From the evaluations of the CORBA object middleware we had to learn that the rich functionality of such an architecture limits the efficiency. Especially with new generation computing and data communication hardware the performance gap between low-level systems processing cycles and the application is growing. Recently the Object Management Group published a request for proposal (RFP) asking for contributions concerning the use CORBA in industrial data acquisition systems[Gro99]. Although the RFP gives a detailed list of the issues that are expected to be tackled in the response from industry, necessary modifications or possibilities for configuration in the object request broker architecture are omitted. Although efficiency questions, such as buffering of sensor data or timed invocation are mentioned, no call for assessment of the ORB core support is requested. Interfaces alone however cannot guarantee that a quality of service is met. There is more to it than having the mere possibility of specifying requirements. Requirements finally have to be met and support for meeting them has to be provided. We therefore argue that standardisation of middleware for systems with high efficiency requirements should also concentrate on core support. Higher levels can be still be defined using a standardised interface definition language along with a set of common interface services.
The simple and minimal cross-object invocation scheme as proposed in this thesis has been shown to perform efficiently. It is a self contained module on top of which richer services can be placed. Recently the toolkit was reimplemented making use of a portable operating system library. It has been tested on real-time systems like VxWorks and LynxOS on Intel and PowerPC platforms as well as on ordinary Linux and Solaris systems for Intel and Sparc platforms respectively. The absence of COTS solutions to high performance communication at high abstraction level caused us to adopt this scheme for our future design studies.