Toward a simulation-generated knowledge base of service performance
Michael A. Smit, A. Nisbet, Eleni Stroulia, Gabriel Iszlai, Andrew Edgar · 2009
The wide adoption of service orientation and the growing complexity of service-oriented applications presents new challenges in configuring and tuning these applications, both at deployment time and at run time. This paper describes our simulation-based approach to addressing this problem. We have developed a general SOA simulation framework, which includes a tool for automatically generating significant portions of the simulator based on the WSDLs of the web services. Our methodology stipulates the systematic exploration of varying SOA deployment configurations in order to collect behavioral profiles. One can use these to examine the trade-offs between configuration cost and performance, ultimately identifying a cost-effective configuration that complies with a Service Level Agreement. The results can also used to identify configurations that allow for more accurate load-balancing and thus better overall performance.