So far, the application level support and transport level support for real-time multimedia data transfer is most band-aid solutions....
In a nutshell, all methods we have encountered so far try to detect if there is congestion and adjust the transmission rate accordingly
In a sense, that all we can do when we work on the end-to-end level...
At the end-to-end level (transport level and above), the network is an unalterable "black box" and the only thing that the source can do to affect the video quality is to adjust the transmission rate so that it transmits at the highest possible rate without dropping an unreasonably large amount of packets.
Things gets much more interesting when we try to provide support for real-time (multimedia) transmission at the network level
The network level can in fact provide very hard guarantees in supporting real time transmissions.
One example of such support is the plain old telephone network:
- Voice packets transmitted over the telephone network is guaranteed
to arrive with almost the speed of light to the destination.
- Voice packets will not be dropped by the network
- A good illustration of how the telephone network can
provide such strong guarantee in transporting voice packets
is the "luggage transport" conveyer
belt system where the belt spaces is divided into groups of "buckets":
-
The belts run from one "exchange" node to another and rotates
at constant speed.
-
Each group of buckets contains the same number of buckets and
a "connection" reserves a bucket space on every belt along
the path from the source to the destination.
- This conveyer belt system can guarantee that a luggage will be transfer from the source to the destination almost without any delay.
- The guarantees provided by the telephone network does not come cheap.
- In order to provide such rigorous performance support to video transfer,
the telephone network makes unsharable reservation of
the network resources to each telephone connection.
- It is extremely expensive if you do not share
(but clearly extremely convenient -
you get what you pay :-))
- If we would run the Internet like a telephone network, your ISP will be charging you long distance fees for every page you download from Yahoo... clearly not what you want (nor I)
- The theory on providing support for real-time transmission
on the Internet began well before any work on
multimedia data transmission on the Internet.
- First work on support for real-time transmission on the Internet
appeared around 1989, while video and audio trasnmission
application on the Internet did not appear until 1992.
- We will next study a number of algorithms for providing
support for real time transmissions that make
SHARABLE reservations
- these techniques can
be used to support multimedia transmissions as well - although
they were not primarily developed with multimedia transmission
in mind.
- The study of supporting real-time transmission in the network
layer is highly theoretical - in the sense that
there are no real implementations available
- Nevertheless, real solutions do exist and are implemented,
but not on real networks.
- Algorithms for the network layer are tested using
simulations
- Hence, I will also introduce you (briefly) to the Process based smulation technique