Integrated networking
Columbia University
ELE6905, Spring 2004
Thursdays 10:00-12:30, Mudd Building rm. 545
Class #8
March 11, 2004
Instructor
Stephen Weinstein (Dept. of Elect. Eng.), sbw@cttcgroup.com
Office hours: Thursdays 1:30-3, or by appointment.
Lecture notes and references posting:
http://www.cvn.columbia.edu/courses/Spring2004/ELENE6905.html
Today, class #6
Wired Access Networks
xDSL, cable data, PON
But first, a bit more discussion of token/leaky bucket
traffic smoothing
- Token bucket imposes an average rate constraint, leaky
bucket a peak rate constraint. Both offer some control
of the duration of a high-speed burst from the source.
- Can be used to police a DiffServ Service Level Specification
(part of an SLA) between two network domains, specifying how
traffic crossing the boundary of the two domains is to be treated.
In particular, how each other’s traffic is conditioned (limited and
smoothed) at the boundary, and how packets can be labeled or
relabeled for different treatments.
Token bucket (limits average rate)
Periodic token deposit, Rav tokens/sec
(independent of packet arrivals)
Bucket capacity C tokens
One token withdrawn to serve a packet whenever a packet
is present
Server
queue
Shaped output traffic
Arriving traffic
rate r packets/sec
Leaky bucket (limits peak rate)
One token deposited each time a packet arrives
Bucket capacity C tokens
Tokens "leak" out at periodic rate Rp
Server
queue
Arriving traffic
rate r packets/sec
Shaped output traffic
Burst length analysis of token bucket
Bucket size C.
Tokens deposited in bucket at constant rate Rav.
Token removed from bucket each time a packet is served.
Packets arrive at a rate r packets/sec.
Assume token bucket initially full (C) after a period of little activity.
High-speed burst of packets arrives at rate r > Rav and is immediately
serviced, passing through at rate r, until the bucket empties (after
which service is at Rav rate).
Tb = maximum burst length time until the bucket empties.
C + RavTb = no. of tokens already in bucket or added over Tb.
rTb = ni, of tokens withdrawn over Tb.
0 = C + RavTb - rTb, or
Tb = C/(r-Rav)
Digital access networks
The communications facilities used between local/personal networks
and metropolitan/core networks.
Access
networks
PON
T-carrierCellular mobile
services
Cable data
Wireless MAN
xDSL
(and ISDN)
UWB
(Ultra Wideband)
IEEE 802.16
LEOS, direct satellite
Powerline communications
Where is access networking heading?
The generic goal: An IP-oriented optical/electrical
and optical/optical convergence architecture
Optical metropolitan
and core networks
Ethernet frames?
IP services
overlay
optical fiber
Optical node
Access network
(optical, wireless, coax, twisted pairit doesn't matter much)
Subscriber
End-to-end Ethernet: A new communications
paradigm, or just a fad?
Ethernet bridge
Ethernet switch
Optical metropolitan
and core networks
Access
network
Ethernet
switch
Ethernet
bridge
OXC
Ethernet switch
Access
network
Optical framer
Ethernet
bridge
(IEEE802.1D)
Ethernet switch
IP stack
LAN
IP on Ethernet
Ethernet
frames
Ethernet card
Why are there many more cable data than ADSL
subscribers in the U.S.?
-Cable systems 15 years ago began major upgrades to
HFC (hybrid fiber-coax) to reduce maintenance costs
and improve transmission performance, before digital
services came along. They were better prepared for
digital services.
Headend
Runs of up to 20 amplifiers contributed
to noise, distortion, frequent failures
-Cable systems have monopolies, but local telephone
companies must lease facilities to competitors,
discouraging major capital investment in access by
telcos.
-Cable operators began with a broadband services
perspective, while telcos were associated with (and
thought like) telephone companies.
-ISDN was too little, too late.
Is any access system free from congestion
problems?
Basically no. All access systems have capacity
bottlenecks because no carrier wishes to overinvest
in excess capacity. Most capacity bottlenecks can
be relieved with additional investment in facilities.
xDSL (Digital Subscriber Line)
- Uses the telephone subscriber twisted pair, bypassing the voice
switches with their 4KHz channel filters.
- Supports (in some versions) normal analog telephone in addition
to data services.
- Comes in various symmetric and asymmetric versions (next page).
- Performance, especially maximum dependable data rate, is
dependent on distance from Central Office (or fiber node) and on
crosstalk between twisted pairs in the same bundle.
Digital Subscriber Line Types (24-26AWG twisted pair)
(POTS: Plain Old (analog) Telephone Service)
Name
Rates down/up
ISDN (Basic Rate) 144kbps/144kbps
ADSL (Asymmetric 1.5-8 Mbps/128-640Kbps
Digital Subscriber Line)
G.lite (ITU-T G.992.2) 0.78-4Mbps/<512Kbps
"splitterless"
SHDSL (or just HDSL) 0.768Mbps each way
(Symmetric High Speed DSL)
G.shdsl (ITU-T G.91.2) 0.144-2.32Mbps each way
POTS Distance
no
18Kft
yes
6.5-24Kft
yes
<24Kf
no
<18Kft
no
12-20Kft
"Multi-rate and extended reach"
[http://www.cisco.com/warp/public/cc/so/neso/dsso/global/shdsl_wp.htm]
VDSL
26-55Mbps/2Mbps
(Very High Speed Digital Subscriber Line)
no
~1Kft
ADSL standard: T1.413-1998, "Network to Customer Installation Interfaces Asymmetric Digital Subscriber Line (ADSL) Metallic Interface: November, 1998"
available for $350 at www.atis.org/atis/docstore/doc_display.asp?ID=159
ADSL can operate much faster than a telephone
modem because the data signal bypasses the
voiceband filters of telephone switches
a) Dialup Modem
PSTN
voice carrier
system
ISP
PSTN
switch
3KHz voiceband filter
Line concentrator
(the bottleneck)
b) ADSL
Digital Subscriber Line Access Multiplexer
voice carrier Line
concentrator
system
(voice bottleneck)
PSTN
ADSL subscriber
termination
Voiceband filters
Data bottleneck
trunk(s)
Router
...
DSLAM
Data network(s)
ISP
ADSL modem with data passband filter
Typically an ATM switch
Commercial ref. for a DSLAM: http://www.adtran.com/static/docs/DOC000965.pdf
"G.lite"
(G.992.2)
Subscriber ADSL modem plugs into computer, just like a
voiceband modem. So-called "Splitterless" system. Lower rate,
smaller BW (512KHz upper edge).
ADSL subscriber
termination
To DSLAM
Convenient installation, but unreliable inside wiring
usually degrades performance
Commercial example: www.hellosoft.com/products/hdsl/hdsl.htm
ADSL modulation formats
- Primary standard: DMT (discrete multitone)
Generate with Discrete Fourier Transform, as described previously.
Different data rates and amounts of transmitted power can be allocated
to different subbands, in accordance with transmission rate requirement
and quality of channel at different frequencies.
sinc(f) spectrum (magnitude shown here) has nulls at carrier
frequencies of adjacent subbands, eliminating interband
interference (in a system with no channel distortion).
Upstream
4KHz ~20KHz
(POTS)
138KHz
1.1MHz
May use QAM signal constellations of different sizes
in different subbands.
DMT/OFDM performance costs to maintain
orthogonality of subbands
- Overhead from cyclic extension (extending blocks beyond multipath
dispersion)
- Virtual (unused) subbands on the edges of the total band to avoid
interference with other bands.
138KHz
1.1MHz
Filtered MultiTone (FMT)
A variation on DMT applied in VDSL. Mitigates DMT/OFDM
performance costs.
"With FMT, orthogonality between subchannels is ensured by using
non-overlapping spectral characteristics instead of overlapping
sinc(f) type spectra. Since the linear transmission medium does
not destroy orthogonality achieved in this manner, cyclic prefixing is
not needed".
Bandpass versions
of a standard low-pass filter
Parallel
data
IDFT
Line signal
Parallel/serial
Ref: I. Berenguer & I. Wassell, "FMT Modulation: Receiver filter bank definition for
the derivation of an efficient implementation"
www-lce.eng.cam.ac.uk/~ib226/papers/fmt_modulation.pdf
- Secondary standard: CAP (Carrierless Amplitude-Phase Modulation)
Essentially the same as QAM (quadrature amplitude modulation),
generated by direct inband digital techniques.
Passband signal
samples
Information stream Store of inband
D/A
digitally modulated
signal sequences
By proper design, can generate
signals for an infinite variety of
information streams from a finite
set of stored inband digital segments
Line
signal
ADSL modem
KHz
26
138
ADSL modulator/filter
Subscriber Hybrid
line
coupler
Telephone
filter
KHz
0.3 3.3
Splitter
Pulse
shaper
Data from
computer
Ethernet
Receiving
filter
Demodulation/
equalization/
detection
MHz
1.1
.138
RJ-11
jack
Coder/
decoder
ADSL protocol stack
ISP (Internet
Exchange office
DSLAM
Data
TCP/UDP network Protocol services
such as DHCP
terminal
IP
IP
IP
PPP
ATM
ATM
ATM
(optional)
Service
Provider)
(optional)
SONET
(optional)
SONET ADSL
Subscriber
ADSL
terminal
Ethernet
ATM
(optional)
ADSL
Computer
(or appliance)
TCP/UDP
IP
PPP
Ethernet
The future of xDSL?
-Existing full-length subscriber line ADSL doesn’t work (at
any attractive rate) on a significant percentage (30%?) of
subscriber lines.
VHDSL will develop as fiber nodes come closer to
subscribers.
-Requires capital investment in metropolitan data
networking to alleviate congestion as subscriber population
grows.
Telcos want exclusive right to offer xDSL on their subscriber
lines; enthusiastic deployment and performance advances
depends on that. [They object to unbundling rules.]
-Unbundling rules
(carriers required to "unbundle" their facilities and lease the
pieces to service competitors at reasonable rates)
Telcos may be holding back on broadband access
(especially VHDSL) until their monopoly is comparable
with cable's.
Central Office
Telco ADSL
Telco-built subscriber lines
Co-located competitor ADSL?
Cable Data System
DOCSIS: Data-Over-Cable Service Interface Specifications: Radio
Frequency Interface Specification SP-RFIv1.1-I06-001215, Cable
Laboratories, December 15, 2000
Main DOCSIS Technical Specifications
Modulation Bandwidth (MHz) Data Rate(Mbps)
Down 64 or 256 QAM
6
27 or 36
Up
QPSK or 16-QAM
0.2-3.2
0.32-10
In both directions: MPEG-2 framing, Reed-Solomon forward error
correction coding, DES encryption.
Upstream Medium Access Control: Packet-based, contention and
reservation slots, QoS capabilities.
Management: SNMP, with MIB definitions.
Residential network interface: 10BT Ethernet (USB and IEEE 1394
planned).
Business network interfaces: 10/100BaseT, ATM, FDDI
Spectrum utilization (within the cable)
Upstream
(0.2-3.2MHz channels in usable parts of this noisy spectrum)
Downstream
................................
5 42 50
6MHz channelization
MHz
860
Each downstream channel can be used for:
-One analog TV signal, or
-Six-seven 4Mbps MPEG-2 digital TV signals, or
-One 19Mbps HDTV signal plus two digital TV signals, or
-Data (e.g. Internet downloads) at 30Mbps.
64-QAM or 256-QAM used downstream for high spectral
efficiency.
To PSTN Telco return access
concentrator
(TRAC)
Switch or
network adaptor
Digital satellite
Local server
programming
facility
Backbone network
Remote
server
facility
Cable Headend
Analog headend term
Cable Modem Network termination
Termination
R ... R T ... T
System (CMTS)
Operations
Splitter
Support
Security &
Access
Controller Single-modeO/E
E/O
T ... T
Coaxial cable
distribution network
T T T ... T
Combiner (mux)
Analog signal modulators
(6MHz channels)
optical fibers
125-500
OC3-OC12 Fiber node subscri(O/E, E/O)
Analog
satellite
programming
bers
QAM receiver MPEG,
control functions
Set-top box
Digital or
Analog TV
Ethernet
Contention in shared
coaxial cable tree
Cable modem:
QAM receiver,
QPSK transmitter
Contention for (upstream) bandwidth in the
DOCSIS MAC (medium access control)
Resources are allocated as "minislots" of upstream
transmission time.
Client
sources
Requests
Alloc.
Mediator
Medium
Mapping MAC frame into minislots
In the upstream direction, transmission time is slotted into minislots
for TDMA (time division multiple access). The time duration of a
packet transmission is a power of two multiple of 6.25s minislot
increments. If grant isn’t sufficient, MAC frame is fragmented and
part is sent later.
fragmentation
Example MAC frame
Minislots
Grant: 4 x 6.25μs
6.25μs 6.25μs
Used by others
6.25μs
Later grant for remainder of request
Minislot allocation for upstream traffic
(Cable Modem Termination System)
Medium Access Protocol management message
CMTS
Requests
Slots previously
mapped
Information elements assign slots to different
modems
Cable modem
transmit opportunity
Request
contention
area
Slots not yet
mapped
Maintenance
Request contention resolved through backoff algorithm.
The future of cable data and the comparison
with xDSL?
-Cable operators (at least in U.S.) are advanced in broadband
digital services (entertainment as well as Internet access) and
with further innovations may continue in lead.
-Congestion on shared coaxial cable tree can be resolved by
splitting the fiber node (making two nodes, each of which
takes half the lower distribution tree).
(This requires capital investment!)
Fiber node
(O/E, E/O)
400 subscribers
Fiber node
(O/E, E/O)
Fiber node
(O/E, E/O)
200 subs
(more)
200 subs
-Modern cable plants can offer high-speed data service to
almost all of their customers, unlike ADSL that can only
be offered to about 70%.
-Telcos still ahead in quality of plant engineering and
maintenance.
-Telcos may catch up and pass cable operators when and if
the investment cable operators made in fiber nodes is
matched by telco investment in fiber nodes (to implement
VHDSL).
-Both systems have capacity bottlenecks, but cable's may
be more expensive to resolve.
-xDSL and cable data likely converge to a common "fiber to
the neighborhood" architecture with a variety of "last mile"
transmission media.
PON (Passive Optical Network)
A passive splitter in the field reduces cost and makes
fiber to the home/business more practical. May replace
T-carrier access systems and services.
Telco serving
office
Metropolitan/
core networks
CO
terminal
ATM or (newer)
Ethernet based
transmissions
Passive splitter
Up to 64 broadcast drops
ONU
PON interfaces
DS-1
DS-3
CO
OC-3
terminal OC-12
IP router
ATM switch
ONU
DS-1
DS-3
POTS
PBX
Enet
Layers 2/3 switching & routing
Data provisioning in 64kbps increments
up to 1 Gbps
Dedicated wavelength
Tutorial ref: www.iec.org/online/tutorials/epon/topic04.html?Next.x=37&Next.y=14
EPON (Ethernet PON)
Objectives:
- Point-to-multipoint, low-cost architecture using single-mode fiber.
- Access network distances (at least 10km)
- Standard Ethernet frames, no contention
- Standard 1 Gigabit Ethernet rate
- Minimum 1:16 split
- Replacement for earlier ATM PON
Ref (brief tutorial): www.ieee802.org/3/efm/public/jul01/tutorial/pesavento_1_0701.pdf
EPON (Ethernet PON)
Optical network unit
Downstream:
Headend (e.g. CO terminal)
1 3 1 2
OLT
Users
1 1
ONU1
1
1 3 1 2 ONU2 2
2
1 3 1 2
splitter
Optical line
terminal
1 3 1 2
ONU3 3
3
All packets broadcast to all ONUs, where Ethernet frames for particular
users are separated on the basis of the MAC (medium access control)
addresses.
Ethernet packet
Header
Type, or length of data field (2 bytes)
Checksum (2 bytes)
Data field (up to 1500 bytes for 10Mbps)
Preamble
(7 bytes)
Source address (2 or 6 bytes)
Destination address (2 or 6 bytes)
Start of frame delimiter (1 byte)
Pad (0-46 bytes)
6 bytes (each hexadecimal pair is one byte)
Example of MAC address: 00-50-DA-CE-E2-76
EPON
Users
Upstream:
11
ONU1
1 1
OLT
1 1 2 3 3 3
2
ONU2
splitter
Optical line
terminal
3
3
3
ONU3
2
1
2
3 3 3
3
Synchronized system does upstream time slicing, so there are no collisions
and no need for packet fragmentation.
"MAC uses existing PAUSE control frame or other control messages."
EPON optical aspects
OLT
Medium T
WDM
access
R
logic
λ1
1:N optical
splitter
ONU1
ONU2
λ2
ONU3
WDM
T Medium
Data
access
R logic
Full duplex operation using separate wavelengths.
(example: 1550nm/1310nm)
Headend permits only one subscriber at a time to transmit.
End users see only traffic from headend, not from each other.
Review for the midterm exam
Analog to digital conversion
Sampling theorem:
For x(t) bandlimited to (-W, W) Hz,
x(t) = x(n/2W)sin[2W(t-n/2W)]/[2W(t-n/2W)]
n
3T 4T
Interpolation function (what is its
frequency spectrum?)
time (sec)
5T
T 2T
T= 1/2W
Reconstruction formula above realized in a low-pass filter limited to
what frequencies?
What are advantages of digitized media?
Does a digitized media stream necessarily use less bandwidth than the
original analog media signal?
Infrastructure network types
Metropolitan
Local area
networks Access area networks
networks (MANs)
(LANs)
Core (or long haul)
networks
PON
Optical Core Network
Cellular
mobile
T-carrier
(DWDM)
services
SONET ring,
Cable
data
Wireless LAN
RPR
Resilient Packet Ring
(IEEE 802.11)
Subscriber
1-10Gbps
line
Ethernet
Satellite
IEEE 802.16
UWB
transport,
LEOS, direct satellite
broadcasting
IR
Wireless local loop
("wireless MAN")
Bluetooth
Switched
Ethernet
Personal Area Networks
Protocol layers offering services to higher layers
and peer-to-peer interaction across networks
Application Information unit transfer
Application
request transport
Transport
Transport data package transfer
Transport
request packet forwarding
Network
Packet transfer
Network
request link access & physical commun.
Physical network
Link, Phys
Link, Phys
Virtual Circuits, e.g. in ATM
CBR
VBR
ABR
Pools of
capacity for
different services
CBR
VBR
ABR
guaranteed peak capacity
guaranteed average capacity
whatever is left over
CBR: Continuous bit rate
VBR: Variable bit rate
ABR: Available bit rate
UBR: unrestricted bit rate, best effort service
Virtual Private Network
Encrypted end-end packet
Enterprise
network
segment
(or individual
remote host)
Encapsulating "tunnel" packet
addressed to firewall
Public Internet
(or any data network)
Firewall
router
Enterprise
network
segment
Destin.
host
VoIP (general concepts of Internet-PSTN interworking)
IP telephone (H.323 compliant)
Mike
Spkr
codec
Telephony
Application
Port zz
Peer-to-peer
IP communication
IP tel.
Addr. Dir.
buf
buf
RTP
Signaling (e.g. SIP)
Media
UDP/IP/
Internet gateway
Phys
controller
IP address xxx.xx.xxx.xx
Echo
canceler
PSTN
PSTN
telephone
Tel number (212) 854-xxxx
H.248
Internet/PSTN
Gateway
Spectrum policy (e.g. renting spectrum in bandwidth
and time) and applications of software-defined radio
- Multiple air interfaces.
- Agility to move between available time/bandwidth slots
rather than stick to fixed assignments.
Frequency
Time
Infrastructure and ad-hoc modes
Infrastructure mode
Access
Point
Access
Point
Ad-hoc mode
(peer to peer relay)
Backbone network,
wired or wireless
Network interoperability at PHY/MAC levels vs.
network interoperability at network level (IP)
Challenges for PHY/MAC interoperability:
- Matching different protocol implementations of different
operators
- Matching rate offerings in the digital hierarchy
- Coordinating operations and management functions
(protection/restoration, traffic engineering, comparable
service features and performance, ...)
At network level: How does IP avoid these complexity
problems? What kinds of agreements are still needed?
Network vulnerabilities
- Physical damage and congestion.
- Congestion.
- Cyber attacks.
- Cascading failures.
Everyday restoration mechanisms, such as SONET ring, and breaks
that cannot be managed by everyday restoration mechanisms.
X
Damaged conduit
Normal
path
Restoration
path
Add/drop multiplexer
Packet networking advantages
e.g. efficient use of capacity from statistical multiplexing
of different sources (What is the difference from a circuit-switched
network?) and resilience (why?)
silent
packetizer
talking
silent
packetizer
packetizer
Aggregated traffic
Scheduling
algorithm
Router functions
Lookup
table
Destination
address
IP packet
Routing
Algorithm
(what is
OSPF?)
to output
port 1, top
priority
(what kinds of
service disciplines?)
Link states from
neighboring nodes
Transfer
Input port 1
Classifier
to output
port m,
top priority
Scheduler
Server
State
Queue 1
Queue n
Output
port 1
Scheduler
Input port m
Transfer
Classifier
Server
State
Queue 1
to output
Queue n
port m,
lowest priority
Output
port m
IP protocol stacks
HTML
SIP
Application Information unit transfer
JAVA, CORBA, RTP, RTP
SOAP
Session
(stream IDs)
Client-server requests/responses
HTTP
Transport
TCP, UDP
Transport data package transfer
(port numbers)
Network
IP, DiffServ, SLA, traffic
smoothing & policing
MPLS
Multi-Protocol Label Switching
Packet transfer
(IP addresses)
Link or MAC
Link-level framing, circuit switching,
medium contention resolution
Physical
Plug interfaces, modulation
Physical network
Successive encapsulation
UDP protocol unit
IP packet
IP header
Application-level
information unit (e.g. media AIU
header
encapsulation)
Ethernet frame
UDP header
IPv4 address classes A,B,C – How the 32 bits
are used
CLASS A
01
8
Network
0 ID
Bit number
16
24
Host ID
31
Fraction of all
addresses
1/8
For the relatively small number of networks supporting a very
large number of hosts computers.
Can address up to 27 = 128 networks, each with up to
224 = 16,777,216 host computers.
What is the address depletion problem addressed by IPv6, and what
are the major differences between IPv4 and IPv6? How can address
translation mitigate the address depletion problem?
TCP and UDP
Provide an end to end (application to application) transport
service.
TCP (transport control protocol): Reliable, connectionoriented service.
UDP (user datagram protocol): Unreliable datagram
service (but no delays for retransmissions!)
What are the major similarities and differences between TCP and
UDP? What is sliding window flow control in TCP?
RTP Audio and Video Sessions
Application
RTP1
RTP2
UDP
IP
Physical
Session 1 (audio)
port x
Session 2 (video) port y
Internet
Control messages
Stream Data
Application
RTP1
RTP2
UDP
IP
Physical
Internet QoS with DiffServ, SLA, MPLS
Client
network
DSCP (Differentiated Services CodePoint) marking
(for class of traffic).
Possible traffic smoothing.
Traffic conditioning part of SLA regulates
volume of submitted traffic for each DSCP
Traffic-engineered
links using MPLS
ISP ingress
router
DiffServ PHBs invoked
by DSCP markings
Policing of
submitted
traffic
Direction of traffic
Client
network
ISP network with
DiffServ-capable routers
DiffServ Classes
EF (expedited forwarding): Specified PIR (peak
information rate) and bounded delay (e.g. for
voice). The EF PHB implements a pass-through
service with no queueing delays or delay jitter .
AF1-AF4 (assured forwarding): Preferred forwarding
at routers to minimize packet losses. Each level
has a specified CIR (committed information rate)
and PIR.
BE (best effort): Packets utilize whatever capacity is
left after preferred classes are accommodated.
What is a DiffServ codepoint? What is an SLA? What is "color
marking"? What are token and leaky buckets?
Traffic conditioning in a network router
DiffServ class marker
(if not already marked)
Packets
EF
profile
Conformance marker
conforming
To PHB
Meter/ Non-conpassed scheduler
shaper forming dropper
dropped
profile
conforming
To PHB
Meter/ Non-conpassed scheduler
dropshaper forming
per
dropped
Other AF classes
Best effort
AF1
classifier
Meter measures conformance with SLA
Shaper smooths traffic to conform with SLA
What are IntServ and DiffServ?
No. of service classes
Session state maintained
in network?
Advance resource
reservation?
IntServ
2
Yes
Yes (e.g. by
RSVP)
DiffServ
6 (incl AF1-4)
No (only preferred
service discipline)
No (SLA by customer
and class)
Where would they tend to be used in networks?
What are the IntServ and DiffServ classes?
How can IntServ and DiffServ work together (at least be interfaced)
MPLS - what is it and what is it good for?
Paths for forwarding equivalence classes
NY to SF
Chicago All other traffic
Denver
New York
San
Francisco
St. Louis
NY to SF
Expedited Forwarding traffic
What is "label swapping"?
Connectionless and Connection-Oriented
Datagrams
Connection-oriented
stream
What is a virtual circuit?
What is the difference between a switch and a router?
Cell switching (ATM)
-"Asynchronous" because cells need not arrive at a switch
at fixed times.
-Resource reservation setup in advance through signaling.
-Offers quality of service (bandwidth/delay guarantees)
in broadband networks.
Why a 53-byte cell?
What is VCI swapping?
What are the service types?
What is a Virtual Path?
What is an adaptation layer?
Synchronous, asynchronous, isochronous,
plesiochronous
Synchronous: Data stream synchronized to a clock
Data stream 1 (e.g. a tributary as on last slide)
Data stream 2
What are TDM and TDMA?
time
Optical networks: Metropolitan/core network integration
DWDM core network
PSTN access and
metropolitan network
ADM
PBX, LAN
SONET
ring
Framing (SONET
and/or OTN)
Terabit
router
Ethernet
LAN
λ2
OXC1
OXC2
(Transp.)
OXC3
(Opaque)
λ1
O/E E/O
λ1
λ1
λ2
λ3
λ2
OXC4
(Opaque)
High-speed Ethernet switch
(Metropolitan Area 1Gbps/10Gbps)
Lightpath shown with two red
links and one green link
OTN: Optical Transport Network, ITU-T G.709
Lightpaths
What are WDM and DWDM?
What are opaque and transparent optical switching?
What is hierarchical optical switching?
optical
fibers
One fiber
Repeater
cross-connect
Routing and wavelength assignment through an optical core
network continues to be a research topic
SONET framing
(155.52Mbps SONET/SDH frame)
Virtual container (including path overhead)
t=0 Transport overhead
Payload
First byte
(section and line)
in frame
Pointer to
start of
virtual
container
........................
........................
........................
........................
........................
........................
9 rows
........................
........................
........................
270 columns
What is GFP?
Last byte
in frame
t=125μs
(entire frame)
Fiber access systems
What are FTTC, FTTH, PON? What is VDSL?
Telco serving
office
CO
terminal
SONET
Broadband network
(OC-3 to OC-48)
Example product for either
Fiber with subcarrier-multiplexed
FTTH or FTTC:
40-870MHz band downstream VDSL signal
on 1310nm and 1550nm
ONU
carriers, carrying "a full
(fiber node)
complement of analog
and digital signals"
VDSL on twisted pair, < 1000ft
[http://www.synchronous.net]
Up to 50Mbps (asymmetric)
TV
What is modulation?
Baseband/passband, PCM and PAM (M-level), Fourier transform
(spectrum), spectral efficiency, PSK, QAM, CDMA, DMT/OFDM,
signal constellations, noise immunity, UWB
Baseband information
signal
signal
levels
encoder
Carrier waveform
modulator
Modulated transmission
signal
Next week (March 18): Spring Break
Next class (March 25)
Midterm exam