Improving Real-Time System Con.guration via a QoS-Aware CORBA Component Model

• Authors: Nanbor Wang;Christopher Gill
• Venue: HICSS '04 Proceedings of the Proceedings of the 37th Annual Hawaii International Conference on System Sciences (HICSS'04) - Track 9 - Volume 9
• Year: 2004

A fundamental tension exists in systems research and practice between (1) hiding extraneous details to simplify the system programming model, and (2) revealing crucial details to allow better customization to particular application requirements, data characteristics, or other constraints. For many modern systems, such as distributed real-time and embedded (DRE) systems and multimedia applications, this tension is even stronger: the correctness of the system depends not only on functional behavior (e.g., correctness of interfaces and the algorithms behind them) but also on quality of service (QoS) properties such as CPU speeds, frame rates, and end-to-end timeliness requirements. Satisfying both the functional and QoS requirements of modern applications running in various run-time environments requires programming models that can allow selective visibility and (re)con.gurability of key details in both the application and the services it uses.Although prior research and practice has focused on decoupling system properties from the application logic to facilitate modification at different points in the system lifecycle, some system properties cross-cut both (1) multiple layers (application and supporting infrastructure) and (2) multiple configuration phases in the system lifecycle. To date, there has been little support for configuring such cross-cutting concerns in a systematic and standardizedway across different kinds of applications. This has resulted historically in laborious and ad hoc hand-customization of systems to meet real-world constraints, especially for DRE systems.However, the convergence of model-integrated computing with component-oriented middleware in the DRE middleware research community offers a new opportunity for automated coordination of system modifications at different layers and at different points in the system lifecycle, to meet application requirements end-to-end. We are extending the Component-Integrated ACE ORB (CIAO), to provide a unified system configuration capability that can be leveraged by higher-level modeling tools.This paper provides two main contributions to the state of the art in component middleware for model integrated computing in DRE systems. First, we present an argument based on a representative real-world example from the avionics mission computing domain, that multiple points of coordination in the CORBA Component Model (CCM), i.e., componentpackaging, assembly and deployment, are needed to assure feasibility and increase performance at run-time. Second, we summarize key lessons learned for configuring QoS properties such as event dependencies, rates of execution, and feasibility of deadlines, and describe which of the alternative configuration points in CCM are most appropriate in our example under different design considerations.