next up previous contents
Next: Achievement of the Presented Up: Introduction Previous: Data processing   Contents

Problem Statement

Development of software for high energy physics experiments is highly prototype driven. The object-oriented approach is an applicable paradigm for addressing such working environments. It allows the encapsulation of implementation details into self contained units. These units can be abstracted to general ideas which can be reused. Thus modular composition and hierarchical design complete the model[Boo94, pp. 40]. All these cornerstones of object-oriented design enable us to produce componentware for inter subsystem communication. These components act as a glue between the application components of the data acquisition system.

This thesis covers two aspects of object communication in data acquisition: highly efficient interprocess communication and system management for a distributed system. From the use cases[Jac95, pp. 101] and the quantitative system requirements[Jac95, pp. 251], we conclude that one common object-oriented communication tool is not sufficient to cover all aspects. Three dimensions of performance in distributed DAQ systems ( speed, scale, configurability) have to be addressed, but currently no off-the-shelf solution covers them all. Most approaches that we find are either not flexible or not fast enough. Moreover the complexity of some tools degrades their usability. Simplicity shall therefore drive the design of our proposed solutions. Figure 3 shows how we structure our research. This partitioning is reflected in the thesis.


\begin{Figure}
% latex2html id marker 176\begin{center}
\epsfig {figure=over...
...ct oriented communication in data acquisition systems.} \end{center}\end{Figure}

First we identify the requirements of the software for the DAQ system. Two of the issues, namely high speed distributed object communication and management of the distributed system, are the subjects of our research. For each of them we first aim at finding a commercial solution. As we are not satisfied with what we found on the market we present custom approaches for both topics. These solutions involve components of finer granularity. For high speed distributed object communication we have to find solutions for portable and efficient interprocess communication as well as for concurrency mechanisms. A management tool for a distributed data acquisition system must address the problems of application control and policies for performing its tasks. Again, both problems involve an investigation of concurrency mechanisms.


next up previous contents
Next: Achievement of the Presented Up: Introduction Previous: Data processing   Contents
Johannes Gutleber
1999-10-29