Streaming - A Practical Guide to Video and Audio Compression - page 358

There are very few opportunities at the server end to make this more efficient. Some
servers will cache the content so that, when it is requested again, it can be delivered from
local memory rather than from hard disks. This model is used in edge-server architectures.

20.9.2 Multicasting

The technique of multicasting has been developed over the last few years to provide a way
to deliver the same stream to many users at once. This brings the benefits of the broad-
casting model, where everyone receives the same content even though the loading on the
master server is just one stream. With a combination of the multicasting and uni-casting
models, you achieve the ideal situation, where you can serve video on demand but also
gain efficiencies through shared streams when the opportunity presents itself.
The multicasting model is illustrated in Figure 20-5. It is more efficient than the uni-
casting model.

20.10 Implications for Codec Choice

A DV file is already somewhat compressed but not enough to stream it to a desktop. There
are two commonly used bit rates for DV files in broadcast studios. DV50 runs at 50 Mbps.
DV25 is compressed to half that but the compression is fairly light and you cannot see the
artifacts. But this is not compressed enough for deployment to the public. A more compact
format must be used. Your application determines what compression trade-offs you make.
If you share video during the production process and need to synchronize accurately for
edits and audio dubbing, then you must ensure there are no dropped frames. You may be
able to tolerate some picture degradation, however.

One client

One client

Stream
replicator

Server process

One client

One client

One client

One client

Stream
replicator

Server process

One client

One client

Figure 20-5 Multicasting model.