Predictability can be regarded as an effect of choosing suitable hardware and software architectures (in a wide sense) that lead to systems whose worst-case behaviour is easy to predict, and of utilizing analysis techniques that are able to provide these guarantees for the chosen system architecture.

Important architectural considerations occur on many levels in a system hierarchy. As a general rule, static allocation of resources leads to predictable systems, whereas dynamic allocation makes predictability difficult.

 

 

Challenges addressed by this activity appear at all levels of abstraction in the design process:

• Modeling and Validation of systems and of components: Principles and structures for system and component modeling that are conducive to achieving predictability, by allowing a priori predictability analysis and by allowing mappings to platform architectures that preserve predictability. Investigations of how modeling and analysis techniques extend to non-traditional system structures, including distributed and networked architectures, for which predictability is more difficult to achieve. Precise definitions and characterizations of the central concepts, including predictability, robustness. Exploring trade-offs between predictability, resource consumption, and performance.
• Compiler Techniques and Program Analysis: Timing analysis, i.e., predicting the worst-case execution time (WCET) of a piece of code, is a hard problem, but significant breakthroughs have been obtained in recent years for many types of processors. Commercial tools, all from Europe, are available. Research in timing analysis is closely dependent on research on system-design concepts that increase predictability The issues stretch from the processor architecture across all layers to the application and is caused by the variability of execution times. The goal is to increase the predictability of system behaviour. An important issues is also timing analysis for compilation, especially in the light of multiple processors and other architectural features. An important goal is to marry timing analysis with compilation, in order to make timing properties immediately visible to the embedded systems developer.
• OS/MW/Networks: On the operating system level, scheduling and reservation of resources is a widely researched topic, with a vast literature. Operating system mechanisms, such as scheduling, mutual exclusion, interrupt handling and communication, can heavily affect task execution behaviour and hence the timing predictability of a system. For example, preemptive scheduling reduces program locality in the cache, increasing the worst-case execution time of tasks compared with non-preemptive execution. The object-oriented programming style, although attractive as a software development methodology, introduces dynamics into the execution time by the dynamic binding of methods to calls. Techniques that improve predictability include schemes that a priori reserve resources in a wide sense. This can be in the form of reserving time slots for execution of tasks, reserving time slots for communication between tasks (e.g., in the time-triggered architecture and in the synchronous programming paradigm). In future research, it is important to explore the tradeoff between performance and predictability in scheduling. Also important is to investigate of software architectures for time-predictable real-time operating systems, with the goal to avoid that the execution of OS code adversely affects the time-predictability of application tasks and vice versa, thus making the computation-time needs of both operating system activities and application tasks easily predictable.
• System and Processor Architecture: Simple processor architectures lead to more predictable systems than complicated ones. Current architectures include many features that decrease predictability, such as implicit concurrency, e.g., pipelining, super-scalarity, out-of-order execution, and dynamically scheduled multi-threading. The restricted processor-memory channel-bandwidth and the growing speed gap between processor and memory has led to the introduction of deep memory hierarchies and several types of speculation. Dynamic power management technology, which is critical for reducing the power consumption of hardware, also has a significant impact on predictability. Research on predictability has considered, e.g., to replace dynamic memory management by static and predictable ones, such as scratchpads, to characterize and develop more predictable replacement policies in dynamic caches,
  The current introduction of multicore processors provides new challenges to predictability, since they introduce new concurrency and communication needs to system development. It is not yet clear how to build predictable and performant systems on multicore platforms.

Réalisation Axome - Création de sites Internet