A Linear Logic Representation for BPEL Process Protocol *

Transcription

1 Applied Mathematics & Information Sciences An International Journal 5 (2) (2011), 25S-31S A Lear Logic Representation for BPEL Process Protocol * Jian Wu 1 and Lu J 2 1 Department of Computer Science and Technology, Zhejiang University, Hangzhou, Cha 2 Department of Computer Science and Technology, Zhejiang University, Hangzhou, Cha Address: jluzju@gmail.com Received June 22, 2010; Revised December 21, 2010 Busess Process Execution Language (BPEL) is a powerful tool for describg web services compositions. The protocol of a BPEL process dicates the order of messages, which it sends or receives messages, as well as the structure of the ternal logic. Although formal methods have been used to model BPEL, the purposes of these researches are to verify BPEL process and elimate ambiguity. Recent research advances dynamic adaptation and reconfiguration of service processes require a new formal method to express BPEL process with the ability of problem solvg. This paper presents a lear logic based representation for BPEL process. Our approach expresses both basic and structured activities BPEL. With the help of proof-searchg tools, our approach set up a formal foundation for (semi)automatically solvg more challengg issues of service computg. Keywords: Web Service, Service Adaptation, Lear Logic, BPEL. 1 Introduction Web services are software entities capable of exchangg XML based data. With well-defed terfaces (Web Service Description Language, WSDL) [1], web services can be quickly composed as service processes to complete more complex tasks. One of the well-received specifications for describg service composition is the Busess Process Execution Language (BPEL) [2]. BPEL is supported by most major software vendors and applied to various fields. BPEL is a complex programmg language for describg sophisticated busess processes. A busess process performs numerous actions to complete a busess transaction. The order which actions are executed is called busess protocol. The same idea of protocol has been embedded to the BPEL language. With structured activities and other trsic mechanisms, such as fault and exception handlg, a BPEL process is capable of describg and executg complex busess processes. * This research is partially supported by National Sci & Tech Research Program of Cha (No. 2009ZX ), the 863 Program (No. 2008AA01Z141), the National NSF of Cha (No , No , No ), Zhejiang Provcial NSF (No. Y106598), and Zhejiang Provcial Sci &Tech Program (No. 2008C03007).

2 Jian Wu and Lu J 26 Sce BPEL is not equipped with formal semantics, the protocol of a BPEL process is very difficult to be formally reasoned. Several formal methods [3] have been used to model BPEL processes. R. Lucchi extended pi-calculus to describe activities and advanced structures, such as compensation handler, BPEL [4]. N. Lohmann provided a Petri-net based model for BPEL and a correspondg tool, BPEL2oWFN [5]. These models are then used to elimate ambiguity and verify various properties. Recent advances service adaptation [6] require formal methods with ability to provide solutions to mismatch problems. Although existg approaches are still used, they are more often used as an termedial model to bridge different process defition languages and avoid ambiguity. The adaptation process is usually constructed by an algorithm proposed by the authors. Without sufficient test cases to prove their algorithms, the methods are less convcg. Lear logic (LL) [7] is a branch of logic system. In combation with proofsearchg tools, such as Coq [8] and llprover [9], lear logic is capable of providg solutions to particular problems (semi)automatically. In our previous work [10] we proposed a lear logic-based automatic method for process adaptation. In this paper we propose a formal representation of BPEL process. Both basic activities and structured activities are specified usg LL sequents. As a result, our approach is capable to capture fully the BPEL process behavior and to generatg a lear logic-based representation of process protocol, which the message orders and ternal structures are defed. The rest of this paper is organized as follows: Firstly we troduce some background formation on lear logic. Then we present the LL representation of BPEL basic activities and structured activities. After that an example is given. Fally we summarize our work and give an outlook to our future work. 2 Lear Logic Background Lear logic (LL) was troduced by Girard [7] to provide a logical way for copg with resources. The fundamental notion LL states that A is consumed while producg B. As a result, the number of formulae is aware to the logic, that is to say one copy of A is different than two or more copies of A. This unique feature has made LL popular among computer scientists. The LL grammar that we use this paper is presented as follows: A:: = A A A A A& A A A! A u: A 1. The simultaneous conjunction ؤA B, also called multiplicative connective, suggests the possession of both A and B at the same time. The disjunction إA B, also called external choice, suggests that either A or B is available. The alternative conjunction A&B, also called ternal choice, represents that either A or B is produced. The lear implication A Bstates that A is consumed while achievg B. The of course modality can be applied to resource A if A could replicate itself without any resource. The

3 A Lear Logic Representation for BPEL 27 label u:a means that A is labeled with u. The trivial goal 1 represents a goal that requires no resources to achieve. In this paper we use LL formulae to represent messages exchanged by different services and processes. Sce LL does not distguish data types, we use different names for formulae to represent different types of messages. In this way we are able to reduce the complexity of reasong process. 3 Lear Logic Model for BPEL Protocol A BPEL process implements its busess logic by performg activities. There are two major categories of activities the BPEL specification: basic activity and structured activity. Basic activities describe elemental steps of the process behavior, while structured activities encode control-flow logic. Basic activities BPEL are capable of exchangg messages, manipulatg data, controllg ternal state etc. They are defed as follows: Defition 1. A basic activity is defed as an 8-tuple BA=<N, I, O, P, E, F, CT, ST>. In the defition, N represents the name of this activity; I and O represent comg messages/put variable, outgog message/output variable; P, E and F represent precondition, effect and fault respectively; CT is the collection of control tokens, which functions as a control mechanism for the execution of each activity; ST, state transition, is the collection of LL sequents that describe the basic activity s behavior. There are three notions with this defition that are not origally from the BPEL specification: effect, precondition and control token. Precondition P and effect E can be used to represent ternal states of a BPEL process. Furthermore we use preconditions and effects to represent the <source> and <target> elements. They are standard elements basic activity to specify the source and the target of a lk, which is a synchronization mechanism for parallel processes. Under the same lk, the source activity must fish before the execution of the target activity. Thus it is possible for activities different processes to synchronize. In our model the source activity generates a formula as the effect while the target activity requires that particular formula as the precondition. Control tokens are used to control the execution of each activity. There are three kds of control tokens the defition of activity: CT, CT next, and CT f. CT represent the control token that activates the activity. CT next represents the control token that activates the next activity while CT f represents fault. By assigng one activity s CT next to the next activity s CT, we are able to embed sequence structure to the defition of activity. Furthermore, sce the proof search of lear logic is undetermed, the application of control tokens can reduce the cost of provg theorems. The logic sequent ST represents the behavior of a basic activity. Due to their different functions not every basic activity shares the same composition and LL representation. The voke activity is capable of callg web services. It is defed as CT P ( CT O)& CT and CT I ( E CT )& CT. temp f temp next f

4 Jian Wu and Lu J 28 The receive activity is used to provide services to partners through bound message activities. It is defed as CT P I ( E CT )& CT. out f The reply activity is used to send a response to a previously accepted request. It is defed as CT P ( O E CT )& CT. out f The assign activity is capable of manipulatg variables and data with the busess process. The ST of the assign activity is defed as CT P I ( O E)& CT. The throw/rethrow activity is used to signal ternal fault explicitly or to propagate existg faults. It is represented as CT P E & CT. The wait activity delays the process for a specific period of time or until a certa deadle is reached. It is defed as گ CT P E & CT f f. Δ represents the time consumed by wait. The empty activity represents activities that do nothg. It is defed as CT P E & CT. next The exit activity is used to termate the busess process stance. It is defed as CT P F. Structured activities describe the order which a collection of activities is executed. By the composition of activities, structured activities are able to express the control patterns, handle faults and external events, and coordate message exchanges between process stances volved a busess protocol. The defition of structured activities is described as follows: Defition 2. A structured activity is defed as an 8-tuple SA=<N, I, P, E, F, EA, CT, TT >, which N, I, P, E, F represent name, comg message, precondition, effect and fault respectively; EA={EA 1,EA 2,,EA n } is the collection of embedded activities. If only one activity is embedded, EA is used to represent the embedded activity; CT={CT, CT next } is the collection of control tokens for this structured activity; TT, token transition, is the collection of LL sequents that describe the activity s structure. The defition of structured activities is similar to that of basic activities. Token transitions, TT for short, are used to describe the control flow encoded to the structured activities. Structured activities clude: sequence, if, pick, flow, while, repeat Until, for Each. The token transitions for each structured activity are defed as follow: The sequence activity allows activities to be executed the lexical order which they are defed. With CT out and CT every basic activity it is possible to control the sequence of activities by assigng one activity s CT out as another activity s CT. f

5 A Lear Logic Representation for BPEL 29 The if activity offers the conditional branch structure that represents ternal choice. It is defed as CT EA 1.CT &EA 2.CT & &EA n.ct. The pick activity offers the ability to respond to external events. It is represented as CT (I 1 EA 1.CT ) (I 2 EA 2.CT ) (I n-1 EA n-1.ct ) EA n. CT. Each sequent I m E m.ct (0<m<n) corresponds to an <onmessage> element and depicts that the pick activity consumes an comg message and execute one activity EA m. The last EA n represents the activity the <onalarm> branch, if defed. The flow activity offers concurrency and synchronization. The TT of the flow activity consists of two parts: start and termation. The start is represented as CT E 1.CT E 2.CT E n.ct. The termation is represented as E 1.CT next E 2.CT E n.ct next CT next. The while activity provides the ability to execute a contaed activity repeatedly. The TT of the while activity also contas two sequents: CT E.CT &CT next and E.CT next CT. The repeatuntil activity also furnishes the function for repeated execution of a contaed activity. The TT of the repeatuntil activity is defed as CT E.CT and E.CT next E.CT &CT next. After both basic activities and structured activities are defed, the BPEL process can be quickly defed as a union of all activities. Defition 3. A BPEL process is defed as a 4-tuple P=<IV, CT, BA, SA>, where IV=<N, I, O, P, E, F> is the terface view of the service process, which consists of canonical name, comg and outgog messages, preconditions, effects, and faults. CT={CT 1, CT 2,, CT n } is the collection of control tokens; BA={A1, A 2,, A m } is the collection of basic activities; SA={S1, S 2,, S k } is the collection of structured activities. 4 Example Walkthrough flow1 Calculate Price voke2 -PO CT8 voke3 -shippginfo CT9 receive3 +Invoice CT10 S receive1 +PO CT1 CT7 CT3 flow1 (Start) Select Shipper assign1 -PO +shippgrequest CT11 CT4 lk1 voke1 -shippgrequest +shippginfo CT5 Production Schedule voke4 CT12 -PO receive2 +shippgschedule lk2 voke5 -shippgschedule CT6 CT13 flow1 (Termate) CT2 reply1 -Invoice F Fig. 1. Lear Logic Model for Purchase Order Process Figure 1 demonstrates the lear logic model for the purchase order process from the section 5.1 of BPEL 2.0 specification 2. Activities are represented usg rectangles. Each activity is labeled its message directions, + for comg message and - for outgog

6 Jian Wu and Lu J 30 message. After a purchase order from the client is received, the process itiates three parallel subprocesses: i) Calculate Price calculates the fal price of the order and receives the voice produced by a third party service; ii) Select Shipper arranges transportation and calculates the shippg price; iii) Production Schedule schedules the production and shipment for the order. Dotted arrows represent control lks used for synchronization across concurrent activities. Lk1 dicates the shippg price is required to falize the price calculation. Lk2 shows that the shippg date is required for the complete fulfilment schedule. After the three concurrent paths have completed their execution, an voice is returned to the customer. The process defition is listed as follows. P=<IV, CT, BA, SA> IV=< Purchase Order Process, POMesage, InvMessage, Null, Null, cannotcompleteorder> CT=<CT 1, CT 2, CT 13, S, Φ> BA=<voke1, voke2, voke3, voke4, voke5, receive1, receive2, receive3, reply1, assign1,> SA=<flow1> The detailed lear logic representation of each activity the purchase order process is listed Table 1. Note that itials are used for clear representation. IN stands for Invoice. SI stands for shippginfo. SR stands for shippgrequest. SS stands for shippgschedule. Activities I O P E F CT ST receive1 PO N/A N/A N/A N/A S,CT 1 S PO CT 1 flow1 N/A N/A N/A N/A N/A CT 1,2,3,6,7,11,13 CT 1 CT 3 CT 7 CT 11, CT 6 CT 10 CT 13 CT 2 reply1 N/A IN N/A N/A N/A CT 2, Φ CT 2 IN Φ voke2 N/A PO N/A N/A N/A CT 7,8 CT 7 PO CT 8 voke3 N/A SI lk1 N/A N/A CT 8,9 CT 8 lk1 SI CT 9 receive3 IN N/A N/A N/A N/A CT 9,10 CT 9 IN CT 10 assign1 PO SR N/A N/A N/A CT 3,4 CT 3 PO SR CT 4 voke1 SI SR N/A lk1 CCO CT 4,5, CT v1 CT 4 (SR CT v1 )&CT CCO, CT v1 SI lk1 CT 5 receive2 SS N/A N/A lk2 N/A CT 5,6 CT 5 SS lk2 CT 6 voke4 N/A PO N/A N/A N/A CT 11,12 CT 11 PO CT 12 voke5 N/A SS lk2 N/A N/A CT 12,13 CT 12 lk2 SS CT 13 4 CONCLUSION & FUTURE WORKS In this paper a lear logic-based formal model for BPEL process is proposed. The contributions of this work clude: i) Lear logic is corporated to our method. With the help of proof-searchg tool and proper problem setup, more challengg service computg problems, such as service adaptation, can be solved (semi)automatically. ii) A lear logic-based formal semantic for basic activities is proposed, which is capable of expressg message exchanges and the lk mechanism. iii) A lear logic-based semantic for structured activities is proposed. Structures, such as sequence, choice, parallel and loop, can be easily described. In our future work more mechanisms, such as fault handlg, event handlg and compensation, will be cluded. We are currently workg

7 A Lear Logic Representation for BPEL 31 on an automatic tool that can transform XML based BPEL description to LL sequents. References [1] Web Service Description Language (WSDL) Version 2.0, W3C Recommendation (2007). [2] Web Services Busess Process Execution Language (BPEL) Version 2.0, OASIS Standard (2007) [3] F. v. Breuge and M. Koshka, Models and Verification of BPEL, (2006). [4] R. Lucchi and M. Mazzara. A pi-calculus based semantics for WS-BPEL. Journal of Logic and Algebraic Programmg. 70 (2007) [5] Niels Lohmann. A Feature-Complete Petri Net Semantics for WS-BPEL 2.0, In Web Services and Formal Methods, Forth International Workshop, Milan, [6] M. Dumas, B. Benatallah and H.R.M. Nezhad, Web Service Protocols: Compatibility and Adaptation, IEEE Data Engeerg Bullet. 31 (2008) [7] J.-Y. Girard, Lear logic, Theoretical Computer Science, 50 (1987) [8] Coq, a formal proof management system. [9] llprover, A Lear Logic Prover, [10] L. J and J. Wu, A Lear Logic Based Approach for Generatg Deadlock Adapters, Proceedgs of 2010 Asia Pacific Service Computg Conference, Hangzhou, Cha. Jian Wu received his B.Sc and Ph.D degree from Zhejiang University 1998 and 2004, respectively. He is now an Associate Professor the Department of Computer Science and Technology. His research terests clude Service Computg, Data Mg, Grid Computg and Semantic Web. Lu J received his B.Sc from Zhejiang University He is now a Ph.D student the Department of Computer Science and Technology Zhejiang University. His research terests clude Service Computg and Semantic Web.

8 Applied Mathematics & Information Sciences An International 5 (2) (2011), 33S-37S