Proceedings of DET2007 4th International Conference on Digital Enterprise Technology

Bath, United Kingdom 19-21 September 2007

SERVICE-ORIENTED WORKFLOW: CONCEPTS, ARCHITECTURE AND KEY TECHNOLOGIES

Bo Liu Department of Automation, Tsinghua

University, Beijing 100084, P.R. China liubo03@mails.tsinghua.edu.cn

Yushun Fan Department of Automation, Tsinghua

University, Beijing 100084, P.R. China fanyus@tsinghua.edu.cn

ABSTRACT The advent of Grid and SOA (Service-Oriented Architecture) technologies has brought new opportunities and challenges into the area of workflow. Service-Oriented Workflow (SOWF), which integrates web service and workflow technology, has become a mainstream for flexible integration among heterogeneous systems. In this paper, firstly the definition of the service is proposed with a three-dimensional model and based on a detailed analysis of the characteristics of SOWF, the architecture of SOWF including the user interface layer, the operation logic layer, and the persistent storage layer is presented. As key technologies, process modeling, process execution, workflow implementation technologies, transaction management and performance evaluation are discussed respectively. Finally, an application demonstration of SOWF is given to prove the advantages of applying service-oriented technology in the field of workflow.

KEYWORDS Service, Service-Oriented Workflow, System Architecture, Process Modeling

1. INTRODUCTION Web services technology has become a new wave for Internet-based business applications (W3C, 2002). It has brought to the world a loosely coupled environment that enables flexible integration among heterogeneous systems. Under such circumstances, service and workflow are close related: workflow can be constituted with service, and workflow itself can be encapsulated into service as well. Thus it appears a new tendency that combines web service and workflow technology together. Because of the loosely coupled, autonomous and dynamic nature of service, the operation mechanism of workflow meets some new characteristics and difficulties.

Many researchers have carried out their studies in this area. For example, Zhou et al (2004) regarded every activity in workflow as service, automatically executed by the agent. Xiao et al (2004) proposed a conceptual model of Web services workflow. Zhao and Liu (2006) studied the modeling and implementation of organization-centered workflows in the Web service environment. IBM developed a

workflow management system “intelliFlow” on the basis of SOA for business process management. HP Laboratories proposed a service composition model and realized a prototype system (Casati and Shan, 2001). Grid workflow has also become a hotspot (Oinn et al, 2004).

As a result, a new workflow paradigm, Service-Oriented Workflow (SOWF), comes up into being. However, most above research merely concerns a specific part of service-oriented workflow, lacking the research on integral architecture and implementation technologies. Consequently, it is hard to unify existing research to a common foundation. Moreover, few of them proposed an explicit definition of service-oriented workflow, thus the understanding of which is confused.

In order to address these issues, we analyze the characteristics of SOWF and give out the concepts of service and service-oriented workflow. Then the architecture of SOWF is proposed, based on which the key technologies are discussed and summarized in detail.

539

The rest of the paper is organized as follows. With a definition of service in Section 2, we present the characteristics of SOWF in Section 3. The architecture and five key technologies of SOWF are described in Section 4 and 5. Section 6 introduces an application demonstration of SOWF. Finally, conclusions are given in Section 7.

2. DEFINITION OF SERVICE A new field called Service Science has become the focus in recent years, but there still lacks a uniform definition of service. In this paper, a service is defined as an IT-enabled or IT-innovated functionality involving certain business process or activity, which is offered by a provider. A three-dimensional service model is proposed as shown in Figure-1, which describes service from three views.

Figure 1 –Three-dimensional service model

1. Construction dimension consists of five parts. Description contains the name, identifier and domain of the service. Configuration includes related information used for configuring service, such as specific coherences. Inputs/Outputs describe the input and output parameters. Constraints mean pre- or post-conditions, rules, etc. QoS & Measurements provide the performance indicators of the service.

2. Lifecycle dimension involves four periods. During Design period, service providers define the basic construction of the service, especially Description part. Then, when the service is ready to release to the Network, its coherences need to be confirmed according to the user’s demand. After Deploy period, it comes to Execute period, at this time the service is realized. At last, in maintain period, the service is monitored and managed, being upgraded or withdrawn.

3. Genericity dimension encompasses three levels. Generic level comprises a collection of services that have the widest application in the

representation of service domains. Partial level contains sets of partial services, each one being applicable to a specific domain. Particular level is concerned solely with one particular service domain. Three levels of genericity are ordered, in the sense that Partial level is a specialization of Generic level, and Particular level is a specialization of Partial level.

In this model, service is constructed, realized and specialized gradually, and the nature of service is shown in the round.

3. WORKFLOW IN SERVICE-ORIENTED ENVIRONMENT Because of the nature of service such as loosely coupling, coarse granularity, access transparency, platform independency and business orientation, workflow in Service-Oriented Computing also presents many new characteristics:

1. Services are implemented by workflow. Workflow technology enables the resilient and dynamic composition of services.

2. Workflow is just another kind of services. Business process itself may be distributed among dispersed partners, and the involved applications as well as the whole workflow are delivered as services.

3. There are multiple processes running at the same time. They communicate with event/messages (No explicit control link among them) and share the resource or data. New processes/services are dynamically created or invoked.

4. The processes change dynamically along with the changes of services. Because of the autonomy of services, it requires ensuring the usability of services and selecting service components in real time during the operation of processes, which also results in the difficulty in evaluating workflow performance.

Based on the above characteristics, we proposed the definition of SOWF.

Service-oriented workflow (SOWF) is the business process partly or totally executed by the computers automatically in service-oriented environments, partial or entire activities in the business process are completed by services in Network. In other words, SOWF is a composition of web services for the purpose of special tasks.

4. ARCHITECTURE OF SOWF As shown in Figure-2, the architecture of service-oriented workflow management system is put forward. It is a hierarchical structure which contains three layers, i.e., the user interface layer, the operation logic layer, and the persistent storage

540

layer. The user interface layer includes interfaces for workflow model presentation, execution management and user interaction, while the persistent storage layer comprises databases (DB) for storing the workflow models and relevant data.

The core element of the operation logic layer is the workflow engine, with the participant allocator, the model interpreter, the ECA (Event- Condition-Action) rule parser and the time controller as supportive components. The engine is basically composed of a logic controller and an event handler,

with the former dealing with the navigation of control flows as well as data flows in a workflow model, and the latter handling particular events in service computing. The analysis engine deals with the data from persistent storage layer and supports the data analysis tools, the result of the analysis is fed back to business knowledge DB as rules or knowledge for future use. The interaction engine is a bridge between this workflow engine and other workflow engines by event communication and data correlation.

Figure 2 – Architecture of service-oriented workflow

In service-oriented environments, there are two kinds of activities in business processes, one is normal task executed like traditional activity, and the other is service which needs special mechanism. When the workflow engine encounters a service to handle, it turns to the enterprise service bus, in which the agent is responsible for the execution of the service through querying the service management server who stores the directory of all the service. Finally a simple service or composite service (constructed by composing several simple services according to certain regulation) in Network is selected to match the requirement of specific service. The quality of service is guaranteed by the QoS monitor modular.

In addition, in order to ensure the performance of entire SOWF system, a performance evaluation

model is built to support the operation of SOWF and improve the business performance.

5. KEY TECHNOLOGIES FOR SOWF There are a large number of challenges to be addressed in SOWF, and among these challenges, process modeling, process execution, workflow implementation technologies, transaction management and performance evaluation are most important. The five aspects follow the "buildtime - runtime - management - performance evaluation" approach in the research in workflow area. The following sections discuss and summarize related technologies across-the-board.

5.1. PROCESS MODELING

541

Different from traditional process modeling, service-oriented workflow model includes normal tasks and service nodes, so the main issue of process modeling is the mapping from business processes to service resources and the composition of services resources.

5.1.1. Service Process Modeling Method Existing process modeling methods comprise two types: formal method based on Petri-net and process algebra, informal method based on business process description language. Table-1 lists several typical methods.

Table 1- Typical process modelling methods

Category Language name Description Proposed

year Organizations

or persons

Informal method

WSFL (Web Service Flow Language)

An XML language for the description of Web Services compositions (OASIS, 2002). The composition comes in two types: The first type allows specifying the logic of a business process (called flow models); the second type allows defining the mutual exploitation of Web services of participants in a business process (called global models).

2001 IBM

XLANG

An extension of WSDL (Web Service Description Language) which provides both the model of an orchestration of services as well as collaboration contracts between orchestrations. Although some orchestration engines continue to support XLANG, the momentum has clearly shifted toward BPEL4WS.

2001 Microsoft

BPEL4WS

BPEL4WS (Business Process Execution Language for Web Service) enables the specification of executable business processes and business process protocols in terms of their execution logic or control flow (Ankolekar et al , 2004). It has integrated XLANG and WSFL.

2002 Microsoft, IBM, BEA and so on.

GSFL (Grid Service Flow Language)

An effort to examine technologies that address workflow for Web services and leverage this technology for Grid services. It contains both an XML schema definition for the specification of workflow in Grid environments and a reference implementation of the Workflow engine for use in the Open Grid Services Infrastructure (OGSI). (Laszewski, 2002)

2002 Argonne National

Laboratory

Formal method

Petri-net

The Web Service process model built by Petri-net represents service processes with graphics mode vividly, convenient for various persons to participant in constructing business processes. All the analysis methods of Petri-net could be used to validate the correctness and analyze the performance of web service process model.

1968 A.W. Holt

Process algebra

Process algebra normally uses Pi-calculus as formalization representation, describes complicated process model by composing simple structured process control modules, and verifies the correctness of models.

Late 1970s - early 1980s

Robin Milner, Charles Antony Richard Hoare

Totally speaking, BPEL4WS has become the standard of Web Service supported by OASIS (Organization for the Advancement of Structured Information Standards). It is considered as the most mature and complete business process modeling language.

5.1.2. Service Matching and Composition In SOWF, process is constituted with services or mapped to a service. Service matching is the basis

of process modeling, while service composition is the basis of workflow execution.

As far as service matching, there are normally two implementation methods: the method based on UDDI keyword mapping and the method based on artificial intelligence technology. The former one strictly follows the UDDI rules and well supports the discovery of services described by WSDL, but shows the deficiency of low flexibility, lack of measurement of mapping degree and weak support on semantic operation. Consequently, most

542

researchers concentrate on the latter one, i.e. the method based on artificial intelligence.

Wu et al (2005) proposed a Web Service discovery method based on ontology and semantic similarity degree. Zhang and Lin (2006) presented a fast finding service algorithm in the distributed network by genetic algorithm with the help of mobile agent. Chakraborty et al (2006) proposed a distributed service discovery protocol for pervasive environments on the basis of concepts of peer-to-peer caching of service advertisements and group-based intelligent forwarding of service requests.

As far as service composition, there are mainly two methods. One is based on workflow description language using XML, typically WSCL, BPEL4WS, etc. The other is based on semantic description of Web Service, such as OWL-S.

Tan and Fan (2005) proposed a semantic-based framework to support the matching, composition and operation of manufacturing services with the purpose of integrating various services in networked manufacturing environment. Zhou et al (2006) analyzed the current solutions for Web service composition and brought up a design of Web service composition platform (WSCP). In addition, referring to the composition of grid service, OGSA (Open Grid Services Architecture) integrated grid and web service and identified the framework of distributed systems, but didn’t define corresponding standards about grid service composition.

5.2. PROCESS EXECUTION Different from traditional workflow, the execution of service-oriented workflow concentrates on distributed workflow engine, the interaction of processes and the dynamic scheduling of resources. In fact, the execution of a service process is a procedure of optimization and instantiation for composite services.

5.2.1 Workflow Engine The workflow engine in service-oriented environments should be distributed, including several operation engines and one control engine. Each engine should support specific mechanisms for external and internal event interactions, which will inevitably involve such issues as data correlation as well as asynchronous communication between different workflows and the like.

To deal with application requirements on distributed workflow engine from enterprises and government, Xin and Wang (2005) designed a self-managed distributed workflow engine based on Java Management Extensions. They proposed the central engine which acted as the object and the executive engines which acted as the observers to observe the state changes of the central engine. The executive

engine registered with the central engine during the start time, and logout during the close time, therefore the self-managed workflow system was implemented.

Considering the situation that lots of users invoke grid workflow services in a very narrow time interval, Cao et al (2006) discussed the structure and mechanisms of grid workflow engine based on Jini platform. By adding or removing enactment services, the grid workflow engine can be reconfigured dynamically.

5.2.2 Process Interaction The interoperation and collaboration between different services are the main issues on process interaction.

WFMC has defined a standard “Wf-XML” for the exchange of workflow data. Kuechler et al (2001) described a new workflow interoperability model, the monitored-nested model (MNM), and showed that it can support optimized, extended e-commerce transactions that were not supported by current models. He et al (2005) presented a cross-enterprise collaborative manufacturing platform based on agent and workflow, in which a set of intelligent agents cooperated to perform manufacturing jobs and process modeling, dynamic task scheduling and dispatching, resource management as well as manufacturing process management on the basis of a shared domain ontology.

5.2.3 Dynamic Resource Scheduling The uncertainty of service provider determines the dynamic character of service process execution. The service is determinate while the performer of service is accidental. So how to select the most suitable service provider from a mass of available services, and how to choose the substitute when a service failed pose the problems of dynamic resource scheduling. According to recent research status, the scheduling of composite services still lack efficient algorithm to satisfy multiple performance indicators.

Yuan et al (2005) analyzed the QoS properties in Manufacturing Grid (MG) and presented a scheduling architecture for MG workflow. Workflow engine can dynamically schedule the resources according the activity requirements and resource QoS capabilities. If multiple resources are available to an activity at a time, workflow scheduling can select the 'best match' resource according to scheduling algorithms.

Hao et al (2006) proposed a component called Agent-based Web Service (AWS) to provide manufacturing scheduling services for enterprise collaboration. The scheduling process of an order is orchestrated on the Internet through negotiation among AWSs and the bid proposed by an AWS is

543

supported by the dynamic scheduling results of manufacturing resources inside the enterprise.

As a matter of fact, the ultimate difference between the scheduling in SOWF and that in traditional workflow lies in the fact that the service scheduling is across enterprises but the enterprises are selfish, so only when we take into account the benefit of both the service provider and service consumer can the service scheduling meaningful.

5.3. WORKFLOW IMPLEMENTATION TECHNOLOGIES The implementation of workflow systems normally utilizes existing distributed process technologies, including Web Service, Grid, agent and P2P.

5.3.1 Workflow Based On Web Service SOWF has inherent relation with Web Service (WS). WS itself could be composed and accessed through XML-based message protocol, thereby realizes the seamless integration of applications in heterogeneous environment. The advent of related technologies and standards for WS such as SOAP, WSDL and UDDI promote the fast development of workflow technology to service-oriented direction.

The above mentioned service process description methods, WSFL, XLANG and BPEL4WS, are all based on Web Service technology.

Existing applications of workflow systems based on WS are limited to the composition and integration of total systems, and seldom refer to modeling and engine design. You (2006) did some research on this area, presenting a three-layer modeling framework, improving the architecture of workflow engine, and promoting the flexibility and interoperability of workflow.

5.3.2 Workflow Based On Grid Service With the gradual matureness of grid technology, the composition of grid service and web service through workflow technology has attracted the enterprise’s attention. Typical framework includes GridFlow, GridAnt and Grid-WFS.

1. GridFlow (Cao et al, 2003) project is a large collaborative project to integrate a wide range of existing grid middleware. Existing systems for resource scheduling, resource management, and performance services were integrated with the Globus toolkit. Gridflow uses an XML specification to represent workflow and offers visual editing tools.

2. GridAnt (Laszewski et al, 2004) is an extensible client-side workflow management system developed by Argonne National Laboratory. Begun as GSFL in 2001, it allows for the construction of

client side workflow for Globus Toolkit and the specification of precondition and parallel tasks.

3. Hwang and Kesselman (2003) proposed a grid workflow system (grid-WFS), a flexible failure handling framework for the grid, which addressed grid-unique failure recovery requirements. The framework utilized workflow structure as a high-level recovery policy specification, and enabled users to rapidly prototype and investigate failure handling strategies.

5.3.3 Workflow Based On Agent As a promising tendency of SOWF, the integration of agent and WF technologies could take advantage of each merit respectively.

Nowadays, there exist two kinds of integration, agent-enhanced workflow and agent-based workflow (Yan et al, 2001).

The former is the addition of agent applications on primary workflow engine, for example, user interface agent, automatic task agent, interoperation agent, etc. When the task is started or executed, the agent is invoked to complete specific tasks, and the workflow engine controls the creation and elimination of the agent. In this mode, centralized workflow engine dominates all the tasks, so it fits to inner-enterprise workflow systems.

The latter is the distributed system constructed totally by multiple independent agents, each one executing a process segment. In this mode, entire business process is accomplished by these agents who provide a distributed architecture and satisfy the business requirements of large-scale enterprises and cross-organization enterprises. Accordingly the latter mode is more suitable to construct loosely-coupled cross-enterprise workflow.

5.4 TRANSACTION MANAGEMENT The high dynamic and heterogeneous nature of the Internet brings uncontrolled factors while integrating and invoking outside web services in workflow. So the failed execution in service-oriented workflow system occurs more often than that in the traditional workflow system. Thus transaction management is a key issue to ensure the successful execution of SOWF (Yan et al, 2005).

Yan et al (2005) considered service-based workflow process as a long running transaction which composed of several short transactions. They proposed a transaction management framework to maintain the integrity of workflow transactions. This framework made some improvements to the traditional workflow system including workflow definition tool, compiler tool and enactment tool.

Moreover, because the cross-enterprise operation often results in large number of long transactions,

544

the SOWF shall improve the ability to monitor performance and process transaction. The research on long transaction has not developed yet.

5.5 PERFORMANCE EVALUATION Workflow performance includes several aspects, such as time, cost, quality, reliability and agility. The evaluation of single performance indicator is obviously unilateral. Performance evaluation should be toward multi-indicators synthetically (Lin et al, 2004). So the key problem is to design a performance evaluation model, by which multi-indicators and their relationships could be reflected, and users could utilize it to understand the operation status of business process in a more lively way.

In the service eco-system, workflow performance evaluation faces new challenges, for instance, how to identify the availability of services, how to guarantee the quality of service, what is the relationship between business performance, workflow performance and IT service performance.

Chen (2003) analyzed workflow performance through generalized stochastic Petri nets, yet needed further research on resource distribution and data share. Menasce (2002) discussed the QoS issues in web services. Song and Lee (2005) introduced a performance analysis tool sPAC especially toward web service, but only time properties are illustrated. Furthermore, it still lacks an efficient and general performance evaluation model.

6 APPLICATION OF SOWF The development of SOWF demands the researchers to create special theory, technology and method to

synthetically apply Web Service, Grid, SOA, agent, workflow and some other technologies and develop enabling platforms in order to completely support the applications of SOWF, thus improving the resource utilization rate and the level of system integration.

Undoubtedly, the research and implementation of the theory, method and application of SOWF will result in revolutionary changes in operational mode of workflow, and bring about remarkable commercial benefits for enterprises. The advantages of developing the research and application of SOWF are illuminated as follows:

It can enhance the cooperation among global enterprises and interoperability of different applications, utilize global resources to develop the co-design and distribution of products, and satisfy the demand of market effectively.

It can enrich the alternate models in Web service, and expand Web service to a wider operation of enterprises through offering various alternate models that the enterprises need most.

It is beneficial for improving collaborative services among different enterprises, developing existing application systems operated independently in each company to integrated, hierarchical distributed systems, and achieving dynamic integration of diverse kinds of services.

It can achieve the share and integration of resources in the whole society to the full extent, and increase the resource utilization rate.

It can settle a solid technologic base for the realization and development of newly prosperous enterprises.

545

Figure 3 – An online shopping example

Taking an online shopping process for example, Figure-3 illustrates the scenario of a business process in service-oriented environments. When the customer wants to buy something, he submits the order with his customer information, such as user ID and password. Then, the system checks whether the customer exists or not, if yes, continue to check whether the inventory of selected goods is enough, if not, give an error report to the customer and remind him to register again. When the inventory is lacking, the system notifies the customer and commends him to purchase other related goods. While if the inventory is enough, inform the customer to select pay mode according to his preference, such as bank transfer, remit, pay on delivery and e-Pay. After receiving the money, the company delivers the goods in the way that the customer prefers. If the customer is satisfied with the goods, he accepts it. However, if he is not satisfied, the goods is returned to the company.

In this process, some activities need to communicate with the customer, and some activities work as services, requiring agents in Enterprise service bus to perform corresponding functions. For the activity “check customer”, the agent queries the CRM system of this company, which provides query service and other services. Similarly, for the activity “check inventory”, the agent queries the ERP system. The functions of CRM and ERP systems are encapsulated into inner-enterprise services which could be invoked by other systems.

For payment service, there are mainly four kinds of services, bank transfer, remit, pay on delivery and e-Pay. Suppose the customer selects e-Pay to pay the cost online, then he decides the delivery mode including home-delivery, mail, express and EMS (Express Mail Service). The payment service and delivery service are all inter-enterprise services that are open to users out of the domain of their providers. Each kind of service may have hundreds of providers, so the agents also need to choose the most suitable service with best performance and lowest cost.

In traditional situations, ERP and CRM system may be invoked from different portals and cannot be integrated seamlessly. The payment and delivery activities also need to be completed by searching service providers artificially. For example, if the customer selects the home-delivery mode, the seller has to ask various logistic companies for the cost, time and security of home-delivery service, then to decide which logistic company will perform this service according to the customer’s requirements.

However, through encapsulating inner-enterprise services including legacy systems like CRM, ERP, SCM and PDM systems, and through developing inter-enterprise services covering physical resources, information resources and service resources distributed in the world wide, the issues of information integration, process integration and application integration are resolved.

546

7 CONCLUSIONS In this paper, a three-dimensional service model is proposed and the characteristics of Service-Oriented Workflow are depicted. Then the architecture of SOWF is described which explains the operation mechanism and function modular of SOWF. Some key technologies for SOWF are discussed, especially process modeling, process execution, workflow implementation technologies, transaction management and performance evaluation. Finally, the advantages of developing the research and application of SOWF are illuminated with an example of online shopping process.

As a new technology, SOWF is far from mature now, so there are many fields deserve attention, such as the process modeling of SOWF, the resource discovery and matching, the cooperation of multiple engines in distributed workflow system, and the performance monitoring of workflow in loosely coupled environments.

ACKNOWLEDGEMENTS This work was granted by National High-Tech R&D (863) Plan of China (2006AA04Z166) and National Natural Science Foundation of China (60674080).

REFERENCES Ankolekar, A., Martin, D., McGuinness, D., et al,

“OWL-S' Relationship to Selected Other Technologies”, 2004, Retrieved: 10th Nov 2006, <http://www.w3.org/Submission/OWL-S-related/#bpel>

Cao, J., Li, M.L., Wei, W., et al., “A Distributed Re-configurable Grid Workflow Engine”, Lecture Notes in Computer Science, Vol. 3993, 2006, pp 948-955

Cao, J.W., Jarvis, S.A., Saini, S., et al., “GridFlow: Workflow Management for Grid Computing”, Proceedings of the 3rd international symposium on cluster computing and the grid, IEEE Computer Society Press, 2003, pp 198-205

Casati, F., Shan, M.C., “Definition, Execution, Analysis, and Optimization of Composite E-Services”, IEEE Data Engineering Bulletin, Vol.24, No. 1, 2001, pp 29-34

Chakraborty, D., Joshi, A., Yesha, Y., et al., “Toward Distributed Service Discovery in Pervasive Computing Environments”, IEEE Transactions on Mobile Computing, Vol. 5, No. 2, 2006, pp 97-112

Chen, X., “Performances Analysis of Workflows Based on Generalized Stochastic Petri Nets”, Computer Integrated Manufacturing Systems, Vol. 9, No. 5, 2003, pp 399-402

Hao, Q., Shen, W.M., Wang, L.H., “ Collaborative manufacturing resource scheduling using Agent-based

Web Services”, International Journal of Manufacturing Technology and Management, Vol. 9, No. 3-4, 2006, pp 309-327

He, Y.L., He, W.P., Yang, H.C., et al., “Workflow and agent based platform for collaborative manufacturing across enterprises”, China Mechanical Engineering, Vol. 16, No. 22, 2005, pp 2010-2015

Hwang, S. and Kesselman, C., “Grid Workflow: A Flexible Failure Handling Framework for the Grid”, Proceedings. 12th IEEE International Symposium, 2003, pp 126-137

Kuechler, W.Jr., Vaishnavi, V.K., Kuechler, D., “Supporting optimization of business-to-business e-commerce relationships”, Decision Support Systems, Vol. 31, No. 3, 2001, pp 363-377

Laszewski, V.G., Amin, K., Hategan, M., et al., “GridAnt: A Client-Controllable Grid Workflow System”, Proceedings of the 37th Hawaii International Conference on System Sciences, 2004, pp 1-10

Laszewski, V.G., ”GSFL: A Workflow Framework for Grid Services”, 2002, Retrieved: 15th Jan 2007, < http://www-unix.globus.org/cog/projects/workflow/>

Lin, C., Shan, Z.G., Lin, F.Y., “Quality of Service of Computer Networks”, Tshinghua University Press, Beijing, 2004.

Menasce, D.A., “QoS issues in Web services”, IEEE Internet Computing, Vol. 6, No. 6, 2002, pp 72-75

OASIS, “Web Services Flow Language (WSFL)”, 2002, Retrieved: 3rd Feb 2007, < http://xml.coverpages.org/wsfl.html >

Oinn, T., Addis, M., Ferris, J., et al, “Taverna: A Tool for the Composition and Enactment of Bioinformatics Workflows”, Bioinformatics, Vol. 20, No. 17, 2004, pp 3045-3054

Song, H.G., Lee, K., “sPAC (Web Services Performance Analysis Center): Performance Analysis and Estimation Tool of Web Services”, Proceedings of the 3rd International Conference on Business Process Management, 2005, pp 109-110

Tan W. and Fan Y.S., “Research on Service Matching and Composition in Networked Manufacturing Environment”, Computer Integrated Manufacturing Systems, Vol. 11, No. 10, 2005, pp 1408-1413

W3C, "WSDL 2.0", 2002, Retrieved: 20th May 2006, <http://www.w3.org/2002/ws/desc/>

Wu, J, Wu, C.H., Li, Y, et, al, “Web Service Discovery Based on Ontology and Similarity of Words”, Chinese Journal of Computers, Vol. 28, No. 4, 2005, pp 595-602

Xiao, Y., Chen, D., Chen, M., “Research of Web Services Workflow and its Key Technology Based on XPDL”, Proceedings of 2004 IEEE International Conference on Systems, Man and Cybernetics, Vol.3, 2004, pp 2137- 2142

Yan, S.F., Li, Y., Deng, S.G., et al., “A Transaction Management Framework for Service-based

547

Workflow”, Proceedings of the International Conference on Next Generation Web Services Practices, 2005, pp 377-381

Yan, Y.H., Maamar, Z., Shen, W.M., “Integration of Workflow and Agent Technology for Business Process Management”, Proceedings of the Sixth International Conference on Computer Supported Cooperative Work in Design, 2001, pp 420-426

YOU Junkao, “Research and Implementation of Workflow Management System Based on Web Service”, Master Dissertation, 2006

Yuan, Y.P., Yu, T., et al., “QoS-based Dynamic Scheduling for Manufacturing Grid Workflow”, Proceedings of the Ninth International Conference on Computer Supported Cooperative Work in Design, Vol. 2, 2005, pp 1123-1128

Zhang D.L. and Lin C., “Fast Finding Service on the Network by Genetic Algorithm”, Journal of Tongji University(Natural Science) , Vol. 34, No. 2, 2006, pp 260-263

Zhao, X., Liu, C., “Supporting Relative Workflows with Web Services”, Proceedings of 8th Asia-Pacific Web Conference, 2006, pp 680-691

Zhou C.S., Jiang S., Gu H.J., et al., “Research on Web Service Composition Platform”, Computer Engineering, Vol. 32, No. 5, 2006, pp 269-270

Zhou, X., Cao, J., Zhang, S., “Agent and Workflow Integration Based on Service”, Computer Integrated Manufacturing Systems, Vol.10 No.3, 2004, pp 281-285

548