Transcript Southern Methodist University

Lecture 1
Advanced Networking CSE 8344
Southern Methodist University
Fall 2003
Mark E. Allen
Welcome!
• My contact info:
–
–
–
–
Mark E. Allen
[email protected]
972 747 1490 phone / messages
Email is a great way to reach me.
• Website
– engr.smu.edu/cse/8344
– Will contain syllabus, notes, important dates, etc.
Outline for Lecture 1
•
•
•
•
Preliminaries
Discuss syllabus
Course goals and outline for course
Get into the content
Intro (cont)
• Lecture format
– Power point slides
– Some written examples
– Please ask questions! (unless it’s a tape)
• NOTE:
– Next two lectures will be pre-taped
• Aug 29 10 AM (lecture 2)
• September 5 10 AM (lecture 3)
– Tapes will play at regular time also.
DISCUSS SYLABUS HERE
Motivation
• Purpose of networking: Sharing information
between people.
– Data is “information”
– Voice is “information”
– Evolution of networks
•
•
•
•
•
•
Teletype (Morse code) was low bit rate.
Voice (analog)
Video (analog television)
FAX
Dial-up Modems and DDS circuits
High-speed Internet
Network evolution
• Voice networks
– Analog voice
– Digital trunks
introduced.
– Digital switching
– Out of band
• SS7, AIN, etc.
– Wireless voice
– Voice over IP
• Data networks
– Mainframes connected with
SNA
– Ethernet, Token ring,
Novell IPX
– Ethernet wins out
– Internet
– WWW
– TCP/IP wins out
– GigE and Wireless Ethernet
catching on
Growth of traffic and internet
traffic
Traffic Growth: 1998 to 2002
3000
2500
2000
1500
1000
500
0
1998
1998=100
Source: RHK
1999
2000
Voice
Other data
2001
Internet
2002
Data and multimedia now dominate
traffic on the network
• Eventually the network of the future will
carry all types of service.
• IP looks to be the “convergence layer” of
the future.
• Voice, video and data communications will
eventually occur over a common network.
Motivation (cont)
• What are we really trying to get?
– 1) Convergence: Voice, data, video, etc. all on the same
user terminal
– 2) Low cost: If we can afford it, we’ll use it.
– 3) Mobility: We don’t want to be chained to a desk.
(wireless, and the internet all give us freedom to access
information wherever we are.)
– 4) High bandwidth: Lots of speed will enable new and
useful apps. Games, virtual reality, on demand movies,
video conferencing, etc)
– 5) Consumers want direct access into the data networks
(B to B, e-commerce, databases, etc.)
The requirements drive the
technology
• QoS
– Bandwidth, Delay, Jitter, etc.
•
•
•
•
Mobility
Cost
Power consumption
These things are all related.
– More bandwidth usually consumes more power
– Mobility requires low power.
– Etc. .. etc.
Functions of network elements
• Signaling / Addressing
– Allows the users to control how information flows through the
network (IP, dialed digits, etc.)
• Switching
– Devices necessary for steering information and signaling messages
around the network
• Multiplexing
– Allows several information “flows” to share the same medium
– We will discuss this in detail
Public vs. Enterprise Networks
• Millions of disparate
customers
• Distributed control
• Usage based billing
• 911 and public safety
concerns
• Lots of security concerns
• Legacy infrastructure
• FCC Issues
• Large geography
• 1 “customer”
• Centralized decision
making
• No billing issues
• Limited public safety
concerns
• Limited legacy concerns
• Fewer FCC / regulatory
issues.
• Smaller geography
The layered protocol approach
Applications
“layer 4”
“layer 3”
“layer 2”
“layer 1”
Transport
layer
Network
layer
Datalink
layer
Physical
layer
Ex: TCP, UDP ...
defines how data is transported
Ex: IP, IPX …
defines the logical structure of the network
Ex: ATM, Ethernet, Token ring…
defines how the media is accessed
Ex: 10base2, 10baseT, SONET…
defines the voltages and physical connectors
Limitations of OSI model
• The layered model
– Provided clear demarcation points for protocol
developers.
• Physical layer people needn’t be concerned with software.
– Was intended to be the roadmap for the “OSI” protocol
(never materialized)
• But…
– Often creates duplication of efforts (error correction,
restoration, management, etc.) $$$$
• More on this later
Defining dB
• dB is a convenient way of describing loss and
gain
• dB can be added where multiplication is
normally required
XdB = 10 log (X)
Note:
3dB = 10 log (2)
6dB = 10 log (4)
9dB = 10 log (8)
Ex)
3dB
6dB
G=2
G=4
Gtotal= (2)(4) = 8 = 9dB
Defining dBm
• dBm describes power
– YdBm= 10 log(Ymw)
• Ex, 15 mwatts = 11.8 dBm
30 mwatts = 14.8 dBm
(note: 2X the power is 3dBm more)
Types of transmission mediums
• Open copper pair
– Low attenuation (few hundredths of dB per km at Voice
frequencies)
– Takes up lots of space (not used much anymore)
• Paired wire
–
–
–
–
Many pairs in a bundle (up to a few thousand)
Can be buried or put on telephone poles
Higher attenuation than Open wire
Figure 1 shows the attenuation of copper pairs vs. frequency and
wire gauge
– Example: T1 (~1 MHz signal) experiences 30 dB of attenuation
over 1 mile on 22 gauge wire
– Common in buildings and LAN installs. 100BaseT runs on twisted
pair
Types of transmission mediums
(cont)
• Coaxial cable
– Good for higher bandwidth signals (several hundred
MHz) for several km
– Takes up much more space
– Digital 50ohm used to carry DS3 signals
– Analog 75ohm used to carry TV
• Fiber optic cable
– Extremely wide bandwidth (several THz)
– Low attenuation 0.25dB per km
Loops to the home
• Transmission from phone to CO occurs on a single
pair of wires
– A “hybrid” on either side of the two wire circuits (one
in the phone, on at the subscriber side of the switch)
– Implemented using specialized transformers
– Imperfections in the hybrid can cause echos (see figure
2)
• Loading coils
– Were installed extensively on long loops (3-15 mile)
– Reduces attenuation at VF (~3500 Hz) but sharp cutoff
at higher frequencies
– Big problem for DSL installations
Pair-gain systems
• Used to pack several subscribers onto a single
loop
– Acts time division multiplexor
– Commonly called “subscriber loop carriers”
– Present a problem when installing DSL
• Some pair gain systems use Concentration
– Acts as a statistical multiplexer
– The terminal that needs the line grabs from the
available pool. Some probability exists that the request
is blocked.
Multiplexing scheme
• FDM – Frequency Division Multiplexing
– Several analog VF signals are mixed using different
local oscillators
– A5 Channel bank multiplexor was used to mux 12
voice calls into a group
– See figure 3 for groups, supergroups, etc.
• This scheme worked well with analog Voice
channels and microwave transmission systems.
Transmission impairments (cont)
• Distortion
– Envelope delay refers to the delay seen by a particular frequency
– Loops impose non-uniform envelope delay
– Voice is not severely impacted but it’s a problem for modems
• Echos
–
–
–
–
Occur when there is reflection at the opposite end of the line
Normally causes by hybrid imbalance (2W to 4 Wire)
Attenuation in the circuit helps the problem
Not noticed in short circuits less than 1500 miles (10 msec of delay
per 1000 mile circuit) which experience 30 msec of delay (round
trip)
– Old echo suppressors used active impedence device in reverse path
– New echo “cans” are DSP based and use a variant of adaptive
filters.
Impairments (cont)
• Via Net Loss (VNL) is built-in loss proportional to
the length of the circuit.
– Combats ringing and echo
• Zero transmission level point (TLP)
– 0TLP : Reference point in a circuit into the first switch
– Measurements taken along the path are referenced back
to 0TLP
– (see examples)
Digital Signals
• Two Symbols: Binary Signaling
Symbol is a.k.a. Bit
• M Symbols: M-Ary Signaling
M is usually a power of 2
Log2M bits/symbol
• Baud rates same? Symbol shapes similar? If yes..
Bandwidth required is similar
M-Ary signaling allows increased bit rate
Symbols get closer together if Power fixed
Receiver detection errors more likely in
presence of noise
• Bandwidth, Bit Rate, SNR, and BER related
Example: Binary Signal
• Serial Bit Stream
(a.k.a. Random Binary Square Wave)
– One of two possible pulses is transmitted every T seconds.
Here the symbol is either a positive or negative going pulse.
– When two symbols are used, a symbol is known as a ‘bit’.
volts
+1
If T = .000001 seconds, then
this signal moves 1 Mbps.
0
time
-1
T
Example:M-Ary Signal
• One of M possible symbols is transmitted every
T seconds.
EX) 4-Ary signaling.
Note each symbol can represent 2 bits.
volts
+1.34
If T = .000001 seconds, then
this 1 MBaud signal moves
2 Mbps.
+.45
time
-.45
-1.34
T
M-Ary Signaling
• Bandwidth required
– Function of symbols/second & symbol shape
– The more rapidly changing is the symbol,
the more bandwidth it requires.
– An M-Ary signal with the same symbol rate and similar
symbol shape as a Binary signal has essentially the same
bandwidth.
• The previous two slides show...
– Equal Power & Equal Bandwidth Signals
– M-Ary signal transfers more bits/second BUT detection errors
more likely at the receiver
Wired Physical Links
• Untwisted Pair Cabling
– Highly susceptible to EM interference
– Bad choice for telecom systems
• Example: Speaker Wires, Power Lines
• Twisted Pair Cabling
– Fairly resistant to EM interference
– Bandwidth typically in 1-2 digit MHz
• Examples: LAN wiring, Home telephone cables
Twisted Pair Cables
Wired Physical Links
• Coaxial Cable
– Resistant to EM interference
– Bandwidth typically in 2-3 digit MHz
• Example: Cable TV
• Fiber Optic Cable
– Immune to EM interference
– Bandwidth in GHz to THz
Physical Layer Ailments...
• Attenuation
Signal power weakens with distance
• Distortion
Pulse shapes change with distance
– Copper cabling
High frequencies attenuate faster
Pulses smear
– Fiber cabling
Frequencies propagate at different speeds
Dispersion
Generating a Square Wave...
5 Hz
+
15 Hz
+
25 Hz
+
35 Hz
1.5
0
-1.5
0
1.0
cos2*pi*5t - (1/3)cos2*pi*15t
+ (1/5)cos2*pi*25t - (1/7)cos2*pi*35t)
Effects of Dispersion...
5 Hz
+
15 Hz
+
25 Hz
+
35 Hz
1.5
0
-1.5
0
1.0
cos2*pi*5t + (1/3)cos2*pi*15t
+ (1/5)cos2*pi*25t + (1/7)cos2*pi*35t)
In this example the 15 and 35 Hz signals have suffered a
phase shift (which can be caused as a result of different
propagation speeds) with respect to the 5 and 25 Hz
signals. The pulse shape changes significantly.
Receiver Detection
• SNR tends to worsen with distance
• Digital Receiver Symbol Detectors
– Examine received symbol intervals (T sec.)
– Decide which of M symbols was transmitted
– Single Sample Detectors
Sample each symbol once
Make decision based on sample value
– Matched Filter Detectors (Optimal)
Sample each symbol effectively an infinite
number of times
Make decision based on an average
SNR = (Average Signal Power)
Average Noise Power
4.5
z2 x
k k
0
4.5
0
0
20
40
60
k
Binary Signal
10 Bits showing
80
100
99
SNR = 100
4.5
z2 x 0
k k
4.5
0
0
20
40
60
k
Sequence = 0011010111
80
100
99
SNR = 10
4.5
z2 x 0
k k
4.5
0
0
20
40
60
80
k
Signal a sequence +1 and -1 volt pulses
100
99
SNR = 1
4.5
z2 x 0
k k
4.5
0
0
20
40
60
k
80
100
99
SNR = .1
8.5
z2 x 0
k k
8.5
0
0
20
40
60
k
80
100
99
Single Sample Detector: SNR = 1
Threshold is placed midway between nominal Logic 1 and 0 values.
4.5
0
4.5
0
0
20
40
60
k
80
100
99
Detected sequence = 0011010111 at the receiver,
but there were some near misses.
Matched Filter Detector: SNR = 1
Orange Bars are average voltage over that symbol interval.
4.5
0
4.5
0
0
20
40
60
80
k
Averages are less likely to be wrong.
100
99
Channel Capacity (C)
• C = W*Log2(1 + SNR) bps
– W = channel bandwidth (Hz)
– SNR = channel signal-to-noise ratio
• Maximum bit rate that can be reliably shoved down a
connection
• EX) Analog Modem (30 dB SNR)
C = 3500 *Log2(1 + 1000) = 34,885 bps
• EX) 6 MHz TV RF Channel (42 dB SNR)
C = 6,000,000 *Log2(1 + 15,849) = 83.71 Mbps
Why not just keep amplifying to
counter-act attenuation?
• Amplifiers add noise as they boost the
power.
– For analog signals, this degrades the signal to
noise ratio (SNR)
– With digital signals, the SNR (Eb/No) is
degraded until the system takes errors
• Low noise and multistage amplifiers are
used to combat this problem.
Amplifiers in series
SNRin
SNRout
G1
F1
G2
F2
( F2  1)
Feff  F1 
G1
• A pre-amp with a low noise figure reduces the
overall noise figure while providing high gain
– For example, using this equation, the effective noise
figure of a preamp with a gain of 20dB and noise figure
of 3 dB followed by an amplifier of gain 30dB and
noise figure 9 dB would be 2+(8-1)/100 = 2.07 or 3.16
dB. The total gain would be 50 dB. (make sure to use
non-dB numbers in the equations)
Overview of telephony
• Telecommunication networks were
originally designed for voice
– Analog signals of 4 kHz bandwidth
– Sampled at 8 kHz with 8 bits of quantizing
levels for 64kbps circuit
• Digital TDM multiplexing has been the
technique of choice since the late 70’s.
Multiplexing Formats
• Why multiplex??
– Combine several lower bandwidth signals onto a single
faster channel
– Saves running thousands of individual wires
– Allows single carrier for several signals
• Frequency division multiplexing
– Multiple frequencies “stacked”
– This was done in the analog days.
– Each voice channel was mixed up to higher center
frequency.
– Doesn’t lend itself to digital technology
– Requires very (spectrally) flat channels
FDM Hierarchy
With digital T-Carriers, this is now obsolete
From: Digital Telephony Bellamy, chapter 1
FDM (cont)
From: Digital Telephony Bellamy, chapter 1
Time division multiplexing
• With TDM, every low speed signal gets a fixed
amount of bandwidth in the high speed signal
Low speed
High speed
Current North American TDM
multiplexing scheme
T1 1.544
Mbps
24
DS0
64kbps
44.736
Mbps
Channel
Bank
28
2.5
Gbps
M13
48
10
Gbps
OC48
4
OC192
What is a T1?
•
•
•
•
Four-wire circuit: one pair for transmit, one for receive.
Full-duplex: information goes both ways all the time.
Digital: transports binary data or voice
TDM: Capable of transporting 24 digitized voice
channels.
• Pulse code format: Individual voice channels are normally
digitized using PCM.
• Framed synchronous transmission: samples from each
voice channel (8 bits) are taken and sent sequentially. A
frame bit is added resulting in 193 bit frames.
Analog to PCM conversion
Understanding SONET/SDH, Kartalopoulos
Analog to DSO with CODEC
Understanding SONET/SDH, Kartalopoulos
DS0 to DS1 MUX (channel bank)
Understanding SONET/SDH, Kartalopoulos
European muxing standard
Understanding SONET/SDH, Kartalopoulos
T1(cont)
• Note: each voice channel is sampled at 8000 hz, so a T1
must send 193 bits in 125 usec, or 1.544 Mbps.
• Line coding for T1 is bipolar, AMI. Every other 1 is
represented by a positive pulse. 0’s are represented by no
pulse.
– This gives the signal zero average value.
– The required bandwidth is 770 kHz.
• T1 uses byte synchronous transmission
– Receiving end needs to distinguish which channels are which and
which 8 bits make up the bytes.
– Framing bits make this work
Evolution of T1 muxes
• T1 to DS0 mux commonly called “channel banks”
• D1 channel bank format
– Robbed bit signaling
– LSB of each byte was used to carry signaling for voice
channels (no significant degradation)
– If data was carried, 56k was max amount
• D2 – D3 channel banks were small improvements
• D4 channel bank was common and introduced the
“superframe” format.
T1 Superframe
• Superframe concept: 12 frames together became a
“superframe” framing bits occur in a special
pattern.
– Allows only 6th and 12th frames to give up signaling
bits.
– The 24 signaling bits in the 6th frame are called the A
bits, 24 in the 12th frame are called the B bits.
– B8ZS was introduced to solve the “ones density rule”
when carrying data: In every 24 bits, there must be at
least 3 pulses and no more than 15 consecutive zeros.
• Recall that 8 zeros are substituted with a pattern then a BPV is
generated.
D5 Extended superframe (ESF)
• There was a desire to improve the D4 SF format:
– More performance monitoring capabilities
• Customer equipment didn’t pass through Bipolar Violation
information
– 8kbps F bits are redefined to create a 2kbps framing
sequence, 4kbps data link channel, and 2kbps CRC
channel.
• Data link channel used to control end equipment
• 24 rather than 12 frames are now used, framing bits occur in
frames 6, 12, 18 and 24. These bits now become the A,B,C,D
bits for each channel.
Uses for T1
• T1 was primarily used for voice trunks
between voice switches.
– Now many customers use to connect PBX
– May buy T1 to the internet
• Connect a Router to ISP’s router
• Channel Service Unit (CSU) is often used
Central
Customer premise
office
T1
CSU
Router
Internet
Other common TDM rates
• E1: European version of the DS1, contains 32
voice channels
• DS3: multiplexed version of 28 DS1’s
– DS3 uses “bit-interleaving” rather than byteinterleaving in DS1 format.
– DS3 is 44.736 Mbps
– B3ZS is used to maintain 1’s density.
– Multiplexing is done in two stages:
• 4 DS1 signals are muxed using pulse stuffing synchronization
to form a DS2
• 7 DS2 are muxed using a fixed pulse stuffing synchronization
to form a DS3 signal.
DS3 muxing (cont)
• In DS1 to DS2 muxing, stuffing is used to get rate
to 1.545796 Mbps (with overhead).
– Each DS1 can be running at a different speed ranging
from 1.540 to 1.545 Mbps
– Bit interleaving and bit stuffing allow all the DS1s to
maintain their independent rates.
– 48 bits are collected to make a block, 6 blocks to make
a subframe, 4 subframes to make DS2 frame.
(subframes are not individual DS1s)”
• See figures 1 and 2 in TTC “Fundamentals of
DS3”
DS3 muxing (cont)
• For DS2 to DS3 muxing, 84 bits make a block, 8
blocks make a subframe, 7 subframes make a
frame.
– Bit stuffing may again be used incase DS2s are not at
the same rates.
• Note that C-bits are used in both stages 1>2 and
2>3 to control bit stuffing.
– If all DS2s are in sync, this is not necessary for second
stage
– C-bit format uses C bits in DS2>DS3 stage for other
purposes.
Drawbacks to DS3 based
transport
• Difficult to drop individual voice or DS1 channels
– Bit muxing and bit stuffing makes channels locations
unknown
• Not fast enough for some applications
– No standard Proprietary muxing schemes were
developed to transport 12 or 24 DS3 on a fiber link.
– Interconnects between carriers at rates higher than 45
Mbps were complex
• Required them to purchase same type of fiber muxes.
• Runs on Coax vs. unshielded twisted pair (UTP)
or fiber.
Sonet is defined to address these
problems
• Standardized rates beyond 10Gbps on fiber
• “Synchronous” for improved add/drop
capability
• Standards for both Electrical and Optical
interfaces.
• Backward compatible with T1, DS3, E1,
E3, and other “async” standards.
Traditional multiplexing
hierarchy
DS0
64 kbps
T1 1.544
Mbps
24
44.736
Mbps
Channel
Bank
28
2.5
Gbps
M13
48
10
Gbps
OC48
4
OC192
Traditional hierarchy
Signal
Bit Rate
Channels
DS0
64 kbps
1 DS0
DS1
1.544 Mbps 24 DS0s
DS2
6.312 Mbps 96 DS0s
DS3
44.736
Mbps
28 DS1s
SONET
• SONET byte oriented frame format
– Path, line, and section
– Multiplexing format
• Virtual Containers
• Synchronous payload envelope (SPE)
• Pointers
– Timing issues
– This is what makes SONET synchronous -- the payload can float
in the SONET frame.
• Overhead
– Line, Section, and Path
• Performance monitoring
SONET Hierarchy
Signal
Bit rate
Capacity
STS-1, OC-1
51.840 Mbps
28 DS1s, 1 DS3
STS-3, OC-3
155.520 Mbps
84 DS1s, 3 DS3s
STS-12, OC-12
622.080 Mbps
336 DS1s, 12 DS3s
STS-48, OC-48
2488.320 Mbps
1344 DS1s, 48 DS3s
STS-192, OC-192
9953.280 Mbps
5376 DS1s, 192 DS3s
SONET Network
LINE
SECTION
Terminal
(LTE)
REG
ADM
or
DCS
(LTE)
REG
REG
Terminal
(LTE)
Router
(PTE)
Router
(PTE)
PATH
SONET Networking
810 bytes x 8000 frame/sec x 8 bits = 51,840,000 bps
OH
PAYLOAD
9 rows
OH
PAYLOAD
OH
STS-1
Synchronous
Payload
Envelope
90 columns (87 columns of payload)
3 columns of
transport overhead:
Section overhead
Line overhead
Path overhead
PAYLOAD
BIP-8/BI: Parity Checking
9 Rows
STS-1 Frame Section Trace/Growth
A1
A2
JO/Z0
J1
B1
E1
F1
B3
D1
D2
D3
C2
H1
H2
H3
G1
B2
K1
K2
F2
D4
D5
D6
H4
D7
D8
D9
Z3
D10
D11
D12
Z4
E2
Z5
S/Z1 M0/M1
87 Columns of Payload
STS-1 Synchronous
Payload Envelope
(STS-1 SPE)
BIP-8/B2: Error Monitoring
Z2
90 Columns