Bringing order to service assembly

Back to index page

Previous article: New Taverna Release

Next article: Engage provides a picture of the UK's e-Research community

While working with the UK e-Science community, we spent time thinking about rendering service interfaces for features such as data access and job scheduling. We didn't have to think about how to describe the assembly of a set of such services into a working system, because the open-source Apache Tuscany implementation of the Service Component Architecture (SCA) did it for us.

SCA is a set of specifications currently being standardised at OASIS. The specifications provide a technology-neutral composition capability for assembling applications from services. It defines the simple concepts of components which have references and services. Existing SCA and non-SCA services can be assembled into compositions, and these can be re-used as new building blocks in their own right. SCA (and Tuscany) provides a policy-based model to uniformly describe and realise quality-of-service requirements across assembled services, irrespective of how individual services are implemented or deployed.

SCA scales from fine-grained assembly of local services through to coarse-grained assembly of distributed services. It is a convenient way of describing a network of services that are to be executed across a distributed set of hardware resources. Whether you consider these resources to be a grid, a cloud or more generally the basis for an SOA (Service Orientated Architecture), the SCA concept still applies: a distributed application is modelled as an assembly of wired components.

The motivation of the Apache Tuscany project is to provide an easy-to-use, open-source services infrastructure based on the SCA specifications. Apache Tuscany exploits the extensibility of SCA to support a wide range of service implementation types and communication technologies. Tuscany's OSGi-based distributed SCA runtime allows the application deployer to match the infrastructure, technology and geographic needs of the application—from a full blown JEE application container, a light-weight JSE runtime, or even an Android Virtual Machine.

http://www.tinyurl.com/aej5p3