In this paper, we introduce IDECSE a new integrated approach for composite services engineering which is based on Semantic Web and Data Mining. IDECSE considers semantics in all the composition steps: user query, semantic classi cation of services in the registry, composing services, and verifying the composition process. By considering semantics for describing, discovering, composing, and monitoring services, IDECSE addresses the challenge of fully automating the discovery, composition and monitoring processes while reducing development time and cost. IDECSE appeals for data mining techniques, namely Formal Concept Analysis, for classifying and mining services into service registry in order to anticipate relevant services search and reduce services search space.
Web services are software applications that can be advertised, located and invoked across the Web. Nowadays, an increasing amount of organizations implement their business core and outsource their services over Internet. Thus the ability to e ectively select and integrate di erent services at run-time is an important step towards the development of Web service applications. If no single Web service can satisfy the functionality required by the user, there should be a possibility to create and compose new Web services from existing ones. Considerable academic research and industrial e orts have focused on various aspects of Web service composition ranging from service discovery, to composite service speci cation and deployment. In this context, important initiatives have been conducted to provide tools and languages that allow an e cient integration of heterogeneous services. Standard languages such as UDDI4, WSDL5, and
SOAP6 were proposed to de ne standard ways for service discovery, description, and invocation. WSBPEL7 has focused on representing service compositions where the process ow and bindings between services are known a priori. Later on, following the emergence of the Semantic Web and the fast growth of its related technologies, enhancing Web services description by a semantic level has became one of the basic requirements for e cient services discovery and composition. Since then, several standardization e orts have been done to provide languages which allow to semantically describe Web services on the one hand, and which support e cient automation of the discovery and composition tasks by formal reasoning on services description on the other hand. Standard languages, mainly Web Ontology Language for Services (OWL-S)8 and Web Service Modeling Ontology (WSMO)9, were proposed to allow considering semantic aspects in the description and reasoning about services. Based on such languages, many frameworks was proposed for services composition and deployment [1{4]. Despite these e orts and progresses, Web services composition remains a challenging task for the following reasons. First, the number of available services is dramatically increasing. This requires composition frameworks to be accurate, scalable, and reliable to look up and select the most appropriate services for users requirements. Second, services are developed by di erent organizations based on di erent types of models and platforms. This heterogeneity in services modeling creates semantic gaps between the presentation of their speci cation. This requires composition frameworks to provide e cient tools to support bridging such gaps. Third, Web services can be created and updated rapidly. This requires the composition frameworks should be able to dynamically detect and interact with such changes at run-time. Fourth, specifying composition requirements needs the use of high-level languages that are easy to understand, in order to allow end users to express their functional and non-functional requirements in an e ective way. Fifth, in case of failure to ful ll user's goal, the composition process should be able to iteratively re ning the goal speci cation in an intuitive way to build composite services. Finally, run-time monitoring and adaptation strategies are primordial to ensure the correctness and the scalability of the composition environments. While numerous composition approaches have been developed, very little has been done towards dealing with these challenges. In this context, this paper introduces IDECSE, Integrated Development Environment for Composite Services Engineering, which considers semantics in all the composition steps: analyzing user query, semantic classi cation of services in the registry, composing services, and verifying the composition process. This approach aims to provide an easy way to specify functional and non-functional requirements of composite services in a precise and declarative manner, to guide the user through the composition process, while allowing modi cation or feedback, and nally to enable generating outputs in a deployable language. The rest of the paper is organized
as follows. Section 2, presents the architecture of the proposed framework and details its modules. Section 3 brie y reviews the best known existing approaches before comparing them to IDECSE. Section 4 concludes the paper and outlines current and future work. 2
The IDECSE follows the generic architecture of Web services composition
frameworks [
IDECSE enhances this architecture to address the challenge of fully integrating semantics in all stages of the composition global life-cycle. First user requirements are more understood using re nement techniques such as generalization or speci cation of concepts from a given ontology. Second, IDECSE appeals for data mining techniques for classifying and mining services into service registry based on semantic relations. Third, IDECSE is based on two types of reasoners: a similarity-based reasoner and a logic-based one. The IDECSE architecture is depicted in Figure 2. It consists of ve modules covering the global composition life-cycle (i.e. speci cation, classi cation, composition, deployment, and monitoring ). These modules are described in the following sub-sections.
The Service Request module translates user requirements to an internal language to be used either by the Service Classi cation module or the Service Reasoning module. The Graphical Query Editor relies on domain ontology to "understand" the requirements before enriching them through adding new ontology concepts based on semantic relations such as generalization, specialization, etc. The Query is then parsed to extract functional and non-functional requirements. Functional requirements are modeled using the IOPE Extractor, which extract the Input, Output, Precondition and E ects. Non-functional requirements are speci ed as QoS parameters. Extracted user requirements are then modeled as a new requested service called SR. Given a domain ontology O, a user query Q modeled as SR, consists of a set of provided inputs SRin O, a set of desired outputs SRout O, a set of preconditions SRpre O, a set of e ects SReff O, and a set of quality of service constraints SRqos = f(q1; v1; w1); (q2; v2; w2); :::; (qk; vk; wk)g, where qi(i=1;2;:::;k) is a quality criterion, vi is the required value for criterion qi, and wi is the weight assigned to this criterion such that Pk i=1 wi = 1, and k the number of quality criteria involved in the query. We can model SR as SR = P IOP E + P QoS. 2.2
To deal with the important number of Web services and instead of considering
the whole service registry, this module allows to classify available services
semantically into classes according to their similarities. Its second role is to return only
relevant services to SR from the registry. The module contains four components
which are Service Projector, Service Description Extractor, Service Similarity
Calculator and Relevant Service Selector. The Service Projector selects services
capabilities based on syntactical and semantic description for each service (i.e.
WSDL and OWL-S description for each service) into one interface. The Service
Description Extractor, extracts useful parameters from service capabilities. The
Service Similarity Calculator Sub-Module is based on data mining techniques
for classifying services into classes according to their relevance and similarity
in order to anticipate relevant services search and reduce services search space.
The Formal Concept Analysis (FCA) [
Once the lattice is built, services are grouped into classes according to their similarities. The Relevant Service Selector identi es the most relevant classes of services from the lattice (lattice interpreter), which is more likely to answer the query. The set of relevant services can then be outputted with the appropriate rank with respect to the query needs (Service Ranking) to the next module.
The Service Reasoning module identi es the candidate composition plans that realize the goal through a similarity-based reasoner or a logic-based one. The Similarity Reasoner is based on the value of semantic similarity calculated by Service Similarity Comparator module. When the semantic similarity value between SR and one or more existing services in the registry is higher than a given threshold then SR is satis ed. Otherwise the Logic Reasoner is called to identify the plan of services that achieves the goal of SR. The main components of the Logic Reasoner module are: { Reasoner: checks the ontology consistency in addition to handling the maintenance of state including preconditions evaluation and e ects application. { Filter: avoids redundancy from the plan by identifying service types with potential relevance to the goal and checks the dependency relationships between each two consecutive service types. { Matchmaker: allows querying the service registry for available services in order to match the preconditions of a Web service with the e ects of another. { Abstract Planner: It can be considered as the main component of this module and is responsible for generating a set of abstract plans.
To determine an Abstract Plan, the composition is reduced to a planning problem. A Plan is formalized as a proof of the goal to answer the user query. A Plan P = fAigi=1::n is a sequence of n actions. Each action Ai applies on a state Ei to produce a state Ei+1: 8i 2 f0; ::; n 1g; Ei ^ Ai j= Ei+1. Starting from the initial state E0 the plan P produces the goal G: E0 ^ P j= G. 2.4
The service Execution Module translates the abstract plan into an executable
one by associating to each service type its speci c instances using the Service
Instances Registry. The plan generated by the logic reasoner is considered as
a template for the composite service and drives the process of matching each
service type to a corresponding service instance. The Service Execution Module
is mainly composed of two main components: The Executable Plan Generator
considers non-functional requirements of the goal and enables to concretize the
abstract plan generated by the Service Reasoning Module. The Executable
Composition Analyzer generates executable code and invokes the execution engine
in the Service Monitoring module. Di erent works was proposed in order to
implement the Executable Plan Generator, we can use for example the algorithm
presented in [
Monitoring deals with the actual execution of the composite service and is responsible for monitoring the execution and recording violation of any requirement of the goal service at runtime. If a violation event occurs, an adaptation engine is triggered to handle this violation. There are two inputs to a monitoring framework, a set of constraints that the process must obey and events or messages generated during the execution. A processing engine ensures that all events comply with the constraints and reports any exception. 3
A considerable number of research e orts have focused on various aspects of
Web service compositions ranging from semantic service discovery to
semantic speci cation, deployment, and monitoring. MoSCoE (Modeling Web Service
Composition and Execution) [
IDECSE builds on the approaches mentioned above and aims to provide a declarative approach to service composition engineering to achieve a full manipulation of the composition process. IDECSE appeals for data mining techniques for classifying and mining services into service registry in order to anticipate relevant services search and reduce services search space. It also considers monitoring and adaptation concerns, which are not incorporated in the above approaches. IDECSE provides an easy way to specify functional and non-functional requirements of composite services in a semantic and declarative manner, guides the user through the composition process, while allowing modi cations or feedbacks. 4
This paper describes IDECSE, a new semantic integrated approach for composite services engineering. Compared to existing approaches IDECSE considers semantics in all the composition global life-cycle, addresses the challenge of fully automating the composition processes, uses data mining techniques such as SFCA for classifying and mining services, and proposes new reasoning, monitoring, and adaptation techniques. Our work in progress includes the improvement and implementation of the di erent modules of the proposed architecture. We also plan to extend the framework to include additional features such as failure handling, and an interactive visual environment for testing composite services.