IT practitioners striving to meet changing business needs are using service-oriented architectures to speed development, improve visibility into the business impact of IT events, and lower integration and management costs.
To reap these benefits, IT staffs need to provide some level of semantic integration between IT systems. Semantic integration means that not only do systems need to connect (that is, be able to exchange messages), they also need to have a common language for these messages to be translated into action in an automatic way. Without this shared understanding, messages are exchanged but must be translated by a human operator before resulting in any action.
Semantic integration is achieved via models that contain a description of the elements composing a system, as well as the relationships linking them. While not a cure for all problems, models allow better semantic integration through sharing of model elements and the use of transformations, policies and desired-state type of information.
Models can be easily translated from one modeling language to another, so the invoker of the model and the service providers don't need to use the same modeling language. Service Modeling Language, for example, was designed for that purpose. The sharing of models across the life cycle of the system allows experts to enrich the model with information relevant to each stage of the life cycle (such as design, implementation, testing, deployment, operation and retirement).
WS-ResourceTransfer (WS-RT) plays at the intersection of SOA and model-based management. While its goals are modest and its usage will often be hidden, it meets a critical need in allowing model-driven interactions to be conducted in an SOA.
WS-RT defines a set of Simple Object Access Protocol messages that are used to provide flexible access to a model-driven service. It is fully compliant with the WS-Transfer specification (a World Wide Web Consortium submission that is one of the components of WS-Management) on which it is based.
But while WS-Transfer allows access to the entire representation of the model of a system that is being accessed (in order to read, update, create or destroy it), WS-RT lets individual parts be specified. This capability is useful when interacting with models of individual resources (such as a server) and is critical in interacting with large models that represent complex systems (such as a data center), in which case interacting with the entire model is impractical.
The improvements that WS-RT adds to the WS-Transfer Create operation allows one to specify parts of the model of the system to create. For example, when asking for a server to be provisioned, one might want to specify what operating system it should run and how much memory it should have. But in general the creator doesn't want to have to specify what IP address the server should be assigned (even though this information is part of the model of the server). Rather, the IP address will be assigned automatically at the time of provisioning.
While the ability to provide only a portion of the model at creation significantly improves the usefulness of the Create operation, it still limits it to cases in which the input of the Create operation is a subset of the model of the system. In other words, WS-RT doesn't help when the Create operation needs information outside of the model. But once a system is available (whether it was provisioned through WS-RT or not), WS-RT can be used to manage it through interactions with its model.
While systems management and grids are the most obvious areas of application for WS-RT, it is relevant in all areas where model-based integration is used in coordination with SOA principles, a combination that is key to unlocking many of the benefits expected from SOA and model-based management.