Userbase

Statistical Multiplexing is a method of making the most efficient use of the bandwidth available for transponder or cable transmission, whilst maintaining a quality of
Select a Solution for service across all of the multiplexed channels. It is the process of dynamically allocating bandwidth where it is needed most (i.e. where channel coding is relatively more complex). Despite the impressive bandwidth of today's HFC (hybrid fiber/coax) cable systems, it is still regarded as a limited and expensive resource. Therefore being able to send more services in the same bandwidth by using statistical multiplexing makes it an extremely attractive proposition.

Why is “statistical” multiplexing different from simple multiplexing, and a system that does not allow multiplexing? What are the tradeoffs made by statistical multiplexing?
Statistical multiplexing allows the total multiplexed capacity to be greater than the actual capacity. It depends on the statistical property of the input, i.e. the average and the peak rates. The service rate in Statistical Multuplexingis generally set as
average rate <= service rate <= peak rate.
Tradeoffs: buffering, queuing delays and losses. Also the systembecomes unstable (i.e. infinte delays and unbounded queues) if service rate < average rate.
Simple multiplexing allows multiplexing upto the “actual capacity”.

Of the three types of MAC protocols (Channel partitioning, Random access and “Taking turns” protocols), which classes allow statistical multiplexing? Why or why not?
Taking Turns and Random Access allow statistical multiplexing.
- Taking turns: Is similar to Round-robin scheduling hence it allows statistical multiplexing. If a user has data to send then he will get a turn. Thus it can be understood as a node getting more slots if it has more data to send and no slots if it has no data to send.
- Random Access: Users grab the channel whenever they have data tosend. This implies that if a user has data to send he can utilize the wholechannel, given the channel is not in use.
- Channel Partitioning: Each user is allocated a “fixed length” slot in each round. This implies that if a user has more data to send then he has towait for the next round even if there is an empty slot going (because some other user didn’t have anything to transfer). Thus we are doing a simple multiplexing here.

Why would the statistical multiplexing efficiency offered by these MAC protocols diminish with either increasing in number of hosts or distance spanned.
- Random Access: Number of collisions increase as the number of hosts increases. Also the collision detection will fail if the length of the medium is beyond a certain threshold. (eg. 500m in Ethernet).
- Taking Turns: If the number of hosts in the network increase there is an overhead in “polling” and latency.

Of the three types of MAC protocols (Channel partitioning, Random access and “Taking turns” protocols), which classes allow statistical multiplexing? Why or why not?
Taking Turns and Random Access allow statistical multiplexing.
- Taking turns: Is similar to Round-robin scheduling hence it allows statistical multiplexing. If a user has data to send then he will get a turn. Thus it can be understood as a node getting more slots if it has more data to send and no slots if it has no data to send.
- Random Access: Users grab the channel whenever they have data tosend. This implies that if a user has data to send he can utilize the wholechannel, given the channel is not in use.
- Channel Partitioning: Each user is allocated a “fixed length” slot in each round. This implies that if a user has more data to send then he has towait for the next round even if there is an empty slot going (because some other user didn’t have anything to transfer). Thus we are doing a simple multiplexing here.