- Remember what we set out to do:
- Improve the performance of real-time multimedia transmissions
- Unlike the end-to-end layers (transport & application layers),
we can make major strides towards a real solution
in the network layer (not just a "adjust your rate" kind of solution).
- In fact, the problem of how to modify the network layer so that
the network can make some Quality of Service guarantees,
has been studied extensively even prior to the development
of multimedia applications.
- The first study on providing support for real time network transmissions (not necessarily multimedia data) appeared around 1989.
The following is a list of Quality of Service (QoS) guarantees that a network can make to newly established network connections:
- Loss: the network can make sure that
packets from a new connection will only suffer
a maximum level of packet loss
(e.g., it guarantees that at most 10% of the packets will be lost, or
it guarantees that no packets will be lost).
- End-to-End Delay: the network can make sure that packets
from a new connection
are delivered to the destination within a certain amount of time.
- Throughput Rate: the network can make sure that data from a new connection will receive a minimum amount of bandwidth (e.g., it guarantees that data from a new connection will be transmitted with at least 1 Mbps to the destination.)
- We have seen one extreme example of service guarantee: telephone
- The telephone network provides very strong delay and lossless guarantees to established connections
- The guarantees are provided by exclusive bandwidth reservations
- We can discern the principle behind provisioning of service quality gurantees: isolation through reservation
- By reserving one slot of every 4 slots on the belt,
it is as if we have reserved 1/4 of the space in
the "transmission pipe" between two end-points.
- The "flow of luggages" through the reserved space is not hindered by other flows - isolation.
- Circuit switching is a transmission technique whereby
a certain amount of bandwidth is reserved at connection
setup.
- This is the transmission technique used in telephone networks
- TDM is a bandwidth sharing technique
- A channel of large bandwidth capacity is usually shared
among many users - TDM is one of the common bandwidth sharing
techniques used
- In TDM, the time is divided into slows and the slots are
grouped into periods:
- In Synchronous TDM, slots in the same position
in each period is allocated to the same user
- STDM operates in the same way as the conveyer belt example !
- The good-old telephone system uses STDM and now you should understand why telephone calls are so clear and reliable - delay and lossless guarantees are provided by perfect isolation.
- In Asynchronous TDM (or ATM),
slots are not allocated in a fixed manner.
- You will have to go to my cs558 network class page to read more on ATM. I don't intend to cover ATM here.
- We know that we need to make reservation to provide
a certain degree of performance guarantee to the application.
- So far, we have seen exclusive reservations...
In STDM (or the conveyer belt example), the resource reserved (bandwidth or space on the belt) cannot be used by other user if the owner of the reservation does not use it....
- Such system is clearly undesirable for use on the Internet...
- We need a resource sharing method whereby:
- When the owner of the reservation wants to use its reservation,
it can make use of the reserved resource (bandwidth) immediately
- But when the owner of the reservation currently does not need to use the reserved resource, the system will allow other user(s) to use it !
- When the owner of the reservation wants to use its reservation,
it can make use of the reserved resource (bandwidth) immediately