Skip to main content


An early life-cycle performance modelling approach for reducing the risk of enterprise wide service oriented architecture migration projects


Paul Brebner and Jon Gray



Poster presented at SimTect 2011 Conference


Since 2007 NICTA has developed and trialled a technology for assessing the performance and scalability of Service Oriented Architectures (SOAs). The approach is called “Service Oriented Performance Modelling” and relies on a method and tool support for capturing the performance critical aspects of SOAs in a domain-specific modelling language and predicting performance and scalability metrics through simulation. The method is suitable for assessing risks in large-scale complex software projects at different stage of the software development lifecycle, but has more often been deployed in late phases on partially or fully-built systems during proof-of-concepts, pilots, production switch-on, or major system evolutions. Recently we have been engaged with a project to explore appropriate SOA patterns in a very early phase of a large scale enterprise wide SOA migration. The goals are to inform architectural and procurement decisions through the very early identification of inappropriate software patterns, and to highlight major areas of potential risk to focus on during subsequent phases.

The paper outlines a new “blending" approach we have developed for performance modelling SOAs very early in the development lifecycle, well before a single running implementation of the system exists. This approach utilises information from a variety of sources which are used to construct multiple models, of different types and for different aspects of the system. Information sources include the intended system Use Cases and other existing requirements and architecture documentation, test implementations on multiple different testbeds, future user survey results, vendors and benchmarks. Individual models themselves are based on a blend on information sources, including other models (either through direct re-use and composition of model components, or by incorporating the results of one model in another). Finally, models of increased scope, size and complexity are built to include aspects of the future system that are plausible but which cannot yet be implemented even in testbeds, such as scale, future SOA patterns, and technical infrastructures and topologies.

BibTeX Entry

    author           = {Brebner, Paul and Gray, Jon},
    month            = may,
    year             = {2011},
    howpublished     = {Poster presented at SimTect 2011 Conference},
    title            = {An early life-cycle performance modelling approach for reducing the risk of Enterprise wide Service
                        Oriented Architecture migration projects},
    address          = {Melbourne, Australia}


Served by Apache on Linux on seL4.