Midterm Exam

Computer Networks

CSC522

Fall 2010

Answer any five questions from the following

Policy: Open book, open notes, take-home exam. Each answer should be self-designed.

Date: To be returned by midnight October 19, 2010 as a Word file at

1.  (a) You want to measure the approximate available bandwidth between your host and a distant station during a time when the packet-switched network is lightly loaded. Indicate how you would approach the task. (b) Suppose you have a choice of sending messages either by packet-switching or by circuit-switching. How would you decide?

2.  Suppose a malfunctioning hardware in a character oriented transmission system sets all bits transferred to zero. Does a parity bit catch the problem? Why or why not? Checksum computation is often implemented with conventional computational software, but most CRC computations are performed with special-purpose hardware. Why?

3.  Consider the possibility of implementing FEC between two directly communicating nodes using Hamming code system. Assume the codewords in this system to be 7 bits long with its 4 bits representing its message bits. How many bits of error can it detect and correct if one codeword at a time is sent across? The following are the received codewords. What are the corresponding corrected message bits realizable at the receiving station?

1.  1 0 0 1 0 1 0

2.  1 0 1 1 0 0 0

4.  A Traceroute (in Windows, it’d be Tracert in Command Prompt within Accessories) is a handy utility to learn partially about a network topology. Select a receiver on our continent, and perform a Traceroute checking three different times a day to

a.  Find the average and the round-trip delays between the source and the destination at each of the three settings.

b.  Find the number of routers in the path at each of the settings. Did the path change in any of the settings?

c.  Try to identify the number of ISP switches that the Traceroute packets pass through from source to destination. Addresses with similar prefix should be considered as parts of the same ISP switch. Do the largest delays recurs at the peering interfaces between adjacent ISPs?

d.  Repeat the above for a destination on another continent. Compare and contrast the intra- and the inter-continental data.

5.  Suppose that the average message size of a user is M bits long. Using bits of data per packet you are supposed to transmit your message as a train of packets over a distance of hops. However, each packet would require a header of size bits (. Line bandwidth is a constant bits/sec. and the propagation delay is negligible. What would be the optimal value of p?

6.  Compare and contrast a standard ALOHA with a slotted ALOHA. How would they compare with a reservation ALOHA if reservation is used to indicate who to transmit next? Assume a pure-Aloha set up with an offered load of G frames per second. On a collision, assume that it doubles its last back-off period length and assigns it to be its current back-off period length. On a first collision, the back-off period is B secs, and assume that after a successful transmission, it resets it next back-off period length to B secs. Given this what would be its average transmission delay?

7.  Consider a CSMS/CD system over a 1-km long cable rated at 10 Mbps where the propagation speed is our usual . Repeaters are not used in this system, and the data frames are 256 bits long including 32 bits of header, checksum, and other overhead. Transmitter must seize the cable first before attempting a transmission; we assume it takes altogether to do so. Similarly, the receiver must seize the cable to send its ACK signal, and it must take again time. The ACK is 32 bits long. What is the effective data rate in this case excluding overhead if we assume no collision in the system?

8.  True, or false?

a.  A user requests a Web page that consists of some text and three images. For this page, the client will send on request message and receive four response messages.

b.  Two distinct Web pages like www.mit.edu/research.html and www.mit.edu/students.html could be sent over the same persistent connection.

c.  With nonpersistent connections between browser and the origin server, it is possible for a single TCP segment to carry two distinct HTTP request messages.

d.  The Date: header in HTTP response message indicates when the object in the response was last modified.

e.  HTTP response messages never carry an empty message body.

9.  Consider distributing a file of bits to peers using a client-server architecture. Assume a model where the server can simultaneously transmit to multiple peers, transmitting to each peer at different rates, as long as the combined rate does not exceed .

a.  Suppose that . Specify a distribution scheme that has a distribution time of .

b.  Suppose that . Specify a distribution scheme that has a distribution time of .

c.  Conclude that the minimum distribution time is in general given by

10.  Consider a reliable data transfer protocol that uses only NAK. Suppose the sender sends data infrequently. Would a NAK only protocol be preferable to a protocol that uses ACKs? State your reasons. Now suppose the sender has a lot od data to send and the end-to-end connection experiences a few losses. In this second case, would a NAK only protocol be preferable to one with ACKs? Why? If errors are infrequent, how would NAK-centric protocol perform compared to with ACKs?