Chapter 9: service marts

The use of patterns in data management is not new: in data warehousing, data marts are simple conceptual schemas with exactly one core entity, describing facts, surrounded by multiple entities, describing data analysis dimensions; data marts support special analysis operations, such as roll up, drill down, and cube. Similarly, service marts are simple schemas which match "Web objects" by hiding the underlying data source structures and presenting a simple interface, consisting of input, output, and rank attributes; attributes may have multiple values and be clustered within repeating group. Service marts support Search Computing operations, such as ranked access and joins. When objects are accessed through service marts, responses are ranked lists of objects, which are presented subdivided in chunks, so as to avoid receiving too many irrelevant objects – cutting results and showing only the best ones is typical of search services. This chapter includes a survey of service definition standards (discussing the standards for service description and the current state-of-the-art for service registration and discovery), then introduces a formal definition of service marts and of connection patterns at the conceptual, logical, and physical levels. Then, we show how service marts can be implemented, by taking into account different kinds of data sources, and taking advantage of components (written in Java and SQL) and tools (such as a materialize specifically developed to help service mart implementation). We use such components and tools to build a collection of services used in a running example throughout the chapters of this part.