Image Processing Reference
In-Depth Information
2.4.6 Design to Support the Entire Engineering Life Cycle
Although many middleware solutions address the run-time behavior of networked embedded
systems, few of them address earlier stages of the engineering life cycle. In particular, for networked
embedded systems where simulation is used to gauge performance in the target system prior
to system integration, additional special-purpose mechanisms may be needed. The virtual clock
described in Section .. is a good example of how such special-purpose mechanisms can be
provided in middleware so that () application software is not modified for use in the simulation
environment, and () the mechanism once developed can be reused for multiple applications.
2.5 Related Work
MicroQoSCORBA [] focuses on footprint reduction through case-tool customization of middle-
ware features. Ubiquitous CORBA projects [] such as LegORB and the CORBA specialization of
the Universally Interoperable Core (UIC) focus on a metaprogramming approach to DOC middle-
ware. he UIC contains “meta-level” abstractions that diferent middleware paradigms, e.g., CORBA,
must specialize, while ACE, TAO, and nORB are concrete “base-level” frameworks. e*ORB [] is a
commercial CORBA ORB developed for embedded systems, especially in the Telecommunications
domain.
The time-triggered architecture (TTA) [] is designed for fault-tolerant distributed real-time
systems.WithintheTTA,allsystemactivitiesareinitiatedbytheprogressionofagloballysynchro-
nized time-base. his stands in contrast to event-driven systems, in which system activity is triggered
by events. The Time-triggered Message-triggered Object [,] architecture facilitates the design
and development of real-time systems with syntactically simple but semantically powerful extensions
of conventional object-oriented real-time approaches.
2.6 Concluding Remarks
We have described how meeting the constraints of networked embedded systems requires careful
analysis of a representative application, “as an essential tool for the development of the special-
purpose middleware itself.” In addition, discovering “which” settings and features are best for an
application requires careful design
apriori
. It is therefore important to adopt an iterative approach
to middleware development that starts with specific application requirements and takes simulation
and experimentation results into consideration.
By integrating both real-time middleware dispatching and a virtual clock mechanism used for
simulation environments with distribution middleware features, we have shown how to develop
special-purpose middleware solutions that address multiple stages of a networked embedded sys-
tem's engineering life cycle. We also have empirically verified [] that with nORB the footprint of a
statically linked executable memory image for the ping node scheduling application was % of the
footprint for the same application built with TAO, while still retaining real-time performance similar
to TAO.
Acknowledgments
We gratefully acknowledge the support and guidance of the Boeing NEST OEP Principal Investigator
Dr. Kirby Keller and Boeing Middleware Principal Investigator Dr. Doug Stuart. We also wish to
thank Dr. Weixiong Zhang at Washington University in St. Louis for providing the initial algorithm
implementation used in ping scheduling. This work was supported in part by the DARPA NEST
(contract F--C-) and PCES (contract F--C-) programs.
