Summary

4Networking software in the presence of high efficiency requirements has to be kept simple and small. We were able to show that this is not contradictory with the claim for a flexible and evolvable solution that exhibits a high degree of abstraction. The concept of object-oriented software components manages to bridge this gap. Standard componentware for distributed computing does not meet the needs of high-energy physics data acquisition systems. The need to stay compliant with the standards imposes too many restrictions that hinder optimisation or tailoring to special environments. With the advent of new processing and data communication equipment the performance difference between hardware and software continues to grow. Furthermore distributed systems become more heterogeneous not only in processing and operating system platforms, but also in communication devices. We experience this in highly specialised application domains now, but these technologies will become ubiquitious in the future. It is therefore important to have solutions ready. We were able to show that with the component based software engineering approach we can present solutions that satisfy all needs of such configurations. System management in distributed data acquisition systems has to be shown a necessary research domain. Although the problems can be isolated and solved separately there exists not enough tool support. The key to such a system is the freedom of choice in system management policies. This has been demonstrated to be feasible.

The achievments of the performed work can be summarised as follows:

• Successful application of object-oriented components in hardware/software hybrid systems with the effect of the availability of efficient and yet configurable software.
• Identification of the main limitations in CORBA middleware and assessment of application in Large Hadron Collider data acquisition systems.
• Quantitative analysis of concurrent server architectures for efficient object dispatching. Capturing of the key properties in a novel software component performance model.
• Development and proof of feasibility of a lean distributed object invocation toolkit.
• Design of a system management scheme for heterogeneous distributed systems based on the agent pattern.