Introduction - Scalable Continuous Media Streaming Systems - page 13

Retrieval
delayed

Data packet loss

Server

. . . . . . . . . . . . . . . . . . . . . . . .

Client

Send req

Longer network delay

Playback deadline missed

Figure 1.11 Variations in the system can disrupt continuous media playback

in the network or changes in the path the data took to reach the client. For networks that
support quality-of-service (QoS) guarantees, the delay and delay variations can be known (and
controlled) a priori . Otherwise, for public networks such as the Internet, the delay and delay
variations are often unknown and unpredictable.
Finally, the media client, upon receiving the media data from the network, will buffer them
in local memory to wait for decoding and playback. The decoding process typically takes a
variable amount of time, and even the playback speed may sometimes vary. This is especially
significant in the PC environment where the decoder/player is implemented in software and
has to compete for processor time with other multi-tasking applications.
The above discussions illustrate the fact that variations in a media streaming system are the
norm rather than the exception. Any one of these variations can lead to media data arriving too
late for playback, i.e., losing playback continuity as depicted in Figure 1.11. In the subsequent
chapters we will illustrate some ways to compensate for these variations to ensure continuous
media playback.

1.5.3 Real-time Interactivity
In addition to stored media, some applications such as Internet phone and video conferencing
send live media streams encoded in real-time to participating users. Unlike stored media, which
a few seconds' start-up delay is tolerable, these real-time applications cannot tolerate too long
a delay (e.g., 150 ms one way in voice call). This stringent delay requirement consequently
puts much tighter constraints on the design and implementation of the media streaming system.
It is worth noting that this real-time requirement often conflicts with other design goals, such
as continuity or efficiency, and thus careful trade-offs are needed to balance these conflicting
requirements.

1.5.4 Efficiency
Continuous media typically generate huge amount of data for storage and delivery. Thus, in
designing media streaming systems, efficiency in the various subsystems becomes crucial in