Performance modeling for service oriented architectures

Reviewer: Carlos Juiz

It is enormously difficult to review a paper of only two pages, but it seems to overview a tool's capacity to model performance of service-oriented architectures (SOAs). The tool can be used by system architects early in the software engineering life cycle to predict performance and scalability, to evaluate architectural alternatives, to provide guidance for capacity planning, and to negotiate service-level agreements (SLAs). It directly models and produces metrics for SOA applications, in terms that are familiar to system architects: services, workflows, and composition of services. It seems that the tool can model complex deployment scenarios, such as server virtualization. Development and evaluation of the tool is performed in the context of architectural modeling for large-scale, SOA-based Australian e-government systems. The tool is designed around SOA models, a simulation engine, and a graphical user interface (GUI). The GUI allows users to develop SOA models in terms of the components of an SOA metamodel: services (simple services and composite services/workflows); servers; workloads (workflows consuming services); and metrics and parameters. SOA models are automatically transformed into run-time models for execution by a discrete event simulation engine. However, there is no information about the capabilities of the simulator that obviously determines the power of the tool, in terms of precision of the simulation studies. The tool enables architects to easily produce SOA performance models from existing architectural artifacts, such as unified modeling language (UML) sequence and deployment diagrams. Therefore, another drawback is that UML diagrams are not formal; they are automatically transformed in formal models of the simulation engine. It seems that models are parameterized with measured performance data (from an unloaded system), hardware capacity (for example, central processing units (CPUs)), and optional configuration information (for example, virtual CPUs). The tool automatically provides an extensive set of performance and scalability parameters and metrics appropriate for each type and combination of model components. While the simulation is running, selected metrics are continuously computed and graphed, and selected parameter values are graphed and can be changed (by using, for example, slider controls), giving immediate feedback. In summary, this very short paper describes a simulation tool for SOA performance engineering, based on UML architectural descriptions of services. The tool provides a flexible GUI and quick feedback to designers, in order to simulate different parametrizations of the system. Online Computing Reviews Service