1.
(a)
(b)
(c)
COMP 343-443: Computer Networks
Key to Midterm
Write the common and distinguishing features of
Go-Back-N protocol
Selective Repeat, and
TCP
(20 points)
Solution:
Common Features: All of them use pipelining, provide congestion control and
flow control because they all are Transport Layer, reliable protocols. In addition,
they use time-out to detect lost packets and use sequence numbers and checksum
to detect duplicate packets and errors respectively. Thirdly, the sender always
buffers outgoing segments.
Go-Back-N: (1) Sender uses one timer (2) Receiver sends cumulative
acknowledgment, (3) After timeout, sender resends all segments starting from the
last segment that was correctly acknowledged. (3) Receiver discards out-of-order
segments, (4) Receiver does not buffer.
Selective Repeat: (1) Sender uses multiple timers, (2) Receiver acknowledges
each segment individually, (3) Sender resends segment for which timeout occurs,
(4) Receiver buffers out-of-order segments.
TCP: (1) A hybrid between GBN and SR. (2) Implementations can have buffering
at the receiver side or not, (3) Has several special features not present in either
GBN or SR such as Fast Transmit, and Delayed Acknowledgment, (4)
Cumulative acknowledgement. A key difference between TCP and GBN is that
TCP sender will resend only segment n if the timer for it expires, whereas GBN
will resend segments n, n+1, etc.
2.
It is common for an institutional network to be based on a high-speed
LAN. Suppose that a router in this network is connected to a router in the public
Internet by means of a 3 Mbps access link. Suppose further that the LAN runs at
12 Mbps. Users at the institutional network are accessing the web servers in the
Internet at the rate of 30 requests per second and the average request size is 100
Kbits.
(a) Compute the traffic intensities at LAN and the access link
and comment on the values.
(b) Suppose you add an institutional web cache whose hit rate
is 0.5, how are the traffic intensities affected?
(c) Assuming an Internet delay of 2 seconds (round-trip time
between access link router and an outside webserver on an
average), and 20 millisecond-delay for requests that can be
satisfied by the web cache, compute the average delay for a
cache hit rate of h, where 0<h<1.
(a)
Solution:
Traffic intensity at LAN is:
(30 requests/sec) * (100 Kbits/request)/(12 Mbps) = 0.25
Traffic intensity at access link is:
(30 requests/sec) * (100 Kbps/request)/(3Mbps)=1.0
Traffic intensity at LAN is acceptable. However, at the access link it is too high
and will cause a serious degradation in performance.
(b) Cache hit rate of 0.5 means, that on an average, 50% of the requests will be
served by the LAN and the remaining 50% will go out on the access link. The
LAN continues to generate 30 requests/second and hence will incur a traffic
intensity of 0.25 as before. However, only 50% of 30 requests/second, namely 15
requests/second, will go out on the access link. Hence traffic intensity at access
link will be equal to 0.5
(c)
h*(0.020seconds)+((1-h)*(2.02seconds)=2.02-(2*h)seconds.
(15 points)
3.
Consider sending a large file of F bits from Host A to Host B. There are
m links (and m-1 switches) between A and B. Neglect propagation delays and
queueing delays. Host A segments the file into segments of S bits each. Assume
that F/S is an integer. Host A adds h bits of header to each segment forming
packets of size h+S each. Each link has a transmission rate of R bps. Find the
end-to-end delay in sending the file.
(10 points)
Solution:
Size of Packet + header = h+S.
Time of transmission for Packet 1 on Link 1 = (h+S)/R seconds.
Time of transmission for Packet 1 for m links = m(h+S)/R seconds. (a)
After Packet 1 arrives at the destination, every subsequent (h+S)/R seconds, each
other packet will arrive.
There are (F/S)-1 such packets.
The transmission time for these packets = ((h+S)/R)*((F/S)-1) seconds. (b)
Hence the total end-to-end delay is the sum of (a) and (b) values.
It is equal to ((h+S)/R)(m – 1 +F/S) seconds.
4.
We know from my lecture in class, that streaming audio and video can and
is usually done using UDP sockets. Consider webpages that have embedded
streaming audio/video displays. In this case, you know that webpages use HTTP
whose underlying protocol is TCP; but the inlaid streaming multimedia uses
UDP? Could you discuss the situation, analyze and offer your thoughts on what
may really be going on?
(10 points)
Solution: Will be discussed in class.
5.
Can you think of a useful application in any domain that could use the
internet, use TCP or UDP sockets and not already known publicly? Try your best
to be creative.
(10 points)
Solution: How about this?
A universal appointment system: All of us schedule appointments for various things such as
dental visits, annual checkups, hair salons, spa, Carpet Cleaning, for meeting with friends,
etc. etc. Now here is how things should be: You go to one website, your information/profile
is there. You pick the place for appointment (say Mathangi R. Sekharan, MD on Davis St.
Evanston) and peruse their system as to when appointments are available and pick the one
that is most suitable to you among the ones available. At the end of the session, two things
should happen: your calendar wherever the heck it is, (your PDA or Yahoo!Calendar) and
the appointment or calendar system at the Service Location should get updated. Cool, huh?