Question 1 (6 points) Compare circuit-switching and packet-switching networks based

on​ ​the​ ​following​ ​criteria:

(a)​ ​Reserving​ ​network​ ​resources​ ​ahead​ ​of​ ​data​ ​being​ ​sent:​ ​(2pts)

In ​circuit-switching ​networks, a physical circuit is established between the source and

the destination. As part of ​setting up the circuit, resources are allocated for the connection.
After​ ​the​ ​circuit​ ​is​ ​established,​ ​data​ ​can​ ​flow​ ​between​ ​the​ ​two​ ​end-points.
In p​ acket-switching networks, there is no circuit establishment and therefore no
resources are reserved. As such, data can flow without having to wait for a circuit to be
set​ ​up​.

(b)​ ​Utilizing​ ​network​ ​resources​ ​efficiently:​ ​(2pts)

In circuit switching networks, network resources may not be fully utilized, e.g., during “silent”
times during calls. Since resources are reserved on a per-call basis, they cannot be utilized by
other​ ​calls,​ ​even​ ​when​ ​they​ ​are​ ​idle.

In packets switching, resources are used on-demand, therefore, available resources will be
used​ ​to​ ​fulfill​ ​demand.

(c)​ ​Guaranteeing​ ​uniform​ ​network​ ​quality​ ​of​ ​service​ ​during​ ​the​ ​connection:​ ​(2pts)

Since they reserve resources on a per-call basis, circuit-switching networks can guarantee
uniform​ ​network​ ​quality​ ​of​ ​service​ ​for​ ​a​ ​connection,​ ​while​ ​packet-switching​ ​networks​ ​cannot.
Question​ ​2​ ​(6​ ​points)

(a) What are “loss-intolerant” applications? Give an example of a loss-intolerant Internet

application.​ ​(2pts)

Applications​ ​that​ ​don’t​ ​tolerate​ ​loss​ ​of​ ​data.​ ​Example;​ ​Email,​ ​File​ ​transfer,​ ​Web.

(b) What are “delay intolerant” applications? Give an example of a delay-intolerant

Internet​ ​application.​ ​(2pts)

Applications that don’t tolerate delay in data delivery. Example; Internet telephony, video
conferencing,​ ​...etc.

(c) Based on your comparative analysis in Question 1, does the Internet adequately
support​ ​loss-intolerant​ ​applications?​ ​And​ ​delay-intolerant​ ​applications?​ ​Explain.​ ​(2pts)

The Internet was designed to support loss-intolerant applications but was not designed to
support delay-intolerant applications properly. This is because, as discussed in Question 1, the
Internet,​ ​as​ ​a​ ​packet​ ​switching​ ​network,​ ​cannot​ ​provide​ ​delay​ ​bound​ ​guarantees.
Question 3 (6 points) Consider an Internet application running on two Internet hosts,
Host1 and Host2, which are communicating using a path through router R. Assume that
the average service time at R is 10 ms, packets are 2,000 bits long, and the propagation
delay​ ​between​ ​Host1​ ​and​ ​Host2​ ​is​ ​150ms.

(a) Based on the latency expression we covered in class, how is service time defined?
Write​ ​the​ ​mathematical​ ​expression​ ​for​ ​it​ ​explaining​ ​its​ ​terms.​ ​(2​ ​points)

The​ ​service​ ​delay​ ​is​ ​the​ ​sum​ ​of​ ​the​ ​delays​ ​for​ ​processing​ ​and​ ​queueing.
Processing:​ ​checking​ ​bit​ ​errors,​ ​determining​ ​output​ ​link
Queueing:​ ​waiting​ ​for​ ​output​ ​link​ ​availability

Total​ ​delay​ ​=​ ​processing​ ​+​ ​queue​ ​+​ ​transmission​ ​+​ ​propagation

Service​ ​time​ ​=​ ​processing​ ​+​ ​queue

(b) If the application running on Host1 and Host2 cannot tolerate latencies over 200ms,
what is the minimum capacity of the link connecting the two hosts? Show your work. (4
points)

10ms​ ​service​ ​+​ ​150ms​ ​propagation​ ​=​ ​160ms​ ​->​ ​40ms​ ​for​ ​transmission

Transmission​ ​of​ ​2000​ ​bits​ ​in​ ​40ms​ ​->​ ​2000/.04​ ​=​ ​50​ ​kb/s
Question 4 (4 points) Another way to measure end-to-end network latency is to use the
round-​ ​trip​ ​time​ ​(RTT).

(a)​ ​What​ ​is​ ​the​ ​RTT?​ ​Use​ ​a​ ​time​ ​diagram​ ​to​ ​illustrate​ ​your​ ​explanation.

The​ ​time​ ​it​ ​takes​ ​for​ ​data​ ​to​ ​travel​ ​from​ ​sender​ ​to​ ​receiver​ ​and​ ​then​ ​back​ ​to​ ​the​ ​sender.

(b)​ ​What​ ​is​ ​the​ ​one-way​ ​trip​ ​time?​ ​Illustrate​ ​your​ ​answer​ ​with​ ​a​ ​time​ ​diagram.

It is the the ​time takes for data to be transmitted across a network from sender to
receiver.
Question​ ​5​ ​(8​ ​points)​ ​List​ ​one​ ​advantage​ ​and​ ​one​ ​disadvantage​ ​of:

(a) Peer-to-peer model for networked applications (when compared to the client-server
model).​ ​Explain.​ ​(2​ ​points)
Advantage: self-scalability”, new peers bring new service capacity, as well as
new​ ​service​ ​demands
Disadvantage: Management and security, peers are unreliable and increase
security​ ​risks.

(b)​ ​Web​ ​caching.​ ​Explain.​ ​(2​ ​points)

Advantage: reduce response time for client request, save network- and server
provider​ ​resources.
Disadvantage: on cache miss longer response delay and potential for out-of-date
content.

(c)​ ​Layering.​ ​Explain.​ ​(2​ ​points)

Advantage: It helps designing, maintaining, extending complex systems. Provide
the ability to change the implementation of a service/function without affecting
other​ ​layers​ ​of​ ​the​ ​model.
Disadvantage:​ ​Additional​ ​overhead;​ ​potential​ ​for​ ​duplication​ ​of​ ​functions.

(d)​ ​DNS​ ​(compared​ ​to​ ​a​ ​centralized​ ​name​ ​service).​ ​Explain.​ ​(2​ ​points)
Advantage:​ ​Scalability​ ​and​ ​fault​ ​tolerance
Disadvantage:​ ​Complexity
Question​ ​6​ ​(14​ ​points)​ ​TCP
(a)​ ​Is​ ​TCP​ ​a​ ​full-duplex​ ​protocol?​ ​Explain.
Yes​ ​since,​ ​in​ ​the​ ​same​ ​connection,​ ​data​ ​can​ ​flow​ ​in​ ​both​ ​directions.

(b) When Host1 initiates a TCP connection to Host2, it sends a SYN segment to Host​2​.
As part of the SYN segment, it includes a sequence number SEQX. What is the
purpose​ ​of​ ​sequence​ ​number​ ​SEQX?
SEQX​ ​will​ ​be​ ​the​ ​first​ ​sequence​ ​number​ ​Host​ ​1​ ​will​ ​use​ ​when​ ​sending​ ​data​ ​to​ ​Host2.

(c) Suppose that the Host1 application that initiated the TCP connection has data to
send​ ​to​ ​Host2.​ ​Can​ ​Host1’s​ ​SYN​ ​segment​ ​carry​ ​the​ ​data?​ ​Explain.
No, Host1 needs to wait for the 3-way handshake to complete before sending data to
Host2.

(d) When Host2 replies to the SYN segment from Host2, it uses a SYN-ACK segment
and also includes a sequence number SEQY . What is the purpose of SEQY ? Is SEQX
=​ ​SEQY​ ​?​ ​Explain.

No, SEQX is different from SEQY. SEQY will be the first sequence number Host2 will
use​ ​when​ ​sending​ ​data​ ​to​ ​Host1.

(e) The original TCP protocol only retransmitted upon a timeout event. More modern
variants of TCP use the Fast Retransmit mechanism. Explain how Fast Retransmit
works​ ​and​ ​how​ ​it​ ​can​ ​improve​ ​TCP​ ​performance.
Fast retransmit is when TCP considers a packet lost after receiving a number (typically
3) of duplicate ACKs. This way TCP doesn’t have to wait for the timeout ​event to trigger a
retransmission. By retransmitting sooner, it can try to recover more quickly and avoid the
timeout.

(f) TCP can “pigyyback” acknowledgments on data segments. What field(s) in the TCP
segment​ ​header​ ​are​ ​used​ ​to​ ​do​ ​that?​ ​Explain.
TCP’s​ ​header​ ​acknowledgment​ ​field​ ​and​ ​the​ ​ACK​ ​flag.

(g)​ ​What​ ​is​ ​the​ ​main​ ​advantage​ ​of​ ​“piggybacked”​ ​acknowledgments?

It improves network resource usage by reducing the number of control packets that have
to​ ​be​ ​sent.
Question 7 (12 points) The reliable data transfer, or rdt, protocols we covered in class
provide​ ​reliable​ ​in-order​ ​data​ ​delivery​ ​with​ ​duplicate​ ​detection.

(a) What mechanism(s) used by rdt protocols cause duplicates to be generated?

Explain.​ ​(2​ ​points)

Duplicates are generated because of retransmissions. Consider a scenario in which a

segment was transmitted but the segment got delayed inside the network. It eventually
arrives at the receiver but in the meantime, the sender times out and retransmits the
segment. The receiver would then receive a duplicate and, in order to guarantee
duplicate-free​ ​delivery​ ​to​ ​the​ ​application,​ ​it​ ​needs​ ​to​ ​detect​ ​and​ ​get​ ​rid​ ​of​ ​duplicates.

(b)​ ​What​ ​mechanism(s)​ ​used​ ​by​ ​rdt​ ​protocols​ ​detect​ ​duplicates?​ ​Explain.​ ​(2​ ​points)

Sequence​ ​numbers.

(c) rdt protocols can recover from two types of network losses. What mechanism(s) are
used​ ​to​ ​recover​ ​from​ ​each​ ​type​ ​of​ ​loss?​ ​Explan.​ ​(3​ ​points)

Checksums​ ​for​ ​data​ ​corruption,​ ​and​ ​timers​ ​for​ ​packets​ ​dropped​ ​in​ ​transit.

(d)​ ​What​ ​is​ ​the​ ​retransmission​ ​timeout,​ ​also​ ​known​ ​as​ ​RTO?​ ​(2​ ​points)

It’s the amount of time the sender will wait for an acknowledgement for a transmitted
segment before considering that the segment was lost and starting the retransmission
of​ ​that​ ​segment.

(e) In class we discussed why retransmission timers are critical to the performance of
reliable​ ​data​ ​transfer.​ ​Why​ ​are​ ​they​ ​set​ ​as​ ​a​ ​function​ ​of​ ​the​ ​RTT?​ ​(3​ ​points)

Because the retransmission timer needs to reflect the network’s current conditions. The sender
shouldn’t wait needlessly to retransmit the segment if it was lost, but it also shouldn’t add
unnecessary duplicate traffic if the segment is simply taking longer to be delivered by the
network.
Question 8 (14 points) Two hosts X and Y are communicating over the Internet using
TCP.​ ​Suppose​ ​that​ ​the​ ​TCP​ ​connection​ ​has​ ​already​ ​been​ ​opened.

(a) Assume host X’s window size is 6 and segments 5001 through 5006 have already
been sent. Host X then receives an ACK with ACK sequence number set to 5003.
Segments 5007 through 5011 are waiting to be transmitted by host X. Draw the time
diagram showing this scenario including which segments, if any, X can transmit after
getting​ ​the​ ​ACK​ ​with​ ​ACK​ ​sequence​ ​number​ ​set​ ​to​ ​5003,​ ​and​ ​explain​ ​why.​ ​(4​ ​points)

When the sender receives the ACK for segment 5003, it means that the receiver received
segments up to 5003 (5001-5003). This is because TCP uses cumulative acknowledgment.
Because our window size is 6 and we have only 3 unacked segments, the sender will send
segments​ ​5007​ ​to​ ​5009.
(b) Some time later, X sends segments 5020 to 5025 (having received
acknowledgments for all previously sent segments). Host Y receives segments 5020,
5022, 5024, and 5025. Assuming that segments that were not received were lost, draw
the time diagram for this scenario, including the acknowledgments that host Y
generates assuming it uses a Go-Back-N based variant of TCP. What does host Y do
with​ ​segments​ ​received​ ​out​ ​of​ ​order​ ​and​ ​which​ ​are​ ​these​ ​segments?​ ​Explain.​ ​(4​ ​points)

Because GBN uses cumulative acknowledgments, Host Y will send 3 ACKs for segment 5020.
Only after the first ACK host X will send segment 5026, for the next 3 ACKs it won’t send any
new segments. Once a timeout event for segment 5021 is triggered, Host X will send all
segments​ ​that​ ​have​ ​been​ ​previously​ ​sent​ ​but​ ​not​ ​yet​ ​acknowledged​ ​(5021​ ​-​ ​5026).

Host​ ​Y​ ​will​ ​discard​ ​any​ ​segments​ ​that​ ​were​ ​received​ ​out​ ​of​ ​order​ ​(5022,​ ​5024,​ ​5025,​ ​5026).

(c)​ ​When​ ​X​ ​gets​ ​Y​ ​’s​ ​acknowledgments,​ ​what​ ​will​ ​it​ ​do?​ ​Explain.​ ​(3​ ​points)
Once host X receives the first ACK for packet 5020, it will compare the sequence
number for the ACK with the next expected sequence number(5020). If they match, host X will
restart the timer. Because, the next three ACKs (5020) don’t match the next expected sequence
number (5021), host X will not restart the timer which will lead to a timeout event and Host X will
retransmit​ ​segments​ ​5021-5026.
(d) Suppose the same situation described in (d), but both X and Y use a
Selective-Repeat variant of TCP. What acknowledgments will Y send to X? What does
Y do with out-of-order segments? Explain. And, what will X do when it gets Y ’s
acknowledgments?​ ​Explain.​ ​(3​ ​points)

Host Y will send ACK for segments 5020, 5022, 5024, and 5025. Host Y will buffer any
out​ ​of​ ​order​ ​segments.
Explanation: Because in SR, the receiver individually acknowledges correctly received
segments, Host Y will acknowledge the receipt of segments 5020, 5022, 5024, and
5025 even though they are received out of order. Host Y will buffer segments 5022,
5024,​ ​5025,​ ​and​ ​5026​ ​until​ ​it​ ​receives​ ​segments​ ​5021​ ​and​ ​5023.

Question 9 (20 points) You are studying for the CE 150/L midterm using one of the
computers at the UCSC Science Library. During one of your breaks, you want to check
out the Soccer4All.org Web site for the latest videos of the Soccer World Cup 2018
Qualifyers.

(a) What would be the steps needed before your computer at ucsc.edu can issue a
request for Soccer4All.org? Assume this is the first time anyone at ucsc.edu requests
contents​ ​from​ ​that​ ​Web​ ​site​ ​and​ ​name​ ​resolution​ ​is​ ​done​ ​iteratively.​ ​(4​ ​points)

Computer​ ​sends​ ​DNS​ ​request​ ​to​ ​local​ ​DNS​ ​name​ ​server​ ​at​ ​UCSC.
Local​ ​DNS​ ​name​ ​server​ ​sends​ ​DNS​ ​request​ ​to​ ​root​ ​name​ ​server.
Root​ ​name​ ​server​ ​responds​ ​with​ ​name​ ​server​ ​for​ ​.org​ ​domain.
Local​ ​DNS​ ​server​ ​sends​ ​DNS​ ​request​ ​to​ ​.org​ ​domain​ ​name​ ​server.
.org domain name server replies with the address for authoritative name server for
soccer4all.org.
Local​ ​DNS​ ​server​ ​requests​ ​address​ ​from​ ​Soccer4All.org​ ​authoritative​ ​name​ ​server.
Soccer4All.org​ ​ ​authoritative​ ​name​ ​server​ ​replies​ ​with​ ​Soccer4All.org​ ​Web​ ​server’s​ ​IP​ ​address.
Local​ ​DNS​ ​name​ ​server​ ​forwards​ ​the​ ​IP​ ​address​ ​to​ ​client.

(b) Immediately after you downloaded videos, another friend, Gabriela, who is also at
UCSC, wants to watch the videos as well. What are the steps needed before Gabriela’s
request​ ​to​ ​Soccer4All.org​ ​is​ ​issued?​ ​Explain.​ ​(4​ ​points)

Computer will request the address to UCSC’s local DNS name server which will reply with the
proper IP address without the need to query other servers because it should have the answer in
its​ ​cache.

(c) You finally manage to access the Soccer World Cup 2018 Qualifyers page from the
Soccer4All.org site. The page has 10 videos embedded in it. The processing/service
time within the network is 10ms each way, and the one-way propagation delay is
100ms. Assume that transmission delay is negligible. What would be the response time,
i.e., the time it takes between when your browser requests the page and when the
videos are delivered, if your browser uses non-persistent HTTP (with no parallel
connections)?​ ​Explain.​ ​(4​ ​points)

1 page with 10 videos means 11 objects. Non-persistent HTTP means that we will need 11
requests, each one with its own TCP connection. Each request will take (100 + 10)ms * 2 * 2,
that​ ​is,​ ​two​ ​RTTs​ ​of​ ​220ms​ ​each.​ ​So​ ​in​ ​total​ ​we​ ​will​ ​need​ ​4840ms.

(d) What would be the response time if the browser uses persistent HTTP? Explain. (4
points)
In this case we will only need one connection, so 1 + 11 RTTs, that is, one to establish the
connection​ ​and​ ​one​ ​for​ ​each​ ​objected​ ​requested,​ ​resulting​ ​in​ ​a​ ​total​ ​of​ ​2640ms.

(e) It is now Gabriela’s turn to download the videos. Assuming that the steps in (b) have
already been executed and that you have already been able to download the videos,
what is the average response time Gabriela will experience for each video in Seth’s
Web page if the ucsc.edu’s cache hit ratio is 60%? Suppose that the delay to access an
object​ ​within​ ​ucsc.edu’s​ ​network​ ​is​ ​20ms.​ ​Assume​ ​persistent​ ​HTTP​ ​is​ ​used.​ ​(4​ ​points)

Considering​ ​the​ ​cache,​ ​the​ ​time​ ​to​ ​get​ ​any​ ​object​ ​would​ ​be​ ​.6​ ​*​ ​20​ ​+​ ​.4​ ​(20​ ​+​ ​220)​ ​=​ ​108​ ​ms
Then​ ​we​ ​need​ ​220ms​ ​to​ ​start​ ​the​ ​connection,​ ​plus​ ​108ms​ ​for​ ​each​ ​object.​ ​Total​ ​is​ ​1408​ ​ms.