Sporadic server revisited

The Sporadic Server (SS) has been introduced to over-come the limitations of other Resource Reservation tech-niques in Fixed Priority scheduled systems. However, even if SS provides good performances with relation to both system responsiveness and schedulability, it presents some drawback mainly related to an increased scheduling over-head. Moreover, to our knowledge, it has not yet been adapted to deal efficiently with overrun situations, nor to work on multiprocessor platforms. In this paper we introduce and prove the effectiveness of an improved SS implementation that allows a reduced over-head, as well as a fairer behavior during server overruns. Moreover, we show how this modified Sporadic Server can be efficiently exploited to provide temporal isolation in a multiprocessor platform, adapting already existing schedu-lability tests to the case under consideration. The simu-lations presented prove the effectiveness of our approach, paving the way for an efficient implementation in the Linux kernel.