Skip to Main Content
Service Oriented Architecture (SOA) based information systems and services are finding ever wider applicability in complex, mission-critical operational environments. In order to effectively help design and prepare information sharing services for such complex operational environments, designers have to consider various architecture tradeoffs and interactions across the SOA application, communication middleware and network infrastructure layers. Currently, there is a paucity of tools that address this critical need. In this paper, we describe a novel modeling and analysis framework for analyzing and evaluating the performance of SOA-based information services and their myriad design & deployment options over varying network infrastructures under varying conditions of network load. Our modeling framework defines critical Measures of Effectiveness (MOEs) to characterize performance of candidate applications considering the SOA interaction characteristics, as well as the underlying SOA communication middleware and network infrastructures. These MOEs can quantify & evaluate specific Service Oriented Architectures and design/deployment strategies against required Quality of Service (QoS) metrics. Our modeling framework provides the ability to both analyze as-is application scenarios as well as potential alternative architectures that may include differing application characteristics (e.g. increase in the application's data generation rate), varying the SOA interaction characteristics (e.g. pub-sub vs. request-response communication styles, use of JSON vs. XML for data-interchange, use of SAML security tokens etc.) or network characteristics (e.g. varying background traffic and latency). A key feature of our modeling framework is the ability to uncover issues resulting from adverse or unforeseen interactions between the Application, SOA interactions and Network layers. Our initial results on evaluating Information sharing services across enterprise inter-agenc- infrastructures interconnected by public/private networks demonstrate the viability of our cross-layer modeling framework.