developing, before traversal evolution, all active junction points in the execution route of example to be migrated, add to dynamic The active junction point collection N of state dependency graph_DDGIn, analyzing can control dependence collection S between all active junction points_CDAnd asynchronous call Dependence collection S_AD；

1.2 traversal active junction point collection N_DDGIf, movable σ_iWith movable σ_jBetween exist control rely on or asynchronous call rely on, namely: Movable σ_iWith movable σ_jBetween binary crelation belong to control dependence collection S_CDWith asynchronous call dependence collection S_AD, then by phase The control dependence edge answered or asynchronous call dependence edge add the dependence edge set E of dynamic dependency graph to_DDGIn；

1.3 use dataflow analysis method, are added by all true data dependence edges in dynamic dependency graph DDG, specifically, and traversal Movable σ in execution route σ_jAll input variable var ∈ inPut (σ_j), if movable σ_jEmploy movable σ_kThe variable of definition, I.e. movable σ_kVariable belong to described set ReachingDefs (P₁, σ), then will be with movable σ_jFor starting point, with movable σ_kFor terminal True data dependence edge add in dynamic dependency graph DDG, simultaneously by activity σ_jInput variable add described set Uses to In；

1.4 use dataflow analysis method, join in dynamic dependency graph DDG by all anti-data dependence limits, specifically, time Go through movable σ in execution route σ_jAll output variable var ∈ outPut (σ_j), if movable σ_jDefine movable σ_kThe change used Amount, i.e. movable σ_kVariable belong to described set Uses, then will be with movable σ_kFor starting point, with movable σ_jAnti-data for terminal Dependence edge adds in dynamic dependency graph DDG, simultaneously by activity σ_jOutput variable add movable σ to_jDefined all changes Set Def (the σ of amount_jIn)；

1.5 use dataflow analysis method, join in dynamic dependency graph DDG by all output data dependence limits, specifically, Movable σ in traversal execution route σ_jAll output variable var ∈ outPut (σ_j), if this output variable is by movable σ_iDefinition, I.e. movable σ_iVariable belong to described set ReachingDefs (P₁, σ), and movable σ_jRedefined this output variable, then will be with Movable σ_iFor starting point, with movable σ_jOutput data dependence limit for terminal is added in DDG, simultaneously by activity σ_jOutput become Amount adds described set Def (σ to_jIn), movable σ_iDefined variable add to set Kill (σ_j), this set Kill (σ_j) it is to be migrated Example along execution route σ perform time, all arrival activity σ_jAnd by movable σ_jThe variable-definition being capped after redefining is constituted Set；

All elements in described set Def (σ j) is added to described set ReachingDefs (P by 1.6₁, σ) in, by described Kill (σ_j) In all elements from described set ReachingDefs (P₁, σ) and middle deletion；

1.7 repeat the above steps 1.2 to 1.6, until all activities in complete execution route σ of traversal；And

1.8 establish the activity of dynamic dependency graph DDG according to active junction points all in execution route and the dependence between them Nodal set N_DDGWith dependence edge collection E_DDG。

In the present embodiment, the above-mentioned step that implements can use following algorithm 1 to realize, certainly, and the ordinary skill of this area Personnel can realize in the way of using other, the most only one one exemplary embodiment.

Algorithm 1 builds dynamic dependency graph:

1.begin

2.Defs←Ф

3.Uses←Ф

4.for j←1to|σ|

5.N←N∪{σ[j]}

6.for everyσ[i]∈N

7.if<σ[i],σ[j]>∈SCD or SAD

8.E←E∪{<σ[i],σ[j]>}

9.endif

10.endfor

11.for every var∈inPut(σ[j])

12.E←E∪{<σ[k],σ[j]>|varσ[k]∈Defs}

13.Uses←Uses∪{varσ[j]}

14.endfor

15.Def(σ[j])←Ф

16.Kill(σ[j])←Ф

17.for every var∈outPut(σ[j])

18.E←E∪{<σ[k],σ[j]>|varσ[k]∈Uses}

19.E←E∪{<σ[i],σ[j]>|varσ[i]∈Defs}

20.Def(σ[j])←Def(σ[j])∪{varσ[j]}

21.Kill(σ[j])←Kill(σ[j])∪{varσ[i]|varσ[i]∈Defs}

22.endfor

23.Defs←(Defs–Kill(σ[j]))∪Def(σ[j])

24.endfor

25.end

With reference to shown in Fig. 2, next enter step 2 and operate, yojan dynamic dependency graph DDG.

Step 2, the method passing through Dynamic Slicing backward, delete and do not have influential activity and corresponding dependence edge thereof on instance migration, Wherein, the dynamic dependency graph after yojan is expressed as PDDG=(N_PDDG,E_PDDG), N_PDDGRepresent the active junction point of dynamic dependency graph Collection, E_PDDGRepresent the dependence edge collection between node.As preferably embodiment, it implements and comprises the following steps:

2.1 by Services Composition P₁Project on all variablees under current state Services Composition i.e. BPEL process model after evolution, Obtain the projection ReachingDefs (P of all variablees₁,σ)∩V(P₂), P₂For the Services Composition after developing, V (P₂) represent P₂In All activities；

2.2 traversal projection ReachingDefs (P₁,σ)∩V(P₂All variablees in), it is the most fixed i.e. to find out in dynamic dependency graph DDG The active junction point of justice variable；

2.3 active junction points obtained based on step 2.2, take Dynamic Slicing method backward to obtain directly or indirectly variation value Movable section S；

2.4 deletion dynamic dependency graph DDG are not belonging to the activity of the most movable section S (because of they shapes current to migration example State is to V (P₂) projection do not affect) and dependence edge corresponding with this activity, thus obtain the dynamic dependency graph after yojan PDDG。

Wherein, the realization of step 2.4 comprises the steps:

2.4.1 active junction point collection N is initialized_PDDGWith dependence edge collection E_PDDGFor sky, traversal projection ReachingDefs (P₁,σ)∩V(P₂) In all variablees, if there is variable by movable σ_iDefinition, and movable σ_iThe not active junction point collection of dynamic dependency graph after yojan N_PDDGIn, then by this activity σ_iAdd active junction point collection N to_PDDGNeutralize in a storehouse traversalQueue；

If 2.4.2 storehouse traversalQueue is not empty, then in traversalQueue first activity is popped, if with This activity is that the dependence edge of terminal belongs to dependence edge collection E_DDGBut it is not belonging to dependence edge collection E_PDDG, add this dependence edge to dependence Limit collection E_PDDGIn；If the starting point of this dependence edge is not at active junction point collection N_PDDGIn, add this node to active junction point collection N_PDDGIn, and add in storehouse traversalQueue；

2.4.3 repeat the above steps 2.4.1-2.4.2, until having traveled through projection ReachingDefs (P₁,σ)∩V(P₂All variablees in)；

2.4.4 the active junction point collection N obtained based on step 2.4.1 to 2.4.4_PDDGWith dependence edge collection E_PDDG, after building yojan Dynamic dependency graph PDDG.

In the present embodiment, the above-mentioned step that implements can use following algorithm 2 to realize, certainly, and the ordinary skill of this area Personnel can realize in the way of using other, the most only one one exemplary embodiment.

Algorithm 2 isolates the dynamic dependency graph after yojan:

1.begin

2.N'←Ф

3.E'←Ф

4.Defs←Defs∩V(P2)

5.for each varσ[i]∈Defs

6.N'←N'∪{σ[i]}

7.traversalQueue←{σ[i]}

8.while traversalQueue≠Ф

9.σ[k]←traversalQueue.dequeue()

10.for each<σ[j],σ[k]>∈E

11.if<σ[j],σ[k]>E'and<σ[j],σ[k]>

is not an anti-or output dependence edge

12.E'←E'∪{<σ[j],σ[k]>}

13.endif

14.ifσ[j]N'

15.N'←N'∪{σ[j]}

16.traversalQueue.enqueue(σ[j])

17.endif

18.endfor

19.endwhile

20.endfor

21.PDDG(N',E')←DDG(N,E)↑N'

22.end

As in figure 2 it is shown, it follows that enter:

Step 3: use control flow analysis method to obtain controlling stream graph the Services Composition after developing.

In the present embodiment, the controlling stream graph of the Services Composition after using the control flow analysis method of routine to be developed.Such as, " a kind of program control flowchart Layout Algorithm and the realization thereof " that Yuan Yuan, Jin Maozhong et al. propose；Constructing Control Flow Graph for Java by Decoupling Exception Flow from Normal Flow,Jang-Wu Jo,Byeong-Mo Chang in Computational Science and Its Applications ICCSA 2004 (2004), Volume 3043,2004, pp 106-113；Generation of Control and Data Flow Graphs from Scheduled and Pipelined Assembly Code, David C.Zaretsky, Gaurav Mittal ... in Languages and Compilers for Parallel Computing, Volume 4339,2006, pp 76-90 (2006), be not limited thereto system certainly.

As preferably embodiment, in the present embodiment, the acquisition of controlling stream graph comprises the following steps:

3.1 travel through given BPEL process according to the mode of depth-first, obtain all of active junction point, and to BPEL mistake The root activity of journey is set as that territory<scope>is movable；

3.2 according to the type of acquired activity, and select that the analytical function corresponding with this Activity Type analyze during BPEL is every The attribute of individual active junction point, extracts activity attributes common for all activities and is stored in internal memory, then according to Activity Type, Store attribute specific to each type of activity separately.

Analytical function can be divided into following 4 classes according to Activity Type:

A) attribute of the basic activity that the analytical function of basic activity reads during BPEL, movable attribute mainly comprises name Title, type, the follow-up relation of forerunner, input/output variable, partner's link and port type, in block structure, only in order Just there is the follow-up relation of forerunner between activity under<sequence>is movable, and block structure is by structured activities and its subactivity group Become, there is longitudinal stereovision；

B) for structured activities<if>,<switch>,<pick>,<while>with<repeatUtil>, their input variable is only Can obtain out from conditional statement<condition>, but these conditional statements must are fulfilled for bpel:getVariableProperty (", ' ') or bpws:getVariableData (", ") both forms can be resolved, and output variable in structured activities is all Empty null；

C) select the analytical function movable with loop structure to read the attribute of this type activity during BPEL, wherein input change Amount parses from condition<condition>, and, read the branch under respective conditions<conditon>movable；

D) analytical function of concurrent structure and sequential organization activity reads the attribute of this type activity during BPEL, and reads Take the activity under its Mei Tiao branch.

Operate as in figure 2 it is shown, next enter step 4, the migration validity of respective service example combinations is judged.

Step 4, controlling stream graph based on the Services Composition after dynamic dependency graph PDDG after yojan and evolution, it is judged that this is dynamic Whether there is a topological sorting ρ (i.e. active sequences) under dependency graph PDDG can reappear under the Services Composition after evolution, i.e. Judge whether that a topological sorting ρ is the Services Composition P after developing₂An active sequences, if it is present migrate Validity is met, otherwise is unsatisfactory for.

With reference to shown in Fig. 3-Fig. 4, as one preferably embodiment, migrate implementing of availability deciding and include following step Rapid:

The active junction point collection N of dynamic dependency graph PDDG after 4.1 traversal yojan_PDDGIn all activities, if after yojan The dependence edge collection E of dynamic dependency graph PDDG_PDDGIn, often there is dependence edge with node n as terminal, then node n In-degree indegree [n] adds 1, and wherein this indegree [n] represents the in-degree of node n, and initial value is 0；

If 4.2 node n in-degrees are 0, then add in set zeroIndegreeNodeSet, this set ZeroIndegreeNodeSet represents the node set that in-degree is 0, and initial value is

4.3 repeat above-mentioned 4.1 and 4.2, until having traveled through the active junction point collection N of dynamic dependency graph PDDG_PDDGIn all activities；

If 4.4 set zeroIndegreeNodeSet are notDetect whether to there is an in-degree indegree [n_i] be 0 activity n_iMeet (P₂, S_C)[n_i>(P₂,S_C'), i.e. can be at the Services Composition P after evolution₂Current state (P₂, S_COccur under), if Do not exist, then the migration validity of Services Composition example is unsatisfactory for, wherein (P₂,S_C)[n_i>(P₂,S_C') represent n_iCan be in process P2 S_CPerform and along with n under state_iExecution P₂Can be by state S_CEnter new state S_C’；

4.5 if there is movable n_i, then by node n_iDelete from set zeroIndegreeNodeSet and add topological sorting ρ to In, update P simultaneously₂Current state；

The dependence edge collection E of 4.6 traversal dynamic dependency graph PDDG_PDDG, by all and node n_iAdjacent node in-degree subtracts 1；

4.7 repeat above-mentioned 4.4 to 4.6, until set zeroIndegreeNodeSet isNow if there is an active sequences Topological sorting ρ can be in Services Composition process P after evolution₂Middle playback, then the migration validity of Services Composition example meets.

In the present embodiment, the above-mentioned step that implements can use following algorithm 3 to realize, certainly, and the ordinary skill of this area Personnel can realize in the way of using other, the most only one one exemplary embodiment.

Algorithm 3 migrates validation verification:

1.begin

2.zeroIndegreeNodeSet←Ф

3.(P2,Sc)←(P2,S0)

4.for each n∈N'

5.indegree[n]←0

6.for each<n',n>∈E

7.indegree[n]←indegree[n]+1

8.endfor

9.if indegree [n]=0

10.zeroIndegreeNodeSet←zeroIndegreeNodeSet∪{n}

11.endif

12.endfor

13.while zeroIndegreeNodeSet.isEmpty ()=false

14.ifni∈zeroIndegreeNodeSet,satisfying(P2,Sc)[ni>(P2,Sc')

15.zeroIndegreeNodeSet←zeroIndegreeNodeSet-{ni}

16.(P2,Sc)←(P2,Sc')

17.for each n∈N satisfying<ni,n>∈E

18.indegree[n]←indegree[n]-1

19.if indegree [n]=0

20.zeroIndegreeNodeSet←zeroIndegreeNodeSet∪{n}

21.endif

22.endfor

23.else

24.validity←false

25.return

26.endif

27.endwhile

28.validity←true

29.end

Shown in Fig. 4-7, using the BPEL process model of travel agent as Services Composition example, the present embodiment is carried Specific description is done in the migration availability deciding process gone out and realization thereof, and the BPEL process model of travel agent includes plane ticket booking, Hotel reservations etc. service, and this model is made up of 14 active junction points, including selecting, and the structure such as concurrently.

In order to judge that can example that generate under the definition of the Services Composition process before evolution, that have not carried out end safely and effectively Moving to continue executing with under the particular state definition of the Services Composition process after developing, the enforcement of its decision process includes:

1. analyze Services Composition example P to be migrated₁Execution route σ=< A_1.1,A_2.1,A_3.1,A_4.1,A_5.1,A_6.1,A_7.1,A_8.1,A_9.1, A_1.2,A_2.2,A_3.2,A_4.2,A_5.2,A_6.2,A_7.2,A_8.2,A_9.2,A_10.1,A_12.1,A_11.1> carry out dependence R, obtain dynamic dependency graph DDG, and acquire the set ReachingDefs (P up to variable-definition₁, σ) and={ queryInputA_1.2,airlineQueryA_2.2, hotelQueryA_2.2,airlineResultA_3.2,hotelResultA_4.2,queryOutputA_5.2,AcknowledgementA_8.2, airlineBookingA_10.1,hotelBookingA_10.1,hotelOfferA_12.1,airlineOfferA_11.1},A_i.jExpression activity A_iJth time Occurring, its detailed process is as follows:

1.1 initialize ReachingDefs and Uses isAll work of the execution route of the BPEL process instance before traversal evolution Dynamic node, adds the active junction point collection N of dynamic dependency graph to_DDGIn, obtain N_DDG={ A_1.1,A_2.1,A_3.1,A_4.1,A_5.1,A_6.1, A_7.1,A_8.1,A_9.1,A_1.2,A_2.2,A_3.2,A_4.2,A_5.2,A_6.2,A_7.2,A_8.2,A_9.2,A_10.1,A_12.1,A_11.1, analyze and obtain all movable knots Control dependence collection S between point_CD={ (A_9.1,A_1.2),(A_9.1,A_2.2),(A_9.1,A_3.2),(A_9.1,A_4.2),(A_9.1,A_5.2),(A_9.1, A_6.2),(A_9.1,A_7.2),(A_9.1,A_8.2),(A_9.2,A_10.1),(A_9.2,A_12.1),(A_9.2,A_11.1) and asynchronous call dependence collection S_AD= {(A_7.1,A_8.1),(A_7.2,A_8.2)}；

1.2 traversal active junction point collection N_DDGIf existing between the movable i of fruit and movable j and controlling to rely on or asynchronous call relies on, i.e. its two Unit's relation belongs to control dependence collection S_CDWith asynchronous call dependence collection S_AD, then dependence edge or asynchronous call will be controlled accordingly Dependence edge adds the dependence edge set E of dynamic dependency graph to_DDGIn；

1.3 use dataflow analysis method, by all true data dependence edge (A_1.1,A_2。1),(A_2。1,A_3。1),(A_3.1,A_5.1),(A_4.1,A_5.1), (A_5.1,A_6.1),(A_5.1,A_7.1),(A_7。1,A_8。1),(A_8.1,A_9.1),(A_2.1,A_1.2),(A_3.1,A_2.2),(A_4.1,A_2.2),(A_5.1,A_3.2),(A_5.1, A_4.2),(A_6.1,A_5.2),(A_7.1,A_5.2),(A_5.2,A_10.1),(A_1.2,A_2.2),(A_2.2,A_3.2),(A_2.2,A_4.2),(A_4.2,A_5.2),(A_5.2, A_6.2),(A_5.2,A_7.2),(A_7.2,A_8.2),(A_8.2,A_9.2),(A_10.1,A_12。1),(A_10.1,A_11.1) add in dynamic dependency graph DDG, it may be assumed that Movable A in traversal execution route σ_1.2All input variables because at execution activity A_1.1Time, by A_1.2Input variable QueryInput has added in defined variable set ReachingDefs, so movable A_1.2Employ movable A_1.1The change of definition Amount queryInput, then will be with movable A_1.1For starting point, movable A_1.2True data dependence edge for terminal adds in DDG, with Time queryInput is added in used variables collection Uses；

1.4 use dataflow analysis method, by anti-data dependence limit (A_9.1,A_8.2) add in dynamic dependency graph DDG, it may be assumed that traversal Movable A in execution route σ_8.2All output variables.Because at execution activity A_9.1Time, by A_8.2Output variable Acceptance has added use variables collection Uses to, so movable A_8.2Define movable A_9.1The variable used, then will be with Movable A_9.1For starting point, movable A_8.2Anti-data dependence limit for terminal is added in DDG, is added to by acceptance simultaneously Def(A_8.2In)；

1.5 use dataflow analysis method, by all output data dependence limit (A_1.1,A_1.2), (A_2.1,A_2.2), (A_3.1,A_3.2), (A_4.1, A_4.2), (A_5.1,A_5.2), (A_8.1,A_8..2) add in dynamic dependency graph DDG, it may be assumed that movable A in traversal execution route σ_8.2Institute There is output variable, because at execution activity A_8.1Time, by activity A_8.2Output variable acceptance added to defined Variables collection ReachingDefs, so movable A_8.2Redefine movable A_8.1The variable acceptance of definition, then will be with work Dynamic A_8.1For starting point, movable A_8.2Output data dependence limit for terminal is added in DDG, simultaneously by A_8.2The variable of output Acceptance adds Def (A to_8.2In), by A_8.1The variable acceptance of output add Kill (A to_8.2In)；

1.6 by Def (A_8.2Elements A in)_8.2The variable acceptance of output add defined set ReachingDefs to In, by Kill (A_8.2Elements A in)_8.1The variable acceptance of output add to from dynamically up to definition set ReachingDefs deletes, i.e. uses A_8.2The acceptance of definition overrides A_8.1The acceptance of definition；

1.7 repeat 1.2 to 1.6, until all activities in complete execution route σ of traversal；

1.8 establish the activity of dynamic dependency graph DDG according to active junction points all in execution route and the dependence between them Nodal set N_DDG={ A_1.1,A_2.1,A_3.1,A_4.1,A_5.1,A_6.1,A_7.1,A_8.1,A_9.1,A_1.2,A_2.2,A_3.2,A_4.2,A_5.2,A_6.2,A_7.2,A_8.2, A_9.2,A_10.1,A_12.1,A_11.1And dependence edge collection E_DDG={ (A_9.1,A_1.2),(A_9.1,A_2.2),(A_9.1,A_3.2),(A_9.1,A_4.2),(A_9.1, A_5.2),(A_9.1,A_6.2),(A_9.1,A_7.2),(A_9.1,A_8.2),(A_9.2,A_10.1),(A_9.2,A_12.1),(A_9.2,A_11.1),(A_1.1,A_2。1),(A_2。1,A_{3。 1}),(A_3.1,A_5.1),(A_4.1,A_5.1),(A_5.1,A_6.1),(A_5.1,A_7.1),(A_7。1,A_8。1),(A_8.1,A_9.1),(A_1.1,A_1.2),(A_2.1,A_1.2),(A_2.1, A_2.2),(A_3.1,A_2.2),(A_3.1,A_3.2),(A_4.1,A_2.2),(A_4.1,A_4.2),(A_5.1,A_3.2),(A_5.1,A_4.2),(A_5.1,A_5.2),(A_6.1,A_5.2), (A_7.1,A_5.2),(A_8.1,A_8.2),(A_5.2,A_10.1),(A_1.2,A_2.2),(A_2.2,A_3.2),(A_2.2,A_4.2),(A_4.2,A_5.2),(A_5.2,A_6.2), (A_5.2,A_7.2),(A_7.2,A_8.2),(A_8.2,A_9.2),(A_10.1,A_12。1),(A_10.1,A_11.1), build dynamic dependency graph DDG, such as Fig. 4 Shown in, for dynamic dependency graph DDG of Services Composition execution route before developing.

2., by the method for Dynamic Slicing backward, delete the Services Composition P after developing₂Do not has influential movable A_6.1, A_6.2And Corresponding dependence edge (A_5.1,A_6.1), (A_5.2,A_6.2), (A_6.1,A_5.2)(A_9.1,A_6.2) obtain the dynamic dependency graph after yojan, its concrete mistake Journey is as follows:

2.1 by Services Composition example P₁Project on all variablees under current state BPEL process model after evolution, it is thus achieved that Projection ReachingDefs (P1, σ) ∩ V (P2)={ queryInputA of all variablees_1.2,airlineQueryA_2.2,hotelQueryA_2.2, airlineResultA_3.2,hotelResultA_4.2,queryOutputA_5.2,AcknowledgementA_8.2,airlineBookingA_10.1, hotelBookingA_10.1,hotelOfferA_12.1,airlineOfferA_11.1}；

2.2 traversal ReachingDefs (P1, σ) ∩ V (P2) all variablees, find out the activity of defined variable in dynamic dependency graph DDG Node A_1.2, A_2.2, A_3.2, A_4.2, A_5.2, A_8.2, A_10.1, A_11.1, A_12.1；

2.3 take Dynamic Slicing method backward to obtain the movable section S of direct or indirect variation value；

The 2.4 movable A deleting in dynamic dependency graph the S that is not belonging to arbitrarily to cut into slices_6.1, A_6.2And the limit (A relevant to this activity_5.1,A_6.1), (A_5.2,A_6.2),(A_6.1,A_5.2),(A_9.1,A_6.2), thus obtain dynamic dependency graph PDDG after yojan；

Wherein, step 2.4 is specific as follows:

2.4.1 active junction point collection N is initialized_PDDGWith dependence edge collection E_PDDGFor sky, traversal projection ReachingDefs (P1, σ) ∩ V (P2) In all variablees, variable queryInputA_1.2, the not active junction point collection N of the dynamic dependency graph after yojan_PDDGIn, then will live Dynamic A_1.2Add nodal set N to_PDDGIn and by activity A_1.2Add in a storehouse traversalQueue；

2.4.2 with A_1.2Limit (A for terminal_1.1,A_1.2),(A_2.1,A_1.2),(A_9.1,A_1.2) belong to dependence edge collection E_DDGBut it is not belonging to E_PDDG, add this dependence edge to E_PDDGIn；

2.4.3A_1.1, A_2.1, A_9.1Not at N_PDDGIn, by A_2.1, A_9.1Add N to_PDDGIn；

2.4.4 the analysis method of step 2.4.1-2.4.4 is repeated, until having traveled through projection ReachingDefs (P₁,σ)∩V(P₂Own in) Variable；

2.4.5 by the active junction point collection N obtained_DDG={ A_1.1,A_2.1,A_3.1,A_4.1,A_5.1,A_7.1,A_8.1,A_9.1,A_1.2,A_2.2,A_3.2, A_4.2,A_5.2,A_7.2,A_8.2,A_9.2,A_10.1,A_12.1,A_11.1And dependence edge collection E_DDG={ (A_9.1,A_1.2),(A_9.1,A_2.2),(A_9.1,A_3.2), (A_9.1,A_4.2),(A_9.1,A_5.2),(A_9.1,A_7.2),(A_9.1,A_8.2),(A_9.2,A_10.1),(A_9.2,A_12.1),(A_9.2,A_11.1),(A_1.1,A_2。1),(A_{2。 1},A_3。1),(A_3.1,A_5.1),(A_4.1,A_5.1),(A_5.1,A_7.1),(A_7。1,A_8。1),(A_8.1,A_9.1),(A_1.1,A_1.2),(A_2.1,A_1.2),(A_2.1,A_2.2), (A_3.1,A_2.2),(A_3.1,A_3.2),(A_4.1,A_2.2),(A_4.1,A_4.2),(A_5.1,A_3.2),(A_5.1,A_4.2),(A_5.1,A_5.2),(A_7.1,A_5.2), (A_8.1,A_8.2),(A_5.2,A_10.1),(A_1.2,A_2.2),(A_2.2,A_3.2),(A_2.2,A_4.2),(A_4.2,A_5.2),(A_5.2,A_7.2),(A_7.2,A_8.2), (A_8.2,A_9.2),(A_10.1,A_12。1),(A_10.1,A_11.1) build dynamic dependency graph PDDG after yojan, as it is shown in figure 5, for after yojan Dynamic dependency graph PDDG.

3. the Services Composition P after pair evolution₂Existing control flow analysis method is used to obtain its controlling stream graph CFG, such as Fig. 6 a-6b Shown in, for the example of the controlling stream graph of Services Composition after developing.For the sake of contrast, this example gives Services Composition before developing Controlling stream graph, as shown in Fig. 7 a-7b.

4. dynamic dependency graph PDDG after yojan judges, if there is a topological sorting ρ=< A_1.1,A_2.1,A_3.1,A_4.1, A_5.1,A_7.1,A_8.1,A_9.1,A_1.2,A_2.2,A_3.2,A_4.2,A_5.2,A_7.2,A_8.2,A_9.2,A_10.1,A_11.1,A_12.1> can be at P₂Controlling stream graph CFG is upper to be reappeared, then judge that respective service example combinations meets and migrate validity, and what it judged specifically comprises the following steps that

The dependence edge collection E of dynamic dependency graph PDDG after 4.1 traversal yojan_PDDG, if there is the limit with certain node as terminal, Then in-degree indegree of this node adds 1, finally obtains indegree [A_1.1]=0, indegree [A_2.1]=1, indegree [A_3.1]=1, indegree[A_4.1]=0, indegree [A_5.1]=2, indegree [A_7.1]=1, indegree [A_8.1]=1, indegree [A_9.1]=1, indegree[A_1.2]=3, indegree [A_2.2]=5, indegree [A_3.2]=4, indegree [A_4.2]=4, indegree [A_5.2]=5, indegree[A_7.2]=2, indegree [A_8.2]=4, indegree [A_9.2]=1, indegree [A_10.1]=2, indegree [A_11.1]=2, indegree[A_12.1]=2；

4.2A_1.1In-degree is 0, then add in set zeroIndegreeNodeSet；

4.3 repeat above-mentioned 4.1 and 4.2, until having traveled through the active junction point collection N of dynamic dependency graph PDDG_PDDGIn all activities；

4.4 one in-degree of existence are the movable A of 0_1.1Meet (P₂, S_C)[A_1.1>(P₂,S_C’)；

4.5 by node A_1.1Add in topological sorting ρ, delete from set zeroIndegreeNodeSet and add topology row to In sequence ρ, update P simultaneously₂Current state；

The dependence edge collection E of 4.6 traversal dynamic dependency graph PDDG_PDDG, by all and node n_iAdjacent node A_1.2,A_2.1In-degree Subtract 1, i.e. indegree [A_1.2]=0, indegree [A_2.1]=0；

4.7 repeat above-mentioned 4.4 to 4.6, until set zeroIndegreeNodeSet isNow if there is an active sequences Topological sorting ρ=< A_1.1,A_2.1,A_3.1,A_4.1,A_5.1,A_7.1,A_8.1,A_9.1,A_1.2,A_2.2,A_3.2,A_4.2,A_5.2,A_7.2,A_8.2,A_9.2,A_10.1, A_11.1,A_12.1> can be in BPEL process P after evolution₂Middle playback, can determine that the migration validity of this Services Composition example meets.

Although the present invention is disclosed above with preferred embodiment, so it is not limited to the present invention.The technical field of the invention Middle tool usually intellectual, without departing from the spirit and scope of the present invention, when being used for a variety of modifications and variations.Therefore, originally The protection domain of invention is when being as the criterion depending on those as defined in claim.

Claims (4)

1. a Services Composition instance migration availability deciding method based on dynamic dependency graph, it is characterised in that before and after developing Services Composition and example to be migrated executed path for input, with the migration availability deciding result of example to be migrated be Output, its realization comprises the following steps:

Step 1, analyze the dependence between activity in the executed path of example to be migrated, automatically generate the dynamic of executed path State dependency graph, the dynamic dependency graph of example the most to be migrated, and obtain arrive executed path terminate place all up to variable-definition Set；

Step 2, yojan dynamic dependency graph: by the method for Dynamic Slicing backward, delete and do not have influential activity on instance migration And corresponding dependence edge；

Step 3, to develop after Services Composition use control flow analysis method obtain controlling stream graph；And

Dynamic dependency graph after step 4, controlling stream graph based on the Services Composition after developing and example yojan to be migrated, it is judged that about Whether there is a topological sorting under dynamic dependency graph after letter to reappear under the Services Composition after evolution, i.e. judge whether to deposit An active sequences of Services Composition after a topological sorting is to develop, if it is present the validity migrated is met, Otherwise it is unsatisfactory for；

Wherein, in the implementing of described step 1, the executed path representation of example to be migrated is σ={ σ₁,σ₂... }, σ₁, σ₂... Expression activity, movable σ_iAll input variables be inPut (σ_i), all output variables are outPut (σ_i)；Dependence between activity is closed System is expressed as R=(σ_i,σ_j), σ_iAnd σ_jRepresenting the activity in aforementioned execution route, it comprises control dependence, asynchronous call Dependence and data dependence relation, and represented in dynamic dependency graph DDG by the form of directed edge；Example to be migrated Dynamic dependency graph is expressed as DDG=(N_DDG, E_DDG), N_DDGRepresent the active junction point collection of dynamic dependency graph, by event log All unduplicated movable compositions, E_DDGRepresent the dependence edge collection between dynamic dependency graph node, by controlling dependence collection, asynchronous Call Dependency Set and data dependence relation collection composition；The realization of this step 1 comprises the following steps:

1.1 initialize set ReachingDefs (P₁, σ) and set Uses beSet Uses represents the variables collection used, Set ReachingDefs (P₁, σ) and it is all defined variable set dynamically terminating place up to definition set, i.e. execution route σ, its Middle P₁For the Services Composition before developing, before traversal evolution, all active junction points in the execution route of example to be migrated, add to dynamic The active junction point collection N of state dependency graph_DDGIn, analyzing can control dependence collection S between all active junction points_CDAnd asynchronous call Dependence collection S_AD；

1.2 traversal active junction point collection N_DDGIf, movable σ_iWith movable σ_jBetween exist control rely on or asynchronous call rely on, namely: Movable σ_iWith movable σ_jBetween binary crelation belong to control dependence collection S_CDWith asynchronous call dependence collection S_AD, then by phase The control dependence edge answered or asynchronous call dependence edge add the dependence edge set E of dynamic dependency graph to_DDGIn；

1.3 use dataflow analysis method, are added by all true data dependence edges in dynamic dependency graph DDG, specifically, and traversal Movable σ in execution route σ_jAll input variable var ∈ inPut (σ_j), if movable σ_jEmploy movable σ_kThe variable of definition, I.e. movable σ_kVariable belong to described set ReachingDefs (P₁, σ), then will be with movable σ_jFor starting point, with movable σ_kFor terminal True data dependence edge add in dynamic dependency graph DDG, simultaneously by activity σ_jInput variable add described set Uses to In；

1.4 use dataflow analysis method, join in dynamic dependency graph DDG by all anti-data dependence limits, specifically, time Go through movable σ in execution route σ_jAll output variable var ∈ outPut (σ_j), if movable σ_jDefine movable σ_kThe change used Amount, i.e. movable σ_kVariable belong to described set Uses, then will be with movable σ_kFor starting point, with movable σ_jAnti-data for terminal Dependence edge adds in dynamic dependency graph DDG, simultaneously by activity σ_jOutput variable add movable σ to_jDefined all changes Set Def (the σ of amount_jIn)；

1.5 use dataflow analysis method, join in dynamic dependency graph DDG by all output data dependence limits, specifically, Movable σ in traversal execution route σ_jAll output variable var ∈ outPut (σ_j), if this output variable is by movable σ_iDefinition, I.e. movable σ_iVariable belong to described set ReachingDefs (P₁, σ), and movable σ_jRedefined this output variable, then will be with Movable σ_iFor starting point, with movable σ_jOutput data dependence limit for terminal is added in DDG, simultaneously by activity σ_jOutput become Amount adds described set Def (σ to_jIn), movable σ_iDefined variable add to set Kill (σ_j), this set Kill (σ_j) it is to be migrated Example along execution route σ perform time, all arrival activity σ_jAnd by movable σ_jThe variable-definition being capped after redefining is constituted Set；

All elements in described set Def (σ j) is added to described set ReachingDefs (P by 1.6₁, σ) in, by described Kill (σ_j) In all elements from described set ReachingDefs (P₁, σ) and middle deletion；

1.7 repeat the above steps 1.2 to 1.6, until all activities in complete execution route σ of traversal；And

1.8 establish the activity of dynamic dependency graph DDG according to active junction points all in execution route and the dependence between them Nodal set N_DDGWith dependence edge collection E_DDG。

Services Composition instance migration availability deciding method based on dynamic dependency graph the most according to claim 1, its feature Being, in described step 2, the dynamic dependency graph after yojan is expressed as PDDG=(N_PDDG,E_PDDG), N_PDDGRepresent and dynamically depend on Rely the active junction point collection of figure, E_PDDGRepresent the dependence edge collection between node；The realization of this step 2 comprises the following steps:

2.1 by Services Composition P before described evolution₁Project on all variablees under current state Services Composition after evolution, it is thus achieved that Projection ReachingDefs (the P of all variablees₁,σ)∩V(P₂), P₂For the Services Composition after developing, V (P₂) represent P₂In all Movable；

2.2 traversal projection ReachingDefs (P₁,σ)∩V(P₂All variablees in), i.e. find out in described dynamic dependency graph DDG The active junction point of defined variable；

2.3 active junction points obtained based on step 2.2, take Dynamic Slicing method backward to obtain directly or indirectly variation value Movable section S；And

2.4 deletion dynamic dependency graph DDG are not belonging to the activity of the most movable section S and dependence edge corresponding with this activity, Thus obtain dynamic dependency graph PDDG after yojan.

Services Composition instance migration availability deciding method based on dynamic dependency graph the most according to claim 2, its feature Being, the realization of described step 2.4 comprises the following steps:

2.4.1 active junction point collection N is initialized_PDDGWith dependence edge collection E_PDDGFor sky, traversal projection ReachingDefs (P₁,σ)∩V(P₂) In all variablees, if there is variable by movable σ_iDefinition, and movable σ_iThe not active junction point collection of dynamic dependency graph after yojan N_PDDGIn, then by this activity σ_iAdd active junction point collection N to_PDDGNeutralize in a storehouse traversalQueue；

If 2.4.2 storehouse traversalQueue is not empty, then in traversalQueue first activity is popped, if with This activity is that the dependence edge of terminal belongs to dependence edge collection E_DDGBut it is not belonging to dependence edge collection E_PDDG, add this dependence edge to dependence Limit collection E_PDDGIn；If the starting point of this dependence edge is not at active junction point collection N_PDDGIn, add this node to active junction point collection N_PDDGIn, and add in storehouse traversalQueue；

2.4.3 repeat the above steps 2.4.1-2.4.2, until having traveled through projection ReachingDefs (P₁,σ)∩V(P₂All variablees in)； And

2.4.4 the active junction point collection N obtained based on step 2.4.1 to 2.4.4_PDDGWith dependence edge collection E_PDDG, after building yojan Dynamic dependency graph PDDG.

Services Composition instance migration availability deciding method based on dynamic dependency graph the most according to claim 3, its feature It is, in described step 4, migrates the concrete of validity and judge to comprise the following steps:

The active junction point collection N of dynamic dependency graph PDDG after 4.1 traversal yojan_PDDGIn all activities, if after yojan The dependence edge collection E of dynamic dependency graph PDDG_PDDGIn, often there is dependence edge with node n as terminal, then node n In-degree indegree [n] adds 1, and wherein this indegree [n] represents the in-degree of node n, and initial value is 0；

If 4.2 node n in-degrees are 0, then add in set zeroIndegreeNodeSet, this set ZeroIndegreeNodeSet represents the node set that in-degree is 0, and initial value is

4.3 repeat above-mentioned 4.1 and 4.2, until having traveled through the active junction point collection N of dynamic dependency graph PDDG_PDDGIn all activities；

If 4.4 set zeroIndegreeNodeSet are notDetect whether to there is an in-degree indegree [n_i] be 0 activity n_iMeet (P₂, S_C)[n_i>(P₂,S_C'), i.e. can be at the Services Composition P after evolution₂Current state (P₂, S_COccur under), if Do not exist, then the migration validity of Services Composition example is unsatisfactory for, wherein (P₂,S_C)[n_i>(P₂,S_C') represent n_iCan be in process P2 S_CPerform and along with n under state_iExecution P₂Can be by state S_CEnter new state S_C’；

4.5 if there is movable n_i, then by node n_iDelete from set zeroIndegreeNodeSet and add topological sorting ρ to In, update P simultaneously₂Current state；

The dependence edge collection E of 4.6 traversal dynamic dependency graph PDDG_PDDG, by all and node n_iAdjacent node in-degree subtracts 1；

4.7 repeat above-mentioned 4.4 to 4.6, until set zeroIndegreeNodeSet isNow if there is an active sequences Topological sorting ρ can be in Services Composition process P after evolution₂Middle playback, then the migration validity of Services Composition example meets.