Extending Real-Time CORBA for Next-Generation Distributed Real-Time Mission-Critical Systems

Next generation distributed real-time mission-critical systems, such as integrated distributed medical information systems, collaborating teams of manufacturing robots, and widescale sensor/actuator grids for system damage detection and management, must adapt swiftly to changing environmental conditions. Greater coordination allows elements at all levels to identify and effectively respond to transient opportunities and hazards. Achieving significant levels of coordination requires: (1) systems with an empirically demonstrated ability to accommodate unplanned tasks and changing task priorities in an evolving distributed information and resource availability environment; (2) the ability to share information and control at multiple scales of distribution and timeliness; (3) the ability to trade performance of individual elements for systemlevel real-time performance objectives; (4) the ability to perform adaptive resource reallocations within bounded timescales; and (5) supporting distribution infrastructure to manage resources effectively across both distribution and timescale boundaries. These types of systems have historically been developed largely from scratch, using handcrafted optimizations on each endsystem to achieve the coordination and performance goals outlined above. Unfortunately, expectations of increasing scale and decreasing development cycles make it hard to sustain this development model in a cost-effective manner over long system lifecycles. Solutions built instead using COTS middleware–in particular Real-time CORBA, offer greater reuse of software architectures, patterns, frameworks, analysis techniques, and testing and certification results across entire families of systems. However, these next-generation distributed real-time mission-critical systems require explicit interfaces and mechanisms for key capabilities, such as finegrain adaptive admission control, that are not available in current-generation COTS middleware solutions, such as Realtime CORBA 1.0. Emerging COTS middleware approaches, such as Dynamic Scheduling Real-Time CORBA and the Real-Time Specification for Java[TM], add essential elements for implementing these capabilities, e.g., enhanced threading models and real-time behavioral descriptors. However, unified higher-level approaches and services are needed to realize the