Multimedia Systems

Recent evolution of technology makes personal computers powerful enough to handle audio and video streams, so that there is a great interest in providing support to applications that manage this kind of data, also called Continuous Media (CM). Some examples of Continuous Media applications are desktop videoconferencing, distributed shared virtual environments, and collaboration-support systems.

This continuously growing up need for multimedia information, (requiring the integration of text, graphics, sound, and visual data) causes new requests in operating system design. In fact, Continuous Media streams are characterized by an implicit timing, that is not supported by conventional operating systems. Therefore, there is a great interest in a new kind of operating systems, that integrates the flexibility of traditional general purpouse operating systems (such as Unix or Windows) with the new capability of supporting the temporal semantics of CM.

Two ways have been individuated to meet these new requirements: the first uses the classical real-time theory adapting it to manage multimedia tasks (in particular, by using soft real-time techniques), whereas the second one is based on the proportional share resource allocation. In the first case, classical real-time scheduling algorithms have been extended to cope with unpredictable execution or interarrival times, or with unknown system loads. For this purpose, Resource Reservation techniques appear to be particularly interesting. In the second case (Proportional Share scheduling), the CPU is allocated to the processes in time quanta (like in conventional time-sharing systems). Hence, these techniques can be easily implemented on a traditional operating system, and every process is charcterized by a weight. The goal of the scheduling algorithm is to ensure that each process makes progress at a precise uniform rate (proportional to its weight).

The two kind of solutions sketched above result in two different task models:

Click here to see a comparison between an event-based (constant bandwidth) model and a time-sharing (proportional share) model.

In my research, I am focusing on developing an efficient scheduling support for Multimedia applications, enaching the classical Real-Time scheduling theory. The effectiveness of the proposed algorithms have been proved by simulations and by implementing the algorithms in the Hartik kernel.

Go to HARTIK and Multimedia

Go to HARTIK Home Page