Representing Software Architectures for Large Scale Systems Executive Summary

This paper will focus representing the architecture of large-scale systems. UML has a rich set of primitives and extensibility. While some authors have addressed tried to address the use of UML for large-scale systems [1][2][3], little has been written about good representations for large-scale systems. This paper introduces the layered package diagram, which has proven to be a useful view for understanding the logical architecture of large-scale systems. Problems for Modeling Large Scale Systems Large-scale systems commonly come with a host of complications, including: • large amounts of source code (typically millions of lines) • high complexity of interaction between components • extensive use of off-the-shelf components • use of multiple languages • large numbers of developers (often hundreds, often geographically distributed) • multiple persistence mechanism (files, relational databases, object databases) • distribution of components over several hardware platforms • high amounts of concurrency Each of these attributes makes a common understanding of system architecture amongst team members and stakeholders more difficult. The architect faced with designing, managing development, and communicating a largescale system needs useful views of the system. It is simply naive to believe that modeling all the classes and methods of the system is a possible or useful way to view the logical architecture. Typical UML references such as The Unified Modeling Language User Guide [4] or UML Distilled [5] will not provide much guidance