Question 1:
=============
a) Consider a network in which packets are delivered reliable across each link. That is, for each pair of adjacent nodes A – B, the datalink layer has an ARQ protocol (such as stop-and-wait) that ensures packets sent by A do arrive at B. In this network, why do you think is necessary for the transport layer to also have its own ARQ protocol and do retransmissions of its packets?
b) For each of the following, identify which OSI layer is responsible for performing the described action:
i) Make sure that a packet is delivered reliably across a single physical link
ii) Describes how a binary tree data structure is to be encoded inside the message
iii) Makes sure that a message is delivered reliably from the source host to the destination host
iv) Coordinates the transmission of multiple flows of data across multiple sources and multiple destinations (for example, a video conference)
v) Defines how the bits are converted into voltages
vi) Decides how a data packet is to be routed from one link to the next.
Question 2 –
=============
a) Assume you have a link that has 50ms of round-trip-time, the transmission rate is 3Mbits/sec, and each packet is 1Kbyte. If we are using the concurrent logical channels protocol, how many logical channels do we need to maintain 100% throughput (assuming no message loss)
b) Assume three of the above (i.e. part (a)) logical channels have lost a message and are stuck waiting for the timeout to occur (assume the timeout value unnecessarily large, way too big). How much is the reduction in throughput?
Question 3
===========
Consider the following sequence of data bits
10110100111010100011
How would this be translated into physical bits if we use:
a) NRZ
b) NRZI (high value initially)
c) Manchester
d) 4B/5B
Question 4:
===========
a) Consider bit-stuffing (also known as bit-insertion) the way I present it in the slides, NOT the way it is presented in the book. Consider k = 2. If the data bits of the frame to be sent are
1110011010110011101
what are the actual bits sent by the sender? Please include some idleness bits before and after the frame.
b) Is it possible to have k=1? Why or why not?
Question 5
==========
a) Consider the parity bit protocol with the p’s, q’s, and the additional r bit. What is the hamming distance of this protocol? Briefly explain why
b) Assume we have a CRC protocol that satisfies all the desirable properties that we described in the slides. What is the hamming distance of this protocol? Briefly explain why.
c) For both a) and b), can these protocols be used for error correction, and if so, how many bits can they correct? (i.e., can they perform x-bit correction, and if so, what is x?)
Question 6
==========
a) Assume G = 100101, and M = 10110011. What check bits are sent by the sender?
b) Argue that if G has at least three terms (e.g. G = 1011 or G = 1110 or G = 1001100), then it can catch all errors E of the form x^i + x^(i-1) (i.e., two consecutive bits).
Question 7
========
a) Assume that packet size is 1KB, the round-trip-time is 25ms, and the bandwidth of the channel is 1200KB/s. What is the throughput of stop-and-wait?
b) What is the minimum number of logical channels in the concurrent-logical channels protocol that are needed to ensure we maximize the throughput
c) What if instead of using the number of logical channels you gave in b), we use twice that number. What will happen to the throughput? Briefly argue why.
Question 8
=======
Assume we have the cumulative acknowledgment protocol with unbounded sequence numbers. Assume the channel can lose and reorder messages (but not duplicate). Furthermore, assume that the SWS = 1, and that the timeout of the sender is accurate,. i.e., the last action of the sender is replaced by
when timeout frame#ch.sender.receiver = 0 and ack#ch.receiver.sender = 0 and LAR < LFS then
send frame(LAR+1) to receiver.
Under these conditions, is the sequence number in the acknowledgment really necessary? Briefly argue why yes or why no.
0 comments