Networking - University of Bath

Download Report

Transcript Networking - University of Bath

Networking
• A few questions on the course
• A few questions about the course
The Chancellor of the University is
Alistair Darling
Lord Patten
Glynis Breakwell
Prince Philip
Lord Tugendhat
Lord Hinton of
Bankside
33%
19%
19%
15%
11%
Lo
lis
ta
ir
D
ar
lin
g
rd
G
Pa
ly
ni
tte
s
n
Br
ea
kw
Pr
el
in
l
c
e
Lo
Lo
Ph
rd
rd
ili
Tu
p
H
in
ge
to
nd
n
ha
of
t
Ba
nk
si
de
4%
A
A.
B.
C.
D.
E.
F.
That was the “warmup”: E
A.
B.
C.
D.
E.
F.
Darling is Chancellor of the Exchequer
Lord Patten is Chancellor of Oxford
Glynis Breakwell is the Vice-Chancellor
Prince Philip is Chancellor of Cambridge
Lord Tugendhat is the Chancellor
Lord Hinton was the first Chancellor: the
mullets (stars) on the unicorn and wyvern
supporting the coat of arms are a
heraldic pun on his own arms
Questions on the course
(mostly factual, but there is more
to the course than facts)
How long is an Ethernet address?
Variable length
16 bytes
4 bytes
6 bytes
72%
20%
8%
ria
s
6
by
te
s
by
te
4
es
by
t
16
bl
e
le
ng
th
0%
Va
A.
B.
C.
D.
The answer is D (6 bytes)
If you didn’t get that, you’re still pretty
confused: After last time, and the previous
one, all I can suggest is that you really reread sections 2.2 and 3.2 of the book.
Which of the following will DHCP
not tell you
1. Your IP address
2. A router’s IP address
3. A nameserver’s IP
address
4. A nameserver’s name
5. The subnet mask
48%
15%
A
k
bn
et
m
as
m
e
na
r’s
e
su
er
ve
Th
na
m
es
r’s
IP
a.
..
dr
es
s
11% 11%
er
ve
ad
m
es
na
A
ro
ut
A
Yo
ur
er
’s
IP
IP
ad
dr
es
s
15%
D — the nameserver’s name
• It’s actually not useful
• And anyway you can find it out: how?
The reset (RST) bit is set in the last
packet of a 4-way close
1. True
2. False
83%
ls
e
Fa
Tr
ue
17%
False
• RST is not a normal part of TCP operation:
it is there to indicate some kind of error/
abnormal circumstance
After the 3-way open, MSS is
48%
38%
10%
t..
..
en
t
di
ffe
r
of
e
Po
ss
ib
ly
er
ag
in
tt
w
ha
w
of
Th
e
e
av
m
N
in
im
um
eg
ot
ia
te
d
ha
...
3%
Th
A. Negotiated
B. The minimum of
what the two said
C. The average of
what the two said
D. Possibly different
in the two
directions
D — possibly different
There is no negotiation, and it is perfectly
possible for MSS to be different in both
directions (e.g. one end is on a dial-up
link)
Consider the following TCP Open
scenario:
• P→Q: SYN (with MSS)
• Q→P: SYN (with MSS, and ACK)
• P→lost: ACK
What happens next?
P’s ACKof Q’s SYN is lost, and
The connection hangs
P eventually times out
P starts sending data
P resends its SYN
Q resends its SYN
P sends a RST
65%
15%
12%
P
R
a
se
n
ds
its
s
ST
SY
N
N
SY
nd
re
se
Q
en
d
s
its
da
P
re
s
di
ng
se
n
rts
lly
st
a
P
ta
t
ou
s
tim
es
ha
ng
n
nt
ua
tio
P
ev
e
nn
ec
6%
3%
0%
co
e
Th
A.
B.
C.
D.
E.
F.
C (or possibly E)
A. Hanging would be a failure of TCP
B. P thinks the connection’s open, so why time
out?
C. Therefore can send data (which will
acknowledge the SYN as well)
D. Why should P send SYN: it’s been acked
E. Q hasn’t has SYN acked, so will resend if C
doesn’t happen (e.g. SMTP, where the server
first sends a 220)
F. Why should it send RST?
Which of these is not in the TCP
header
Sequence Number
Options
Receiver Window
Congestion Window
Urgent Pointer
54%
20%
14%
9%
on
C
in
te
r
ow
ow
U
rg
e
nt
Po
W
in
d
n
ge
s
tio
ei
ve
R
ec
en
c
qu
rW
in
d
pt
io
n
O
e
N
um
be
r
s
3%
Se
A.
B.
C.
D.
E.
D — Congestion Window
A.
B.
C.
D.
Sequence number is always there
Options, if present are in the header
Receiver Window is always there
Congestion Window is purely maintained
by the sender, and is not passed to the
other end (its effects may be felt, but
that’s another matter)
E. Urgent pointer is always there
A receiver drops packets to indicate
that its window is full
1. True
2. False
72%
ls
e
Fa
Tr
ue
28%
False
It may drop, and not acknowledge, the data
(as in the persist timer examples), but it
does always respond to the packet, even
if the acknowledgement field is not
incremented.
Network congestion can be
indicated by
A.
B.
C.
D.
36%
21%
15%
12%
6%
6%
in
d
4
AC
2
W
m
e
ou
t
du
o
p
w
Ks
lic
=0
at
of
e
th
A
e
CK
sa
s
m
e
...
A
an
d
A
D
,B
an
d
C
B
an
d
C
3%
Ti
Time out
Window =0
2 duplicate ACKs
4 ACKs of the
same number
E. A and D
F. A,B and C
G. B and C
E (A and D) is right
A. Timeout is a good indicator of congestion
(but the other end might have crashed
…)
B. Not network congestion
C. Not 2 Duplicate ACKs
D. 4, with the same number, means 3 are
duplicates, so this is congestion
The persist timer deals with which
problem?
31%
A. Hosts that go down
B. Hosts that have no
space for more data
C. Lost SYN packets
D. Lost window adverts
E. Lost FIN packets
22%
Lo
st
FI
N
pa
ck
er
ts
ad
v
Lo
st
w
in
d
ow
pa
s.
..
SY
N
no
Lo
st
ve
ha
th
at
ts
os
H
ck
e
n
do
w
go
th
at
ts
os
H
et
s
11%
ts
11%
25%
D: lost window adverts
A.
B.
C.
D.
E.
Hosts that go down: TCP times out
sending
Hosts with no space — there’s no cure
for this!
Lost SYN packets, retransmission deals
with this
The response to the persist timer will
give us a new window size (which may
still be same, if there is still no space)
Lost FIN packets, retransmission deals
with this
“slow start” is somewhat of a
misnomer
1. True
2. False
53%
ls
e
Fa
Tr
ue
47%
Possibly
It does indicate that we start slow (so far, so
good), but we get exponentially faster, so
most people probably wouldn’t call it slow
Sending `window scale’ means
A. I will send scaled
windows
B. I will understand
scaled windows
C. I will send scaled
windows if you will
D. You can send scaled
windows if you want
E. B and C
F. B and D
68%
al
ed
sc
ill
Iw
Iw
un
d
d
se
n
ill
Iw
8%
5%
5%
8%
w
er
i..
s
.
ill
ta
nd
se
nd
sc
Yo
sc
a.
u
..
al
ca
ed
n
w
se
i..
nd
.
sc
al
ed
...
B
an
d
C
B
an
d
D
5%
E is the best answer
A. You can’t scale unless the other end
says it understands scaling
B. If you can send them, you must be able
to understand them
C. Certainly correct
D. Some TCPs send window scale
automatically even if the other end didn’t
(of course, they can’t use it)
In FTP, the control channel
A. Starts in ASCII or
binary depending
on the options
B. Starts in ASCII,
but can be
switched to binary
C. Is always in ASCII
D. Uses XDR
32%
35%
26%
U
se
s
XD
R
AS
CI
I
in
s
w
ay
al
Is
in
ts
St
ar
St
ar
ts
in
AS
C
AS
C
II,
II
o
rb
bu
tc
in
a.
..
.
6%
C – always ASCII
A. The other end doesn’t know what the
options are, so this can’t be right (a
general point about protocols!)
B. The IMAGE command changes the data
channel only
C. Right
D. XDR is connected with NFS, not FTP
One of these statements is true
A. SMTP constructs the
header from the
envelope
B. The envelope determines
who gets the mail
C. The header determines
who gets the mail
D. The header MUST agree
with the envelope on this
E. The header should agree
with the envelope on this
36%
31%
14%
17%
er
s
ad
he
Th
e
he
e
Th
a.
..
d
ho
ul
ST
ag
...
..
ad
ad
he
e
Th
er
M
U
et
er
m
in
er
d
de
e
ve
en
e
Th
SM
TP
co
lo
p
ns
tru
c
ts
th
...
te
rm
...
3%
B: envelope→sender
A. If anything, the sending agent (not SMTP)
constructs the envelope from the header
B. Correct
C. False
D. Not an RFC requirement: The header might
well not agree
E. It is impossible for C’s machine to check
whether C occurs on a mailing list
administered by B, or even whether it is a
mailing list
MIME is
1. A feature of e-mail
only
2. A feature of the Web
only
3. What Marcel
Marceau did
4. A general-purpose
encoding, first
invented for mail
86%
11%
e.
..
...
ge
A
ha
t
ne
M
ar
ce
se
ra
l-p
ur
po
lM
ar
ce
au
...
W
eb
th
e
W
at
ur
e
fe
A
A
fe
at
ur
e
of
of
em
ai
lo
nl
y
0% 3%
D — general purpose
• It did start out in the mail community, but
there is nothing really mail-specific about it
• (Marcel Marceau did Mime: different
capitalisation!)
NFS can be told to handle both
binary and ASCII files
70%
Fa
ls
e
30%
Tr
ue
1. True
2. False
False
• NFS files are fundamentally binary, and
the application has to deal with any
differences (e.g. extra CR/LF in ASCII
from Windows)
Now that we have NFS and the
Web, FTP is obsolete
1. True
2. False
94%
ls
e
Fa
Tr
ue
6%
False
• It’s less common to type the ftp
command, but we see (or at least use)
ftp:// URLs.
• FTP is fundamentally different from NFS
— FTP creates copies, whereas NFS
allows multiple computers to access the
same file.
UDP is used in many
circumstances when TCP could be
used instead
Fa
ls
e
28%
Tr
ue
1. True
2. False
72%
True
• NFS often uses UDP, where TCP would be
as good, possibly better
• DNS uses UDP (which is simpler), and
TCP could be used, though probably less
efficient.
Therefore the internet could be run
without UDP
1. True
2. False
62%
ls
e
Fa
Tr
ue
38%
False
• RIP and BOOTP/DHCP both rely on
broadcast (or multicast for some RIP v2
installations) and therefore can’t use TCP
• Many uses of multicasting (internet radio
etc.) would be impossible, which would
cause bandwidth/load problems as well
It is impossible to build reliability on
top of UDP
1. True
2. False
69%
ls
e
Fa
Tr
ue
31%
False
• TFTP manages to do it.
• TCP builds it on top of IP, and UDP is (by
definition) no worse than IP, so clearly it’s
possible (which doesn’t mean it’s
desirable)
Web hosting firms need HTTP 1.0
(or later) because
A. It allows binary
B. It allows MIME
C. It allows the complete
URL in GET
D. It can send the length
of the data file
47%
29%
21%
gt
h
le
n
th
e
d
th
e
It
ca
n
se
n
s
al
lo
w
It
...
.
m
pl
e.
co
s
al
lo
w
It
It
al
lo
w
s
bi
na
ry
M
IM
E
3%
C — complete URL
A. All HTTP allow binary (in the data)
B. MIME was in 0.9, and anyway isn’t very
relevant
C. Without the complete URL, we wouldn’t
know what page we were called on
D. Size is useful, but not necessary
Which of these is not part of the
NFS complete suite
XDR
Portmapper
TLS
Remote
procedure call
22%
ca
S
R
re
em
ot
e
Po
pr
oc
e
du
rtm
TL
ap
pe
R
ll
8%
r
3%
XD
1.
2.
3.
4.
68%
C — TLS
A. eXternal Data Representation is vital:
times, lengths etc.
B. Portmapper to find RPC programs
C. TLS — Transport Layer Security. Not an
intrinsic part.
D. Remote Procedure Call — the underlying
paradigm
Recall the ISO 7-layer model
1.
2.
3.
4.
5.
6.
7.
Physical
Link
Network
Transport
Session
Presentation
Application
Which layer is MIME?
Physical
Link
Network
Transport
Session
Presentation
Application
56%
14% 14%
14%
3%
ta
tio
A
n
pp
lic
at
io
n
en
Pr
es
Se
ss
io
n
or
t
sp
Tr
an
N
et
w
or
k
nk
0%
Li
ys
ic
a
l
0%
Ph
1.
2.
3.
4.
5.
6.
7.
6 — Presentation
• Describes how the data is meant ot be
interpreted at the other end.
Networking
• A few questions on the course
• A few questions about the course and the
subject
There’s too much on the basics
and not enough on the applications
30%
27%
24%
Strongly Agree
Agree
Neutral
Disagree
Strongly Disagree
14%
St
ro
ng
ag
re
e
ly
D
Di
s
is
a
gr
ee
tra
l
eu
gr
ee
A
N
ly
Ag
re
e
5%
St
ro
ng
1.
2.
3.
4.
5.
I have bought/been given a copy of
the book
o
50%
N
s
50%
Ye
1. Yes
2. No