A framework for building distributed data flow chains in clusters

Abstract

In this paper, we present a software framework designed to set up data flow chains in clusters of PCs or workstations. Communication between the components of such a chain is based on the publishersubscriber- paradigm including bridge components being used to transparently connect components on different nodes. In addition to these bridge components, the framework provides a number of further building blocks to shape the data flow, as well as templates for data sources, sinks, and analysis components. In the bridge components, a class library is used that abstracts from the actual network technology and protocol used and it already exists in versions for TCP/IP and SCI. We will give an overview of the frameworks architecture and present some results obtained in tests of the system.