Page 84 - C:\Users\msi\OneDrive\Documents\Flip PDF Corporate Edition\E-Commerce\

Synchronization, Orchestration & QOS

Architecture

A fundamental requirement for multimedia systems is to provide intermedia and
intermedia synchronization. The intermediate subsystems involved in delivering
a stream may introduce delay, jitter, and errors. Along any path these values are
cumulative, and it is the cumulative delay, jitter, and error rate that must be
managed to achieve the end-to-end QOS requirements.

The management of collections of resource managers to achieve end-to-end
synchronization is referred to as orchestration. QOS parameters are considered to
be a basic tool in orchestration.

The definition of QOS parameters which permit system-wide orchestration is
referred to as a QOS architecture.

Synchronization

Synchronization is the coordinated ordering of events in time, and various
mechanism and formalisms for synchronization have been developed, ranging
from low-level hardware-based techniques to abstractions for concurrent

programming languages.

Systems using continuous media data do not require fundamentally new
synchronization primitives, but do require consideration of two aspects of
multimedia applications

1. synchronization events have real-time deadlines, and
2. failure to synchronize can be handled using techniques such as frame
repetition or skipping such that the application can still continue to execute.

For a single media element which has a presentation deadline tp, if the maximum
end-to-end delay due to retrieval, generation, processing, transmission, etc., is
Dmax, then the scheduling of the presentation steps must begin by time tp-Dmax.
If the media object is a stream of elements, not necessarily isochronous, with
deadlines {tpl, tp2, tp3,...}, then the scheduling problem becomes meeting the
sequence of deadlines {tp1-Dmax, tp2-Dmax, tp3-Dmax,...} for each object being
presented. Any admissibility test which is to satisfy the synchronization
requirement must consider the delay requirements of the application, i.e., Dreq <
Dmax. If the average delay experienced per media element, Davg’ is less than
Dmax, then additional capacity exists to schedule other media objects, though
with increased probability of failure.