Model Integrated Computing
for Embedded, Real-Time Systems

Program

Monday, September 8

9:00 – 9:15

Welcome and Opening Remarks

Andrew Watson

9:15 – 9:45

Model-Based Integration of Embedded Systems

Dr. John Bay, DARPA/IXO

9:45 – 10:30

Open Tool Integration Framework for Embedded Systems

Dr. Gabor Karsai, Vanderbilt University

10:30 – 11:15

Advances in Model Driven Middleware for Distributed Real-time and Embedded Systems

Dr. Douglas C. Schmidt, Vanderbilt University

11:15 – 11:45

Clichés, Languages, and Levels of Abstraction

Dr. Joe Cross, DARPA/IXO

11:45 – 12:00

Closing Remarks

Ben Watson, Lockheed Martin Aeronautics

Abstracts

Title:  Model-Based Integration of Embedded Systems

Speaker:  Dr. John Bay, DARPA/IXO

The terms "model-based," "model-driven," and "model-integrated" can have different meanings depending on the context to which they are applied. In the research activities sponsored by the DARPA Model-Based Integration of Embedded Systems (MoBIES) program, model-based development requires that the tools and techniques used in the development process all have an understanding of the environment in which they run. The environment in this context includes the physical characteristics of the host device, which impose temporal, size, weight, and power constraints on the system design. The environment is also determined by the execution platform and underlying operating system, and by user-defined constraints such as security and reliability. In addition, the MoBIES program considers large-scale and distributed embedded systems, such as those found in aircraft, ships, and communications systems. Together, these characteristics impose formidable demands on the design and development system.

The MoBIES approach is to attempt a unification of the mathematical models of the system, its constraints, its performance requirements, and its physical host. This is done by abstracting the models to the meta-model level, and integrating design tools on the basis of the commonalities in model and view descriptions imposed by the applications domain. This may require translators between meta-model and meta-data representations, but it has two distinct advantages. First, it facilitates the design of an open tool integration framework, to which specialty analysis and design tool providers can interface. Second, it can potentially reduce V&V and testing costs because some of the correctness of the system can be built ("meta-programmed") into the tools themselves. Thus, the MoBIES program seeks both meta-programmable embedded system development tool technologies and translation and interchange technologies in order to facilitate construction of the framework. 

This talk will present a status of the MoBIES program, to include an overview of the tools developed thus far, the testbeds on which they have been evaluated, and the transition applications to which they have been applied.

Title:  Open Tool Integration Framework for Embedded Systems

Speaker:  Dr. Gabor Karsai, Vanderbilt University

Embedded system development necessitates the use of a number of design tools: model editors, verifiers, functional and other simulators, synthesis tools, etc., which were typically not designed to work together. In order to support a design process often a process-specific tool-suite is selected, but then the tool interoperability problem has to be solved. The talk will describe a generic, metamodel-driven, open tool integration framework that can serve as the infrastructure for building integrated tool-suites. The framework is based on metamodels of the tools, and a CORBA-based message passing architecture that coordinates tool adaptors with semantic translators according to a workflow model.

Title:  Advances in Model Driven Middleware for Distributed Real-time and Embedded Systems

Speaker: Dr. Douglas C. Schmidt, Vanderbilt University

Traditional approaches to developing and evolving large-scale Distributed Real-time and Embedded (DRE) systems have been unduly coupled with the technology they are based on, which yielded inflexible software that is hard to maintain over long DRE system lifecycles. The last decade has witnessed significant R&D efforts to alleviate these problems, resulting in the factoring out of reusable software that is now available as COTS middleware components. DRE systems are now being developed and provisioned for end-to-end quality of service (QoS) by composing, fine tuning, and deploying these COTS middleware components end-to-end.

The following significant challenges remain, however, to use COTS middleware components effectively to provision the required QoS properties of DRE applications: (1) determining how to distribute DRE application functionality that will provide effective resource utilization and performance, (2) determining the right set of semantically compatible configuration parameters at multiple layers of middleware that will deliver the required QoS properties to DRE applications, (3) provisioning middleware layers and component to adapt to changing operational conditions, (4) reprovisioning DRE applications as existing middleware and hardware technologies are obsoleted by newer ones.

This presentation describes a model driven middleware approach to address the challenges outlined above. This approach is based on (1) domain specific modeling languages that capture application QoS requirements and middleware behavior as platform-independent models and (2) analysis and synthesis tools that use these models to provision platform-dependent middleware layers to support the end-to-end QoS requirements of DRE applications.

Title:  Clichés, Languages, and Levels of Abstraction

Speaker:  Dr. Joe Cross, DARPA/IXO

This presentation offers some viewpoints on the future of MDA-related technologies upon which consensus has not been entirely, ahem, achieved. These include that the expression of functionality is most cost-effectively achieved at a variety of levels of abstraction, rather than at only one or two levels; and that the greatest near-term achievements of MDA-related technologies will be in the translation of domain-clichés, such as "client-server distributed architecture," into implementation-clichés, such as "CORBA system."