The authors describe a toolset for performance-based design and analysis of distributed real-time systems. The toolset is based on their design methodology, denoted distributed pipeline scheduling, that provides a set of rules that an engineer can follow to design near-optimal, distributed real-time systems with fully predictable, end-to-end performance properties. The methodology provides (1) models for capturing the application, resource, and system design specifications; (2) an analysis algorithm and figures of merit for evaluating a system design; and (3) allocation and scheduling algorithms for navigating the design space to find a near-optimal solution that meets application timing requirements and optimizes a set of system objectives (e.g., minimize total monetary cost of system and the number of resources used). The toolset, denoted the System Engineering Workbench (SEW), aids system engineers to design, maintain, and upgrade distributed real-time systems by encapsulating the complexities of the methodology, while exporting a graphical user interface that is intuitive and easy to learn. The toolset has been applied to the design of several sonar, medical, and multimedia systems that have end-to-end timing requirements.
展开▼
