|
Consider the network depicted in the figure below.
|
|
|
In this network, compute the minimum round-trip-time between `A` (resp. `B`) and `D`. Perform the computation if the hosts send segments containing 1000 bits.
|
|
|
How is the maximum round-trip-time influenced if the buffers of router `R1` store 10 packets ?
|
|
|
If hosts `A` and `B` send to `D` 1000 bits segments and use a sending window of four segments, what is the maximum throughput that they can achieve ?
|
|
|
Assume now that `R1` is using round-robin scheduling instead of a FIFO buffer. One queue is used to store the packets sent by `A` and another for the packets sent by `B`. `A` sends one 1000 bits packet every second while `B` sends packets at 10 Mbps. What is the round-trip-time measured by each of these two hosts if each of the two queues of `R1` can store 5 packets ?
|
|
|
When analyzing the reaction of a network using round-robin schedulers, it is sometimes useful to consider that the packets sent by each source are equivalent to a fluid and that each scheduler acts as a tap. Using this analogy, consider the network below. In this network, all the links are 100 Mbps and host `B` is sending packets at 100 Mbps. If A sends at 1, 5, 10, 20, 30, 40, 50, 60, 80 and 100 Mbps, what is the throughput that destination `D` will receive from `A`. Use this data to plot a graph that shows the portion of the traffic sent by host `A` which is received by host `D`.
|
|
|
Compute the max-min fair bandwidth allocation in the network below.
|
|
|
Simple network topology
|
|
|
Consider the simple network depicted in the figure below.
|
|
|
In this network, a 250 Kbps link is used between the routers. The propagation delays in the network are negligible. Host `A` sends 1000 bits long segments so that it takes one msec to transmit one segment on the `A-R1` link. Neglecting the transmission delays for the acknowledgments, what is the minimum round-trip time measured on host `A` with such segments ?
|
|
|
If host `A` uses a window of two segments and needs to transmit five segments of data. How long does the entire transfer lasts ?
|
|
|
Same question as above, but now host `A` uses the simple DECBIT congestion control mechanism and a maximum window size of four segments.
|
|
|
Hosts `A` and `B` use the simple congestion control scheme described in the book and router `R1` uses the DECBIT mechanism to mark packets as soon as its buffers contain one packet. Hosts `A` and `B` need to send five segments and start exactly at the same time. How long does each hosts needs to wait to receive the acknowledgment for its fifth segment ?
|
|
|
Discussion questions
|
|
|
In a deployed CSMA/CD network, would it be possible to increase or decrease the duration of the slotTime ? Justify your answer
|
|
|
Consider a CSMA/CD network that contains hosts that generate frames at a regular rate. When the transmission rate increases, the amount of collisions increases. For a given network load, measured in bits/sec, would the number of collisions be smaller, equal or larger with short frames than with long frames ?
|
|
|
Slotted ALOHA improves the performance of ALOHA by dividing the time in slots. However, this basic idea raises two interested questions. First how would you enforce the duration of these slots ? Second, should a slot include the time to transmit a data frame or the time to transmit a data frame and the corresponding acknowledgment ?
|
|
|
Like ALOHA, CSMA relies on acknowledgments to detect where a frame has been correctly received. When a host senses an idle channel, if should transmit its frame immediately. How should it react if it detects that another host is already transmitting ? Consider two options :
|
|
|
the host continues to listen until the communication channel becomes free. It transmits as soon as the communication channel becomes free.
|
|
|
the host stops to listen and waits for a random time before sensing again the communication channel to check whether it is free.
|
|
