1999 - 2000 - 2001 - 2002 - 2003 - 2004 - 2005 - 2006 - 2007 - 2008 - 2009 - 2010

Announcement - Agenda - Paper Presentations

FRESCOR is a consortium research project, part of the European Union's Sixth Framework P- gramme, which aims to provide a complete framework to support the whole development of a wide range of realtime embedded applications, including tools to ease the work of developers in the phases of analysis, design and implementation. More precisely, part of this work, which has already been done, consisted in the implementation of a scheduling framework, called FRSH, to provide a clean API which supports a very complex set of functionality to deal with real-time multitasking, QoS, distribution, power management and shared resources amongst others.

The key abstraction in FRSH is the use of a hierarchical scheduling architecture based on virtual resources. In order to keep the framework independent of specific scheduling schemes, it defines an interface between applications and the global scheduler, called the service contract. So, instead of using parameters of a specific scheduler algorithm, the application defines its needs in the form of service contracts. This way, FRSH transparently manages all platform specific details, allowing the developer to concentrate in the specification of the application requirements.

Provided as a user library, the FRSH API needs an underling real-time capable platform to ensure the desired timing behaviour to the application on top of it. The pair PaRTiKle + XtratuM was elected as a target platform to provide such support because of its well-defined and standard interface (PaRTiKle implements the POSIX PSE 5.1).

The main effort in supporting FRSH over PaRTiKle was the implementation of a compatibility layer called "FRSH Adaptation Layer" (FOSA for short), which provides a platform independent interface aimed to permit the FRSH run-time to be ported to other platforms with a minimum effort. Nevertheless, the FRSH run-time, and FOSA at its turn, needed features of PSE5.1, such as real-time signals and thread CPU clocks, that PaRTiKle lacked of. Also, FRSH requires a deep and direct control of the application threads' scheduling actions attached to the FRSH scheduler. In order to accomplish a full scheduling control over these threads, the FOSA layer relies on the Application Defined Scheduling (ADS) mechanism, which was not available for PaRTiKle. Thus, significant work was also done to provide PaRTiKle with these characteristics.

This paper present our experience in both, implementing the FOSA layer on top of PaRTiKle and providing the additional features that FRSH depended on and that were not present in PaRTiKle.