Currently, Service Oriented Architecture (SOA) technologies are being used successfully to solve the syntactic interoperability problem whereas the composition (e.g., orchestration) engines provide increased agility. Nevertheless, such classical SOA technologies and Enterprise Service Bus (ESB) concept technologies as those provided by editors or open source providers don't efficiently support large collaborative business networks, since the service description is still limited (and doesn't provide semantic interoperability) and since the global context (i.e., security, reliability, etc.) isn't taken into account while composing services. Moreover nowadays Web Service standards provide a communications medium for distributed systems, but they can't yet ensure that the communicating parties "speak the same language," which is necessary for smooth, fully automated systems interoperation.
In general, developing large service ecosystems involves taking into account nomadic properties and quality of service requirements, particularly considering that these extended environments may use a large variety of communications infrastructures. Such non-functional requirements can be taken into account in a context-aware service composition process whereas probes can capture the execution context.

The main drivers for next-generation service infrastructure (see Figure 1) are then to:

• Enable a service to be used anywhere from any kind of device (pervasive technologies). These services and related information are stored in a lightweight distributed repository and deployed into a new generation Enterprise Service Bus (ESB) that encompass enterprises borders.
• Address end-to-end Quality of Service (QoS) requirements associated with service provisioning, with a special emphasis on dependability constraints and more specifically scalability, reliability, and security.
• Provide an agile framework supporting business-level Service Level Agreements (SLA) definition and monitor system behavior accordingly to enforce required non-functional properties of potentially composite service execution.

The aim of this article is to describe the components of this framework and show how we take into account the full set of challenges encountered by such large service ecosystems by extending classical ESB and SOA technologies thanks to semantic properties.
The first part is a short presentation of basic semantic service and service infrastructure technologies, followed by an explanation of how to use semantics to improve service governance, SLA definition, and relative enforcement mechanisms. A reactive composition framework is then offered to emphasis the main SOA advances for next-generation service infrastructure.

Semantic Services
In the Semantic Web paradigm, information on the Web is enriched with machine-interpretable semantics to allow its automated manipulation. An entity's semantics encapsulates the meaning of the entity by reference to a structured vocabulary of terms (ontology) representing a specific area of knowledge. Ontology languages support formal description and machine reasoning on ontologies; the Web Ontology Language (OWL)1 was recent standardized by W3C.

Semantic Web Services employ such Semantic Web technology in the Web Services area: service functionality, Web Service inputs and outputs, their preconditions, effect, performance, and other QoS aspects; all are expressed in knowledge representation languages, referring to shared ontological vocabularies. This effort aims at the semantic description of Web Services towards automating Web Services discovery, invocation, composition, and execution monitoring:

• When searching for a service providing a specific functionality, ontology can provide synonyms of words, the taxonomic structure of service capabilities, relationships between service capabilities, etc.
• When trying to harmonize different data formats for two services that have to exchange messages, ontologies can provide elaborated conceptual data models for message descriptions that facilitate automated translation.
• When trying to compose complex business processes from given partial processes implemented by a number of services, automated planning algorithms from artificial intelligence can be employed, provided the semantics of the input services are formally defined. Such semantics can embody QoS attributes that allows a planning algorithm to take them into account.

Hence a number of research efforts have been proposed for the semantic specification of Web Services. The latest WSDL 2.0 standard not only supports the use of XML Schema, but also provides standard extensibility features for using, say, classes from OWL ontologies to define Web Service input and output data types. Actually, SAWSDL2 is W3C's new candidate recommendation here, defining how semantic annotations can be added to WSDL descriptions. Further, employing OWL, OWL-S3 is a high-level OWL ontology for Web Services. A similar recent proposal is the Web Services Modeling Ontology (WSMO)4 that's specified using the Web Service Modeling Language (WSML).

Semantic technologies can be employed further for describing non-functional service properties, such as service QoS, including dependability and security properties, as well as service context. This field is far from being standardized, which raises a number of issues, such as the selection of adequate non-functional properties and their effective description to allow discovery and potential composition of services that will satisfy the required properties.
Key considerations for a semantic service architecture are then:

• Semantic description of services in terms of both functional and non-functional properties; the latter include QoS as well as system and user context;
• Support for QoS specification and enforcement through the use of SLA;
• Service interoperability based on semantics;
• Semantic service registries publishing service descriptions;
• Service selection based on matching between functional and non-functional requirements of a business process and corresponding service properties;
• Support for complex services through service composition and adaptation taking into account both service semantics and behaviour;
• Ensuring QoS, dependability and security properties for service aggregations at the composite service level.

Highly Distributed Service Infrastructure
According to Gartner5, "An Enterprise Service Bus (ESB) is a new architecture that exploits Web Services, messaging middleware, intelligent routing, and transformation. ESBs act as a lightweight, ubiquitous integration backbone through which software services and application components flow."