Interoperating components, implemented in multiple programming languages, are one of the key features of service-oriented architectures. Middleware, like the WSRF implementation in the Globus Toolkit 4, facilitates service-oriented grid applications and thereby widens the potentials of traditional approaches to component-based grid programming focusing primarily on code reuse and the separation of concerns. In a SOA, every concern can be addressed using the most appropriate implementation technology. Such integrated infrastructures possibly increase a system’s efficiency, but, in the majority of cases, also its complexity. Using the discrete wavelet transform as an example application, we show that skeletal programming allows to handle systems comprising a multitude of different technologies in a structured manner. The computational core of our system is implemented using eSkel, a C/MPI-based library for parallel programming, which we embedded into a grid service and connected to a Java-based web application presenting the user with an abstract, convenient and high level interface. 1.
|Number of pages||8|
|Publication status||Published - 2005|