Analysis of Real-Time Avionics Systems from Tracing and Sampling data

The raw tracing data rapidly amounts to millions of events which are impossible to study globally without the help of sophisticated analysis tools. Modern trace analysis tools, for example in the Eclipse Tracing and Monitoring Framework, offer several views such as a detailed event list, statistics (both numerical and as histogram) and a state timeline for processes, CPUs and other resources (e.g. disks). More recently, we have proposed new algorithms for dependency analysis which can quickly determine the critical path between two events specified by the user as initiating event and result event.

Real-time applications have a number of specific characteristics which deserve their own special purpose analysis modules. The designers of real-time applications reason in terms of periodic tasks, deadlines, latency, slack and proportion of CPU usage. The trace data must then contain enough information about the different tasks and the scheduling events to feed analysis modules that will be capable of automatically computing and annotating this information in a special real-time view.

In milestone D3.1, the existing tools used for real-time applications analysis will be extensively reviewed. In parallel, the typical architectures of real-time partially simulated avionic systems will be surveyed in order to identify all the desirable information that a real-time analysis view could compute and display. Then, in milestone D3.2, new algorithms will be developed in order to efficiently measure a number of important real-time properties such as the latency, deadlines, slack, and work time profile and frequency of real-time tasks. These new algorithms will then be prototyped, evaluated and refined in a new Eclipse analysis plugin as part of milestone D3.3. This will be followed in milestone D3.4 by publishing in the scientific literature and transfer to the Linux Tracing Tools.

In milestone D3.5, the dependency analysis algorithms will be extended to include real-time tasks. Their spontaneous collection of data and real-time deadline to activate actuators constitute additional inputs and outputs linked in the dependency chains. These extended algorithms will be prototyped, evaluated and refined in milestone D3.6. Publishing and technology transfer activities will follow in milestone D3.7.