Architectural issues in the StormCast system

Abstract

This paper presents a design methodology for large-scale, wide-area distributed applications involving monitoring, collection, and processing of large amounts of data. Examples include meteorological applications or pollution or radiation monitoring. A typical application is the StormCast system, in which several prototype applications have been constructed in the meteorological domain. The methodology exploits a high-level architecture and well-defined abstractions to master the complexity that application developers are confronted with in this application domain. One important design principle is the use of the StormCast architecture as a model for other applications with similar characteristics. This has been shown to be useful in promoting both reusability and rapid prototyping. Another important principle is the use of performance data in the design phase to help decide among design alternatives.