Innovations in computing technology, decreasing costs, and advances in miniaturization have made it physically and economically possible to incorporate computers into a wide variety of devices ranging from diminutive wireless sensors to large consumer appliances. This proliferation of computing devices has served to create a world full of possibly millions of such computers, scattered ubiquitously across physical spaces. These devices in combination represent a computational fabric which has the potential to enable an ambitious new class of applications that can exploit the resources in the physical world and also affect the environment around them. At present however, devices exist in isolation because their software and hardware architectures are incompatible and they have not been designed with interoperability in mind. This project envisages the design and implementation of a software system that (1) will allow large collections of heterogeneous devices to interoperate via the use of standard interfaces, (2) will enable the representation of a (potentially changing) set of such devices as a single composite device to simplify programming, and (3) will support the construction of complex applications from small modular pieces that run across multiple devices. If successful, the project will achieve a seamless integration of the computing and physical worlds in a manner that is highly sensitive to the needs of individual, groups, and organizations. Intellectual advances associated with the project will be incorporated into the computer science curriculum and will be disseminated as open source software.
The intellectual merit of this research is in its exploration and contribution to the state of the art in system software design and implementation, with respect to three areas: composition and decomposition of embedded system components, real-time virtualization, and scheduling of components with different criticalities. The broader impacts of the research supported by this grant include the publication of research results in rigorously peer reviewed conference and journal venues. This research has also led to the development, extension, and release of the RT-Xen real-time virtualization software and the ioa++ framework, both of which are available as open-source software. The award, and the research conducted under it, have supported the training and education of several graduate students who have been involved in the research, including one completed doctorate on mixed-criticality scheduling in system software. Two additional doctorates are currently in progress as of the end of the award period: one on virtualization of real-time systems, and one on composition and decomposition of concurrently exectuting components with a novel model for managing asynchronously dispatched transactions that span multiple components, which is a fundamental extension of the approach that was used to develop the ioa+++ framework. An additional functing supplement to this award also supported two REU students over the summer in developing and evaluating device information dissemination protocols built atop the ioa++ framework, which they presented as one of the 5 featured student research presentation at our department's summer 2011 REU research symposium.
