Adaptive Deployment and Configuration Frameworks for Component-based Applications

Abstract

Over the past decade component-based technologies, such as the CORBA Component Model (CCM), have found increased acceptance as vehicles for implementing distributed systems at many levels of complexity and size. Examples of such applications occur in such diverse domains as distributed sensor webs, air traffic management systems, industrial control systems, and total ship computing environments. Each of these domains represents a unique set of challenges from the perspective of the deployment and configuration of distributed applications. Software frameworks that support the deployment and configuration (D&C) of distributed real-time and embedded (DRE) systems and applications are typically highly specialized for one particular domain or problem area, such as embedded domains or large-scale grid and cloud-based deployments. Moreover, these D&C tool-chains typically support only one application type or component model and lack standardized mechanisms by which their behavior can be adapted and customized for the challenges unique to a particular domain. This inflexibility can lead to problems in real-world use of D&C technologies because applications must often interact and integrate with either new or legacy systems that use different communication or deployment mechanisms. Additionally, the D&C framework itself may require customization to satisfy domain requirements, e.g., to change the behavior of error handling or event logging. This dissertation presents a set of research contributions to the area of robust and extensible deployment and configuration frameworks for distributed component-based systems. First, the application of component deployment and configuration technology to overcome challenges in mutli-agent software system for a distributed sensor web is described. Next, the Locality Enabled Deployment and Configuration Engine (LE-DAnCE), an adaptive and heterogeneous deployment and configuration framework is described including techniques for performance optimization during the deployment and configuration process. Finally, a connector-based infrastructure that allows extension of component containers with new distribution middlware is described.