Lecture 21 Last Mile Technologies Outline Modems Example

Carnegie Mellon
Carnegie Mellon
Outline
Lecture 21
Last Mile Technologies

Khaled Harras
School of Computer Science
Carnegie Mellon University


Classic view: different types of wires
 Copper: telephone, modem, xDSL
 Cable: TV driven
 Fiber: future proofing  (Wireless: satellite and terrestrial)
Media encoding and streaming
Triple play: IPTV, cable, fiber to the home
15‐441 Computer Networks
Based on slides from previous 441 lectures
Carnegie Mellon
Many types of access
Carnegie Mellon
Modems
SAT

Modem offers a bit stream.

SLIP: Serial Line IP.
 Aggressive signal processing has dramatically increased the available throughput ‐ beats the Nyquist limit!
Home
Network
 Protocol to sent IP packets with minimum framing
 Lacks authentication, error detection, non‐IP support, ..
Broadcast

 Better framing, error control, and testing support
 Can negotiate choice of higher layer protocols, IP address
 Can support unreliable and reliable transmission
Access
Network
Distribution
Network
Cable
Copper or F.O.
Copper
PPP: Point‐to‐Point Packets.
Long
haul
Modem
Bank
Home
PC
Telephone
Network
F.O.
Carnegie Mellon
Example: Modem Rates
Carnegie Mellon
Integrated Services Digital Network (ISDN)
M odem rate
100000


10000
ISDN integrates voice and data services.
Provides a set of bit pipes that can be used for voice, data, signaling.
 Implemented by using time multiplexing

1000
Home
PC
100
1975
Basic rate ISDN offers to 64Kbs data bit pipes and one 16 Kbs signaling channel.
1980
1985
1990
1995
2000
Phone
Year
LAN
Public
Branch
Exchange
Telephone
Network
ISDN
Exchange
1
Carnegie Mellon
Digital Subscriber Line
Carnegie Mellon
DSL: Physical Layer Matters


Squeeze more bandwidth out of the telephone line using advanced signal processing.
(Symmetric) digital subscriber line (DSL).

Asymmetric digital subscriber line (ADSL).

 Carry analog voice signal in 0 – 4 KHz
 1‐pair of voice‐grade unshielded twisted pair (UTP)
 Ends of wiring were conditioned for optimizing low frequencies –
 Same bandwidth both ways, e.g. 768 Kbs
cuts off higher frequencies

 More “download” bandwidth, e.g. video on demand or web surfing
 Initially: T1 incoming path, 64 Kbs outgoing path – now much higher ADSL
Subscription
Unit
Phone
Telephone
Network
Changes needed for higher frequencies





bandwidths
Home
PC
Telephone wiring was design to carry a telephone signal
ADSL
Network
Unit
Change conditioning at the end points
Better coding and modulations
Bandwidth depends on the distance
Can upgrade to better wiring, e.g., 2‐pair DG UTP
Data uses higher frequencies on the wire
Carnegie Mellon
ADSL2+
16 – 25 Mb/s Down video
capabilities.
Up to 800 Kb up
Dist 1.5 km/BW
Higher-optional
Video/ voice/
data/copper
8 Mb/s down
800 Kb/s up
Higher-optional
Voice/data
/copper
Up to 8 Mb/s down Full use existing
copper. Web
Up to 1.5 Mb/s up
brows/Voice
2.7- 5.4 km
Video/ voice/
data (V/V/D)
over copper
Up to 52 Mb/s
down asymm
Up to 26 Mb/s
Symm.
Broadcast
video, VoD,
internet TV,
1.5 km - 300m
Voice/data/
Video
1/2 pairs
192 K to 4.6 M
Steps 8/16 Kb/s
Poss amplif.
V/V/D Services
mainly for
business appl.
Voice data and
video
1/2/3 pairs
784 Kbit/s to
2320 Kbit/s
V/V/D Services
mainly for
business appl.
G.992.5
ADSL2
With and without
splitter – G.992.3 / 4
ADSL
With and without
splitter – G.992.1 / 2
VDSL
Symmetric
Asymmetric G.993.1
SHDSL
Symmetric
G.991.2
HDSL
Symmetric G.991.1
Carnegie Mellon
DSL Speeds
Video/ voice/
data/copper
video capab.
Dist + 200m
ADSL
(Copper)
Carnegie Mellon
Cable Modem

Carnegie Mellon
Fiber – “FTTX”

Traditionally, fiber only used in network “core”

Reach of fiber has expanded over time, i.e., fiber reaches closer to the consumer
Use cable infrastructure for data service.
 More expensive technology – higher capacity
 Inherently has more bandwidth

The last mile is a shared infrastructure that was designed for broadcasting.
 Meaning: the bandwidth is shared by users
 Example: 27 Mbs shared incoming path; 768 Kbs common outgoing path





Fiber
Infrastructure
Fiber to the cabinet
Fiber to the curb Fiber to the home
Options include “active” and “passive”
Trend applies to all copper technologies
 Cable, twisted pair
2
Carnegie Mellon
Optical Access Network Architectures
Carnegie Mellon
Cable versus Fiber
OLT: Optical Line Terminator
ONU: Optical Network Unit

FTTH
O
N
U
Cable Modem Network
 Simplex 6 MHz downstream channels
 Simplex 200 KHz to 6+ MHz upstream channels
 All traffic traverses the Headend
Optical Distribution Network
Fibre
Copper
O
N
U
xDSL
Fiber
Access network
AMP
AMP
2nd Generation
Hybrid Fiber Coax
FTT curb
FTTCab
AMP
...
Headend
SNI
Fibre
Copper
N
T
O
L
T
Fibre
HOME
...
O
N
U
...
FTTB/C
N
T
CopperCoax
Copper
Carnegie Mellon
Passive Optical Network (PON) Access System
BB
NB
Comparison

WDM for upload
vs download
Axs Node
Modems use “worst case” technology.
 Has to fit within any voice channel so encoding suboptimal
 Wires can be very long (end‐to‐end)
ONU
OPT
Splitter
 1310nm
C O
C L
T
Carnegie Mellon


ONU
ISDN can be more aggressive – dated quickly
DSL is highly optimized for the transmission medium
 But there are some constraints on distance
1:32

Cable modem uses a transmission medium that has inherently a higher bandwidth, but the network architecture will limit throughput.

Fiber has high capacity but is a big investment
 1530 nm
 Designed for broadcasting, not for point‐point connections
ONU
 +/- 50 nm
ONU
OLT: Optical Line Terminator
ONU: Optical Network Unit
Carnegie Mellon
Access technological evolution
G-PON
2.5 Gbit/s
OPTICAL
ACCESS
622 Mbit/s
50 Mbit/s
VDSL
Carnegie Mellon
Outline


25 Mbit/s
ADSL2+
ADSL2
HDSL/ADSL
}
8 Mbit/s
2 Mbit/s
640 kbit/s

Classic view: different types of wires
Media encoding and streaming
 Big Picture
 Voice and video compression
 Protocols: SIP and RTP
Triple play: IPTV, cable, fiber to the home
ISDN
128 kbit/s
Analog
modems
56.6 kbit/s
28.8 kbit/s
Year
9.6 kbit/s
1989
1997
2000
2003
3
Carnegie Mellon
Different Classes of Streaming


Carnegie Mellon
Options When Using a Streaming Server
Multimedia streaming covers audio and video
Playback: you play back stored content
 Full content is available up front
 Flexibility in when content is transmitted

Broadcast: transmission of live content
 Content generated on the fly but flexibility in playback delay –
seconds in practice
 VCR/DVR functionality adds a twist

Interactive: voice and video conferencing
 Latency is very critical
 Impacts entire system: encoding, protocols, end‐system design, …
Carnegie Mellon
Compensating for Jitter:
VoIP with Fixed Playout Delay
Carnegie Mellon
Adaptive Video Streaming
Source
Playout delay


Feedback
Playback
Receiver reports dropped packets.
Sender reduces frame size/rate if loss rate is too high.
» Also uses probing to detect additional bandwidth

Similar to TCP, except that lost data is typically not
retransmitted.
Carnegie Mellon
Voice over IP (VoIP)
In voice over IP (VoIP), calls are
digitized, packetized, and transported
over an IP network: either an internal
IP network or the Internet.
Carnegie Mellon
Voice over IP (VoIP)
A media gateway connects
a VoIP network to the PSTN.
This gives VoIP users access
to PSTN users.
The media gateway must translate
between both signaling technology
and transport technology.
4
Carnegie Mellon
Carnegie Mellon
VoIP Signaling and Transport
VoIP Signaling and Transport
VoIP transport consists of a stream of VoIP packets.
Each VoIP packet contains a short amount codec-encoded voice.
There is no time to wait for error correction, so UDP is used.
The Real Time Protocol (RTP) header “fixes” weaknesses of UDP.
First, the RTP has a sequence number to place packets in order.
Second, RTP has a time stamp so that the voice steam
can be played back at the correct time.
Signaling is the transmission
of control messages.
Transport is the actual
transmission of voice.
Carnegie Mellon
Outline



Carnegie Mellon
Steps in Encoding and Decoding

Process has three steps on each side
Digitize: represent information in bits

Compress: reduce number of bits

Classic view: different types of wires
Media encoding and streaming
 Big Picture
 Voice and video compression
 Protocols: SIP and RTP
Triple play: IPTV, cable, fiber to the home
 Sample, quantize, (eg. PCM)
 Audio: GSM, G.729, G.723.3, MP3, …
 Video: MPEG 1/2/4, H.261, …


Send over the network
Reverse process on receive side: uncompress, convert, play – need to match!
Carnegie Mellon
Audio Encoding


Traditional telephone quality encoding: 8KHz samples of 8 bits each – 64 Kbps
CD quality encoding: 44.1 KHz of 16 bits
 1.41 Mbs uncompressed

MP3 compression similar to MPEG
 Frequency ranges that are divided in blocks, which are converted using DCT (Discrete Cosine Transform) , quantized, and encoded
Layer 1
Carnegie Mellon
ITU VoIP Codecs
Range
Ratio
384 kbps
4
Layer 2
192 kbps
8
Layer 3
128 kbps
12
Codec
Transmission Rate
G.711
64 kbps
G.721
32 kbps
G.722
48, 56, 64 kbps
G.722.1
24, 32 kbps
G.723
5.33, 6.4 kbps
G.723.1A
5.3, 6.3 kbps
G.726
16, 24, 32, 40 kbps
G.728
16 kbps
G.729AB
8 kbps
Two phones must
use the same codec
to encode and
decode voice.
They must agree on
one of several standard
codec protocols
through negotiation.
Generally, more
Compression gives
lower sound quality but
lowers transmission cost
5
Carnegie Mellon
Carnegie Mellon
Representative Video Bit Rates (Hi ↓ Lo Quality)
Video Encoding

Once frames are captured ("raw" video) resulting file is very large:

 320 x 240 x 24‐bit color = 230,400 bytes/frame  15 frames/second = 3,456,000 bytes/second  10 seconds takes around 30 Mbytes! (no audio) 



Commonly‐used encoding “tricks”:

 Per‐frame versus inter‐frame encoding

Leverage fact that successive frames tend to be similar
 Impacts latency
 Layered encoding
 Quality improves as you decode more layers



1.2 Gbps Uncompressed HDTV
19.4 Mbps ATSC ( ≈ HDTV quality)
8 ‐ 9 Mbps MPEG4 ( ≈ HDTV quality)
90 Mbps Uncompressed NTSC (SDTV) 3 ‐ 6 Mbps MPEG2 ( ≈ SDTV quality)
1.5 Mbps MPEG4 ( ≈ SDTV quality)
1.5 Mbps MPEG1 ( ≈ VHS < SDTV quality)
Note: ATSC, MPEG2, & MPEG4 support a wide
variety of formats (SDTV ↔ HDTV) Carnegie Mellon
Image Encoding: JPEG 


Carnegie Mellon
MPEG Encoding of Video
Joint Photographic Experts Group
Divide digitized image in 8 x 8 pixel blocks.
The DCT phase converts the pixel block into a block of frequency coefficients.

Moving Pictures Experts Group
MPEG uses inter‐frame encoding.

Three frame types:

 Exploits the similarity between consecutive frames
 Discrete Cosine Transform – similar to FFT





The quantization phase limits the precision of the frequency coefficient.
 This is based on a quantization table, which controls the degree of compression

I frame: independent encoding of the frame (JPEG)
P frame: encodes difference relative to I‐frame (predicted)
B frame: encodes difference relative to interpolated frame
Note that frames will have different sizes

Complex encoding, e.g. motion of pixel blocks, scene changes, ..

MPEG often uses fixed‐rate encoding.
The encoding phase packs this information in a dense fashion.
 Decoding is easier than encoding.
 One block at a time – include the DC component
I
B
B
P
B
B
P
B
B
I
B
B
P
B
Carnegie Mellon
MPEG System Streams


Combine one more MPEG video and audio streams in a single synchronized stream.
Consists of a hierarchy with meta data at every level describing the data.





System level contains synchronization information
Video level is organized as a stream of group of pictures
Group of pictures consists of pictures
Pictures are organized in slices …
B
Carnegie Mellon
MPEG Standards

MPEG 1 (1992)
 Compression of motion video & audio at about 1.5 Mbps (VHS Quality)
 Targeted at digital playback & storage
 Has Random Access capabilities
 Divides picture up into 8x8 pixel blocks ‐ Converts blocks to bit stream 
MPEG 2 (1994)
 Targets higher quality compression, typically at 3‐6 Mbps bit rates
 Being used for Direct Broadcast TV
 Large chunks of MPEG2 used in U.S. HDTV standard
6
Carnegie Mellon
MPEG 4



Carnegie Mellon
H.261, H.263, & H.264
Aimed at Multimedia Coding
Bit rates from 8 Kbps ‐ 40+ Mbps
Can code objects as opposed to NxN blocks


 Ability to interact & manipulate objects


Target real time videoconferencing
Subset of MPEG Wide variety of bit rates

 64 Kbps ‐ 128 Kbps: Face shot (video phone)
 384 Kbps: considered to be minimum speed for decent full screen Standard in 1999
Used in Quicktime 6, Direct TV, IPTV
videoconferencing
 We are using H.263/4, mostly @ 768 Kbps
H.264 quality > H.263 > H.261

 Newer protocols require more processing power
 Older protocols become less popular over time
Carnegie Mellon
Carnegie Mellon
Outline
Codecs

100000
Kbit/ sec
Alca tel Propr. Codec
M- JPEG
MPEG2
10000
Audio CD
MPEG4
1000
100

MP3
G.711
MP3Pro,
AAC
G.726
G.728

Classic view: different types of wires
Media encoding and streaming
 Big Picture
 Voice and video compression
 Protocols: SIP and RTP
Triple play: IPTV, cable, fiber to the home
10
G.729
G.723.1
1
1970
1975
1980
voice
1985
1990
HiFi audio
1995
2000
2005
2010
TV qual video
Source: Alcatel
Carnegie Mellon
Transport Protocol Properties

Reliability.
 Some lost data may be acceptable – depends on encoding and user expectations
 Timeouts typically result in unacceptable delay – there may not be enough time to retransmit data

Congestion control.
 Nature of the flow fundamentally limits its bandwidth
 Reduction of rate in response to congestion should reduce data size, Carnegie Mellon
Real Time Transport Protocol (RTP)

Multimedia senders append header fields before passing to transport layer

RTP logically extends UDP: application layer between UDP and application
RTP does not guarantee timely data delivery.
 Format, sequence numbers, timestamps, …

not transmit rate for samples
 E.g. change frame rate or frame size

 Simply helps applications with formatting and the collection of session information
Flow control: natural pacing.
 Samples should be paced at the rate of the data
 Too slow ‐‐> underflow and missed deadlines
 Too fast ‐‐> buffer overflow and lost data
 Guarantees can only be provided at lower level

The protocol has two parts: Real‐time Transport Protocol (carry data) and Real‐Time Control protocol (monitor quality, participant info, ..)
7
Carnegie Mellon
RTP Packet Format

Real Time Control Protocol (RTCP)

Source/Payload type
 Different formats assigned different codes
 Eg. GSM ‐> 3, MPEG Audio ‐> 14




Carnegie Mellon


Sequence numbers
Time stamps
Synchronization source ID
Miscellaneous fields, e.g., feedback
RTCP packets transmitted by each participant in RTP session to all others using multicast
Distinct port number from RTP
Reports on:





Loss rate Inter‐arrival jitter Identity of receivers Delay, (indirectly)
Control bandwidth sharing
 Needed for scalability
 E.g., 5% of data bandwidth
Carnegie Mellon
SIP Introduction

SIP Requests
SIP is an application level signaling protocol for initiating, managing and terminating sessions in the Internet





Carnegie Mellon

Registrations, invitations, acceptations, and disconnections
Sessions may include text, voice, video
Can use unicast or multicast communication
Client‐server model: request‐reply transaction



Common headers in plain text, similar to MIME/HTTP

 request/response line (e.g., INVITE a@b.com SIP/2.0)
 message headers (identification, routing, etc)
 message body, e.g., session description protocol
REGISTER: register user agents
INVITE: initiate calls
ACK: confirm responses
BYE: terminate or transfer calls
Other methods:
 CANCEL, OPTIONS, INFO, COMET, PRACK, SUBSCRIBE, NOTIFY, REFER

SIP response: HTTP‐like (e.g., SIP/2.0 200 OK)
Carnegie Mellon
SIP Entities

User agent (UA)
 UA client (UAC) and UA server (UAS)

Carnegie Mellon
SIP Header Fields


 request/reply: use the same route
 may be different from the media route
Proxy server
 relay calls; chaining; forking

Redirect server

Call‐ID: identification

Content‐Type/Length: payload info
 redirect calls

Registrar server
 UA registration (UA whereabouts)
From/To: caller/callee URI
Via: proxy routing
 unique at caller
 e.g., Content‐Type: application/sdp
8
Carnegie Mellon
SDP: Session Description Protocol

Carnegie Mellon
SIP Session Establishment and Call Termination
Media type, network/transport parameters
 e.g., media: media, port, protocol, format_list
m=audio 2000/2 RTP/AVP 0 98
a=rtpmap:0 PCMU/8000
 connection: net_type, add_type, address
 c=IN IP4 1.2.3.4/127/3



Ref: http://www.ietf.org/rfc/rfc2327.txt
From the RADVISION whitepaper on SIP
Carnegie Mellon
SIP Call Redirection
Carnegie Mellon
Call Proxying
From the RADVISION whitepaper on SIP
From the RADVISION whitepaper on SIP
Carnegie Mellon
H.323: Standard for Real‐Time Conferencing

Components:
 Standalone endpoints: terminals
 Gateways: permit communication between endpoints and circuit‐switched phones
 Gatekeepers: provide address translation, authorization, bandwidth management, …
Carnegie Mellon
Outline



Classic view: different types of wires
Media encoding and streaming
Triple play: IPTV, cable, fiber to the home
 Multicast, traffic management
 Trends for cable, fiber
9
Carnegie Mellon
Change in User Demand: Triple Play


Delivering “Triple Play” to Consumers
Provisioning over a single broadband connection for: broadband Internet, TV, and latency sensitive telephone services.
Cable companies already offer TV and data service –
natural add phone service
 Very low bandwidth
 Plays well in market – convenient and cost effective
 Services emerged as a fourth component

Carnegie Mellon


 Rest of the lecture

Cable and fiber have plenty of bandwidth – channels can be dedicated to specific uses
 Take “TV” channels and use for Internet service and voice
 Demand for large numbers of channels is starting to stress even cable What about telephone companies
 How do you deliver TV service over a phone wire?
 Necessary to be competitive

Can be viewed at the 1990’s dream of integrated services networks: voice, video, and data
IPTV delivery over DSL optimized to work around last mile bottleneck capacity: deliver less unpopular channels “on demand”, DSL style (switched video)
Voice and TV over the Internet?
Carnegie Mellon
IPTV ‐ Internet Protocol television

IPTV Components
Deliver TV broadcast to users using Internet topology
Design only delivers channels of interest to the home
 Contrast with broadcast nature of cable design
 Reduce load on the last mile link


Streaming server: encoding of live content


IP network with multicast support
DSLAM ‐ Digital Subscriber Line Access Multiplexer

CPE ‐ Customer Premises Equipment
 Also, encryption (DRM), protocol support, e.g., for DVR, PIP, …
 Also, services, data, and voice, i.e. triple play

Carnegie Mellon
 Mux/demux from/to multiple subscribers
Uses IP multicast and RTP for distribution of broadcast TV
 More on this later
 Other traffic is unicast, using different higher level protocols
 Receives IP streams from DSLAM and distributes them throughout the home

STB ‐ Set Top Box
 Decryption, decoding, D/A, channel control, …
Carnegie Mellon
Delivering video services over DSL
Carnegie Mellon
TV over IP using DSL
Switch
Streaming Server


DSL has limited bandwidth (few to few 10s of Mbps)
 Depends on distance, technology
Must limit the number of channels, other traffic
 Read your contract carefully
10
Carnegie Mellon
Multicast – Efficient Data Distribution
TV over IP using FTTH

Carnegie Mellon
Pushing fiber to the home dramatically increases bandwidth, e.g., channels
Src
Src
 Easy to justify for, e.g., apartment buildgin
Carnegie Mellon
IP Multicast Service Model (rfc1112)







Carnegie Mellon
IP Multicast Addresses
Each group identified by a single IP address
Groups may be of any size
Members of groups may be located anywhere in the Internet
Members of groups can join and leave at will
Senders need not be members
Group membership not known explicitly Analogy:

Class D IP addresses
 224.0.0.0 – 239.255.255.255
1 110

Group ID
How to allocated these addresses?
 Well‐known multicast addresses, assigned by IANA
 Transient multicast addresses, assigned and reclaimed dynamically
 Addresses can be centrally assigned in IPTV networks
 Each multicast address is like a radio frequency, on which anyone can transmit, and to which anyone can tune‐in.

For IPTV membership is known and access is controlled end‐to‐end
Carnegie Mellon
IP Multicast Service


Sending – source sends one packet
Receiving – two new operations
 Join‐IP‐Multicast‐Group (group‐address, interface)
 Leave‐IP‐Multicast‐Group (group‐address, interface)
 Receive multicast packets for joined groups via normal IP‐Receive operation

Routers must replicate packets
Carnegie Mellon
Multicast Router Responsibilities

Learn of the existence of multicast groups (through advertisement)
Identify links with group members

Establish state to route packets

Src
 Group management ‐ next
 Replicate packets on appropriate interfaces
 Routing entry:
Src, incoming interface List of outgoing interfaces
11
Carnegie Mellon
IP Multicast Architecture
Carnegie Mellon
Internet Group Management Protocol

Service model

Hosts
End system to router protocol is IGMP
Host who wants to joint multicast message send IGMP join request to its router
 Router will add host’s port to the multicast tree
Host-to-router protocol
(IGMP)

Each host keeps track of which mcast groups are subscribed to

Hosts periodically contact router with groups of interest
Routers
 Socket API informs IGMP process of all joins
 Objective is to keep router up‐to‐date with group membership
 Routers need not know who all the members are, only that members Multicast routing protocols
(various)
exist
Carnegie Mellon
Source‐based Trees
Carnegie Mellon
Multicast Routing Protocols

Router
S Source
R Receiver
 Flood and prune: broadcast pruned to multicast
 Link state: routers broadcast information on its MC receivers
 Distance vector: traditional protocol that generates tree that R
connects all possible members
R
S
Many routing protocols proposed for IP multicast

R

Not used in public internet due to security and public concerns
IPTV network requirement are a good fit for multicast
 Private network: full control over networks, endpoints
 Single source and all destinations are known – routing is easy!
S
R
Carnegie Mellon
Controlling Quality of Service

IPTV and voice have very high QoS requirements
 Need to compete with traditional telephone and cable TV delivery based on circuits
 Bandwidth requirements are known

Data and interactive TV services have traditional web‐like requirements
Carnegie Mellon
Credit

Lecture includes slides from several sources:
 home.ubalt.edu/abento/427/VOIPLASTMILE/VOIPLASTMILE.PPT
 http://www.item.ntnu.no/fag/ttm7/Lectures/5_Convergence_IP_T
V.ppt
 www.okstate.edu/elec‐
engr/scheets/ecen5553/fall12/TCM2930W.PPT
 Optimize response time, e.g., browsing
 Maintain bandwidth, e.g., Netflix
 Bursty bandwidth requirements

QoS control is based on careful bandwidth allocation and enforcing bandwidth limits using traffic shapers
12
Download PDF
Similar pages