IPTV and the Future of Video Services

Ali C. Begen
Video and Content Platforms – Research & Advanced Development

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 Presenter Today – Ali C. Begen

 With Cisco since 2007

Video and Content Platforms
Research & Advanced Development Group

 Works in the area of

Architectures for next-generation video
transport and distribution over IP networks

 Interested in
Networked entertainment
Internet multimedia
Transport protocols
Content distribution

 Received a Ph.D. degree from Georgia Tech

in 2006
 Member of the IEEE and ACM

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 2
 In This Tutorial

 We will study
A detailed overview of IPTV and its building blocks
Architectures and protocols used to carry video over IP in core, aggregation and
access networks
Current standardization efforts
Observations and experiences from deployments

 We will also touch on several concepts such as

The medianet vision
Over-the-top services
Internet/Web TV and adaptive streaming
Targeted advertising

Slides will be posted at [Link] after the tutorial

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 3
 Role of Video in Global IP Traffic

1995: Web overtakes Gopher, FTP Video

Communication
2000: P2P overtakes WWW
2013: Video content overtakes P2P
Video
2025: Video communication overtakes content Content

P2P

WWW

Gopher, FTP

2025+
2013-2025
2000-2013
1995-2000
1993-1995

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 4
 Global IP Traffic Growth
IP Traffic will Increase 6x from 2007 to 2012

50000
Mobility
Business Internet
Business IP WAN
Consumer Internet
PB/Month

Consumer IPTV/CATV
25000

0
2005 2006 2007 2008 2009 2010 2011 2012

Source: [Link] PB: 1e15 bytes

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 5
 YouTube and Hulu Traffic

60

50

40 Hulu
PB/Month

30

20
YouTube
10

0
Entire Internet Backbone in 2000 YouTube and Hulu Sites in 2008
(US) (US)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 6
 Household (US) Bandwidth Needs in 2010
1xHDTV + 1xSDTV + 2xPVRs + 1xVoIP + 2xPCs w/ HSD

1.2
TB/Month

0.6

0
2010
HDTV SDTV HSD VoIP

Twenty such homes would generate more traffic

than traveled the entire Internet backbone in 1994/1995
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 7
 Video 2.0: A Consumer-Driven Experience

Video 1.0 Video 2.0

Traditional Sources Community Sources

 Broadcast  Personalized, on demand, Internet

 Limited content  Unlimited content
 Single device  Multiple devices
 One size fits all content  Real-time mashups
 Scheduled viewing  Rapidly changing content popularity

What they want. When they want it.

Where they want it. How they want it.
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 8
 Consumers Seek A Rich Media Experience

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 9
 Networking Requirements are Changing

From To
Packets Streams
Unpredictable Predictable
Content-agnostic Media-aware

Endpoint-agnostic Endpoint-aware

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 10
 The Internet Needs to Evolve

Technology Platform LANs Internet ???

medianet

Technology Shift Digital Web 1.0 Web 2.0

Business Value Automation Transactions Collaboration

Social Value Efficiency Autonomy Visual

Participation
Networking

1985–1995 1995–2005 2005+

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 11
 Video Platforms, Systems and Solutions

Consumer Service Provider Business

Connected Life Any Play Business Video

Video Systems
Systems and
Experiences Visual Networking

medianet Technologies and Products

Technology
Platform Connected
Home
medianet
IP NGN AVVID

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 12
 Consumer Video
Interactivity, Choice and Mobility Consumer

Interactive Safety and Connected Teleworking

Entertainment Security Life

Social Networking Protecting People Video on the Move Collaborate with

with Video and Assets and Mobility Video

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 13
 Service Provider Video
SPs Transforming to Experience Providers
Service Provider

Consumer Service OpEx and New Business

Experience Assurance Service Velocity Models

Integrated Personalized and Differentiated Delivery and

Offerings Relevant Offerings Proof of Play

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 14
 Business Video
Enabling Process Transformation
Business

Collaboration Safety and Customer Operational

Security Intimacy Excellence

Accelerated Protecting People New Revenue Manage Complexity

Decision Making and Assets Streams and Scale

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 15
 What Is IPTV?
The Fundamental Component for Connected Homes
IPTV = IP Network-delivered Television
- Switched digital video (SDV)
- Electronic program guides (EPG)
- Video recording (DVR/PVR/nDVR)
- Video-on-demand (VoD)
- Interactive TV applications
- Targeted (advanced) advertising

Broadband
IP Access

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 16
 Global Growth for IPTV

90
Asia
75 Europe
North America
IPTV Subscribers (Million)

60 RotW

45

30

15

0
2009 2010 2011 2012 2013
Source: MRG
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 17
 Trends Driving IPTV Adoption

 Subscribers want more choice and control

New generation grew up computer/Internet savvy
Customized for me – One bill, one provider, integrated services
 Codec, access, server and CPE technologies are improving
MPEG-4 AVC (H.264) improvements
New ADSL2+, VDSL2, FTTx, DOCSIS 3.0 access technologies
Moore’s law advancements in processing and memory
 Competition is increasing among service providers
No longer limited by access – All services over any network
Traditional markets are going away – VoIP is almost free
 Video is driving next generation service provider network design
Driven by video’s bandwidth and QoS requirements
Experiencing exponential growth in Internet video usage

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 18
 Telcos, Cable MSOs are Aggressive on Bundles

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 19
 Further Reading

 IP Next-Generation Network (IP NGN)

[Link]
[Link]

 medianet Technologies
[Link]
[Link]
“IPTV and video networks in the 2015 timeframe: the evolution to medianets,”
IEEE Communications Magazine, Dec. 2009

 Cisco Visual Network Index

[Link]

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 20
 IPTV Basics and SLA Requirements

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 End-to-End IPTV Network Architecture

Super Headend (SHE) Video Hub Office (VHO)

Regional Headend (RHE)
Video Switching
Office (VSO)
Acquisition
Encoders
National Encoders
Content Acquisition STB
Ad Insertion
Back Office FTTx STB
Transcoders Billing
Transcoders VoD Servers
Regional
Vault Content STB
Streamers
Metro
VoD Servers Ethernet
CA/DRM STB
Streamers Retransmission
IPTV Service CA/DRM Servers
Assurance uBR
IPTV Content
CMTS STB
Manager

IP/MPLS Metro Aggregation & Access

Core Network Distribution Network Network STB

Aggregation DSLAM
Core Core
Router STB
Router Router

Aggregation
Router STB

Core Distribution/Aggregation Access

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 22
 Broadcast IPTV = IP Multicast

 Various Transports
Native IP multicast, MPLS, L2, optical

 IP Multicast Endpoints
Sources: Encoder, transcoder, groomer, ad-splicer
Receivers: Transcoder, groomer, ad-splicer, eQAM, IP STB

 IP = IPv6 in Japan or = IPv4 (Rest of the world)

No/slow move to IPv6 for IPTV in RotW

 Transport Challenges
Packet loss
Out-of-order delivery
Packet duplication

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 23
 Unicast vs. Multicast

Unicast
Source

Multicast
Source

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 24
 Multicast Services and Protocols

 ASM: Any-source multicast (RFC 1112 – 1989)

Sources send packets to multicast groups
Receivers join to groups (G) to receive from any source

 SSM: Source-specific multicast (RFC 4604/4607 – 2006)

Receivers subscribe (S,G) channels to receive traffic only from source S sent to group G
Primarily introduced (by IETF) for IPTV-like services

 IETF standardized
Receiver-to-Router Protocols
IGMPv3 (IPv4) and MLDv2 (IPv6) with (S,G) signaling
Router-to-Router Protocols
PIM-SSM
IGMPv3 Proxy Routing, Snooping on HAG and L2 access devices

 IETF have not standardized anycast source redundancy (yet)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 25
 Real-Time Transport Protocol (RTP)
[Link]
 Basics
First specified by IETF in 1996, later updated in 2003 (RFC 3550)
Runs over any transport-layer protocol (Typically over UDP)
Runs over both unicast and multicast
No built-in reliability
 Main Services
Payload type identification
Sequence numbering
Timestamping
 Extensions
Basic RTP functionality uses a 12-byte header
RFC 5285 defines an RTP header extension mechanism
 Control Plane – RTCP
Provides minimal control and identification functionality
Enables a scalable monitoring functionality (Sender, receiver, extended reports)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 26
 RTP Transport of MPEG2 Transport Streams

Religion

Politics

Application
IPTV Application
Presentation MPEG2-TS

Session
RTP
Transport UDP IGMP

Network IP

Data Link

Physical

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 27
 Packetization into MPEG2 Transport Streams
Single Program Transport Streams (SPTS)
Video
PES
Video

Video
Encoder Packetizer
Audio

MPEG2-TS Mux
PES
SPTS
Audio

Audio
Encoder Packetizer

Additional Audio Tracks

Application Data
PAT (PID=0), PMT
Video
PES

Video
Video
Depacketizer
Decoder
MPEG2-TS Demux Audio
PES
SPTS

Audio
Audio
Depacketizer
Decoder

Additional Audio Tracks

Application Data
PAT (PID=0), PMT
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 28
 TS Packets and Multi-Program Transport Streams

188-byte MPEG2-TS Packet

Video
PES
0x47
PID

Program A Audio
PES
TS Packet Header:
- Sync byte (0x47)
- Transport error indicator

...
- Payload unit start indicator

MPEG2-TS Mux
- Transport priority
Video
- Packet ID PES MPTS
- Scrambling control
- Adaptation field control
Program Z Audio
- Continuity counter PES
- Adaptation field (optional)

CAT (PID=1)
Program & Systems Information (PSIP)
PAT (PID=0), PMT (for each program)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 29
 RTP Transport of MPEG2 Transport Streams
[Link]

V=2 P X CC M PT Sequence Number N 188-byte MPEG2-TS Packets

Timestamp
... ... ...
Synchronization Source (SSRC) Identifier
Contributing Source (CSRC) Identifiers MPEG2-TS Payload
…

RTP Encapsulation

RTP MPEG2-TS Payload

8 bytes
UDP Encapsulation

UDP RTP MPEG2-TS Payload

20/40 bytes
IP Encapsulation

IP UDP RTP MPEG2-TS Payload

Default IP header size is 20 and 40 bytes for IPv4 and IPv6, respectively
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 30
 End-to-End IPTV Network Architecture

Super Headend (SHE) Video Hub Office (VHO)

Regional Headend (RHE)
Video Switching
Office (VSO)
Acquisition
Encoders
National Encoders
Content Acquisition STB
Ad Insertion
Back Office FTTx STB
Transcoders Billing
Transcoders VoD Servers
Regional
Vault Content STB
Streamers
Metro
VoD Servers Ethernet
CA/DRM STB
Streamers Retransmission
IPTV Service CA/DRM Servers
Assurance uBR
IPTV Content
CMTS STB
Manager

IP/MPLS Metro Aggregation & Access

Core Network Distribution Network Network STB

Aggregation DSLAM
Core Core
Router STB
Router Router

Aggregation
Router STB

Core Distribution/Aggregation Access

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 31
 Video Headend Building Blocks

Satellite MPEG-2 Transcoding CAS/DRM

ACQUISITION

ENCODING

PROCESSING

MANAGEMENT
reception MPEG-4 AVC Transrating Remote ops
Off-air reception Standard def. Splicing Metadata
Fiber reception High definition Multiplexing Billing
Signal conversion Audio encoding Ad-insertion VoD servers
Video apps

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 32
 Role of IPTV Middleware

 Middleware provides interoperability for system components

 Middleware enables
Revenue producing IPTV services
Differentiation for service providers
Consistent and extensible consumer experience
Delivery of rich media to consumers
Compelling GUIs

 Middleware ties together all parts of an end-to-end IPTV system

EPG, content navigation
CAS/DRM, VoD servers, asset management, service packaging
EAS, STB support, network management
Billing, triple-play integration, subscriber management

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 33
 Telco IPTV System Reference Architecture

OSS/BSS

VoD IPTV IPTV IPTV

Content
Backoffice Network Application Application User
Provider
System Server Platform Server Interface
IPTV
Content Client
Delivery Platform
System
CA/DRM
Internet Core Network OLT/ Fiber/ ONT/
DSLAM Copper Modem IP STB

Video
Headend
CA/DRM
System

IP Content and Delivery over Fiber/xDSL Access

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 34
 Cable IPTV System Reference Architecture

OSS/BSS

VoD IPTV IPTV IPTV

Content
Backoffice Network Application Application User
Provider
System Server Platform Server Interface
IPTV
Content Client
Delivery Platform
System
CA/DRM
Core Network HFC Cable
Internet CMTS IP STB
Modem

Video
Headend
CA/DRM
System

IP Content and Delivery over DOCSIS (VDOC)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 35
 Efficiency Gains From IPTV in Cable Delivery

Variable Bitrate

• VBR provides a bandwidth savings of 40-60%

• IPTV is the best choice for narrowcast statmux and AVC statmux

Switched Video

• Switching is the way to offer unlimited channels

• IPTV provides built-in switching functionality

Advanced Coding

• AVC provides a bandwidth savings of 50% over MPEG2

• IPTV solves the problem of slow channel change

QAM Sharing

• Convergence provides further bandwidth savings

• We can share QAMs for VoD and SDV as well as for video and DOCSIS

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 36
 Types of Video Services

 Transport (Contribution and Primary Distribution)

 IPTV (Secondary Distribution) / CATV
IP multicast distribution from centralized super headends
Driving enhanced multicast features and functions
 VoD (Secondary Distribution)
Distributed architecture for better scalability
Non-real-time content distribution to caches
More impact on metro and access networks, less impact on the core
 Enterprise
mVPN based
Driving enhanced multicast features and functions
 Over-the-Top (e.g., Hulu, Apple TV, Netflix)
Approaches are still evolving
Players are in and out everyday

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 37
 Market Dynamics

 Operators are moving from TDM to IP or IP/MPLS core networks

To integrate disparate networks
To converge IP data (business/residential) and IP video services on one IP/MPLS backbone
To reduce costs
 FCC says “If you use public airways to distribute cable TV networks, you must provide this
content to the satellite competition as well”
MSOs are leasing dark fiber to move from satellite to IP terrestrial backbones
 It is a private network, so FCC rules do not apply
 This enables operators to distribute exclusive channels
 Cable MSOs are also moving to SDV – Switched Digital Video
To enable multicast-based delivery (through QAM or IPTV)
To send only the active channels to cope with the last-mile bandwidth limitations
 Support for VoD and IPTV from telcos is still in its infancy
Need to deal with the limited local loop bandwidth
Need for retail and wholesale services support

Next round of core and edge network upgrades for many SPs
will be influenced by support for video services
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 38
 IPTV must Deliver Entertainment-Caliber Video
Tolerance is One Visible Artifact per Movie
Contribution
Network
Studio Studio Primary
Distribution
IP/MPLS Secondary
Core
Distribution
IP/MPLS
Homes
Core
(Millions)
Final IP/MPLS
Studio Mobile
Studio Super Core
Headend Regional
Reliability
Access is more
(x2) Headend difficult to achieve here
Video
National Content (x10s)
Switching Office
Acquisition Regional Content (x100s)

Encoders Acquisition Ad Insertion

Vault Encoders Back Office/Billing

Streamers VoD Servers

Streamers

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 39
 Taxonomy of Video Service Providers

Studio to Studio
Uncompressed/Lossless compression
- SD: 270 Mbps (SMPTE 259M)
Higher Bandwidth

- HD: 1.5 – 3 Gbps

(SMPTE 292M, 372M, 424M)
P-to-P P-to-Mp Contribution

Provider to Subscriber
Compressed
- SD: 2 – 6 Mbps
- HD: 6 – 16 Mbps
P-to-P for VoD
Primary
P-to-Mp for IPTV Distribution
May be wholesaled

Owner to Provider
Secondary Compressed (High quality)
Distribution P-to-P  P-to-Mp

Stricter Requirements
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 40
 Digital Video Bandwidths

Uncompressed Digital Video

SDTV (480i CCIR 601 over SD-SDI SMPTE 259M) 165.9 – 270 Mbps

EDTV (480p or 576p via SMPTE 344M) 540 Mbps

HDTV (1080i or 720p over HD-SDI SMPTE 292M) 1.485 Gbps

HDTV (1080p over Dual Link HD-SDI SMPTE 372M) 2.970 Gbps

MPEG-2 Compressed Video

SDTV Broadcast (3.75 Mbps for cable VOD) 3 – 6 Mbps

HDTV Broadcast (19.3 Mbps for ATSC DTV) 12 – 20 Mbps

SDTV Production (Contribution – 4:2:2 I-frame only) 18 – 50 Mbps

HDTV Production (Contribution – 4:4:4 I-frame 10-bit) 140 – 500 Mbps

MPEG-4 AVC / H.264 Compressed Video

SDTV Broadcast (~50% less than MPEG-2) 1.5 – 3 Mbps

HDTV Broadcast (1080i about 4x SDTV) 6 – 9 Mbps

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 41
 Video SLA Requirements

 Throughput
Addressed through capacity planning and QoS (i.e., Diffserv)
 Delay/Jitter
Controlled with QoS
Absorbed by de-jittering buffer at IP STB
 We desire to minimize jitter buffer size to improve responsivity
 Jitter originating in the core is rather insignificant
 Loss
Controlling loss is the main challenge
 Service Availability
Proportion of time for which the specified throughput is available within the
bounds of the defined delay and loss

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 42
 Video SLA Requirements

Network Approaches
Over-engineered - Fast convergence
Cost and Complexity

- Fast reroute
- Spatial diversity
- Source diversity
Viable Application Approaches
Engineering - Forward Error Correction (FEC)
- Temporal diversity

Must be re-engineered

Loss Occurrence
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 43
 Further Reading

 Articles
“Not all packets are equal, part I: streaming video coding and SLA requirements,” IEEE
Internet Computing, Jan./Feb. 2009
“Not all packets are equal, part II: the impact of network packet loss on video quality,”
IEEE Internet Computing, Mar./Apr. 2009

 Recent Special Issues and Sessions

IEEE Communications Magazine (Multiple issues in 2008)
IEEE Internet Computing (May/June 2009)
IEEE Transactions on Broadcasting (June 2009)
IEEE CCNC 2008, 2009 and 2010

 Industry Tests
Light Reading: Cisco Put to the Video Test
[Link]
EANTC Experience Provider Mega Test
[Link]
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 44
 Lossless Video Transport in the Core Networks

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 Four Primary Causes for Packet Loss

 Excess Delay
Renders media packets essentially lost beyond an acceptable bound
Can be prevented with appropriate QoS (i.e., Diffserv)
 Congestion
Considered as a catastrophic case, i.e., fundamental failure of service
Must be prevented with appropriate QoS and admission control
 PHY-Layer Errors (in the Core)
Apply to core and access – Occurrence in core is far less
Considered insignificant compared to losses due to network failures
 Network Reconvergence Events
Occur at different scales based on topology, components and traffic
Can be eliminated with high availability (HA) techniques

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 46
 What are the Core Impairment Contributors?

Impairment Rate
Trunk failures .0010 /2h

Hardware failures .0003 /2h

Software failures .0012 /2h

Non-stop forwarding (NSF) and
Stateful switch-over (SSO) help here

Software upgrades (Maintenance) .0037 /2h

Modular code (IOS-XR) helps here

Total .0062 /2h

Note that average mean time between errors on a DSL line is in the
order of minutes when no protection is applied

Back of envelope calculations across several SPs show

mean time between core failures affecting video is > 100 hours

Source: Data from industry standards, customers and assumptions

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 47
 Il Buono, il Brutto, il Cattivo

No Loss – Perfect Quality

0.5% Packet Loss 5% Packet Loss

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 48
 MPEG Frame Impact from Packet Loss
GoP Size: 500 ms (I:P:B = 7:3:1)

100
90
I-Frame Loss Probability (%)

80 FUD on 50-ms Requirement

70 - Legacy telco voice requirement
60 - Artifact of SONET/SDH
50 FACTS
40 - IP multicast is UDP based
30 - The only acceptable loss is 0%
20 - At 50 ms outage, an I-frame will
10 be impacted with a chance of 34%
0
50 100 150 200 250 300 350 400 450 500 550 600
Outage Duration (ms)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 49
 Fast Convergence or Fast Reroute

Core
Router

Video Core Edge

Source Router Distribution

Core
Router

 Network reconverges (reroutes) on a core network failure (link or node)

 Fast Convergence (FC) – Fast Reroute (FRR)
 Lowest bandwidth requirements in working and failure cases
 Lowest solution cost and complexity
! Requires fast converging network to minimize visible impact of loss
 Is NOT hitless – Loss of connectivity before connectivity is restored

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 50
 Multicast (SSM) Fast Convergence (CRS-1)
Convergence Time Following a Failure

4000
Number of IPTV Channels

Test shows that

no more than one I-frame is
800
lost even in the worst case
Prefix prioritization allows important
prefixes (e.g., premium channels) to
converge first
400 Max of Max
Advances in IP optical integration
can further reduce the outage to
Median of Median
sub 20 ms in many cases (lossless
0
in some cases) 200 400 600 800 1000
Reconvergence (ms)
Tested with 2500 IGP prefixes and 250k BGP routes
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 51
 Forward Error Correction (FEC)

Core
Router

Video Core Edge

Source Router Distribution

Core
Router

 FEC adds redundancy to the source data to allow the receiver to detect and repair
losses (within some bound)
 FEC
 Is hitless from loss due to core network failures if loss can be constrained
 Does not require path diversity – Works for all topologies
! Requires fast converging network to minimize FEC overhead
 Incurs delay – Longer outages require larger overhead or larger block sizes

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 52
 Temporal Diversity (Redundancy)

Delayed Stream

Primary Stream
Video Core Core Edge
Source Router Router Distribution

 Let Q (ms) denote the max outage duration that is intended to be repaired
 Packets are transmitted twice, each separated by Q-ms delay
 Temporal diversity
 Is hitless from loss due to core network failures if loss can be constrained
 Does not require path diversity – Works for all topologies
! Requires fast converging network to minimize Q
 Introduces 100% overhead and Q-ms delay

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 53
 Spatial (Path) Diversity – Live/Live

Primary Stream

Primary Stream
Video Core Core Edge
Source Router Router Distribution

 Two streams are sent over diverse paths in the core

 Spatial diversity
 Introduces no delay if the paths have equal propagation delays
 Requires network-level techniques to ensure spatial diversity
 Incurs 100% overhead
– May not be an issue where redundant capacity is normally provisioned
– E.g., dual-plane core networks

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 54
 Source (Site) Diversity

Primary Stream
Video Core Core
Source Router Router

Edge
Distribution
Primary Stream
Video Core Core
Source Router Router

 Source (Site) diversity

 Introduces no delay if the paths have equal propagation delays
 May not require network-level techniques to ensure spatial diversity
– Topology dependent
 Incurs 100% overhead
– May not be an issue where redundant capacity is normally provisioned

This provides protection against single point-of-failure at the source,

but this is not a hitless recovery as the sources are not in sync
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 55
 Spatial Diversity through Multiple Interfaces

Primary Stream
Core Core
Router Router

Video Edge
Source Distribution
Primary Stream
Core Core
Router Router

 Assume the headend is connected to two disjoint networks

 This approach
 Incurs 100% overhead
– May not be an issue where redundant capacity is normally provisioned
 May not require network-level techniques to ensure spatial diversity
– Topology dependent
 Offers a hitless recovery as the both primary streams are in sync
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 56
 Towards Lossless IPTV Transport
Deployment Scenarios
Increasing Cost
and Complexity
TE +
Re-engineering

Live / Live
Network

MTR
+ Live / Live

MPLS TE FRR MPLS TE FRR

+ TD
Increasing
Loss
MoFRR +
Re-engineering

Live / Live
No Network

Fast
Convergence + MoFRR
FEC or TD

Fast
Convergence

Lossless One GoP Impacted

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 57
 Further Reading

 Articles
“Network-reconvergence technologies for low-loss video transport,”
URL: [Link]
“Deploying diffserv in backbone networks for tight SLA control,” IEEE Internet
Computing, Jan./Feb., 2005
“Designing a reliable IPTV network,” IEEE Internet Computing, May/June 2009

 Related Specifications
[Link]
[Link]
[Link]
[Link]

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 58
 Reliable Video Distribution in the Access Networks

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 VQE – A Unified QoE Solution

 IPTV viewers have two criteria to judge their service

Artifact-free audiovisual quality
Packets dropped in access and home networks must be recovered quickly
Packet loss may or may not be correlated in spatial and/or temporal domain
Loss-repair methods must be multicast friendly
Short and consistent zapping times
Compression and encryption used in digital TV increase the zapping times
Multicasting in IPTV increases the zapping times
Zapping demand varies the zapping times
 Service providers need a scalable unified solution that
Is standards-based and interoperable with their infrastructure
Enables versatility, quick deployment and visibility into the network
Extends the service coverage area, and keeps CapEx and OpEx low
 Our goals are to offer
Glitch-free audiovisual quality, short and consistent zappings even in low-bandwidth networks
Monitoring tools that isolate and pinpoint the problematic locations

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 60
 A Simplified Model

VQE Client
Retransmission Server
(w/ Feedback Target) STB VQE Client

STB
VQE Server VQE Client
DSLAM
Transport & STB
Access
Distribution Network VQE Client

STB
RTP Source Aggregation
Router
VQE Client
Channel 1 DSLAM
Channel 2 STB

 Each TV channel is served in a unique (SSM) multicast session

IP STBs join the respective multicast session(s) for the desired TV channel
Retransmission servers join all the multicast sessions
 (Unicast) Feedback from IP STBs are collected by the feedback target
NACK messages reporting missing packets
Rapid channel change requests
RTCP receiver and extended reports reporting reception quality

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 61
 Packet Loss Rate Tolerance Limits
Each Random or Bursty Loss Counts for One Artifact

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 62
 Impairments in xDSL Networks

 Twisted pair is subject to

Signal attenuation  Use shorter loops
Cross talk  Use Trellis Coding and RS-based FEC
Impulse noise  Use RS-based FEC with interleaving

 There are three types of DSL impulse noise

REIN  Short burst of noises (< 1 ms)
PEIN  Individual impulse noise (> 1 ms, < 10 ms)
SHINE  Individual impulse noise (> 10 ms)

 We observe different noise characteristics

Among different SP networks
Among different loops in the same SP network

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 63
 ADSL Transmitter Reference Model
ITU-T Recommendation G.992.1

Fast Path

AS0
AS1 Scrambler
CRC

Tone Ordering
& FEC
AS2
MUX/Sync

AS3 DMT
LS0 Symbols
LS1 Scrambler
LS2 CRC Interleaver
& FEC
OAM

Interleaved Path

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 64
 ADSL and ADSL2+ Configurations

ADSL ADSL2+
Data bytes per RS codeword, K 239 bytes 69 bytes
Parity bytes per RS codeword, R 16 bytes 10 bytes
Correctable byte errors per RS codeword, T = R / 2 8 bytes 5 bytes
Total bytes per RS codeword, N = K + R 255 bytes 79 bytes
# of RS codewords per DMT symbol, 1/S 1 11
DMT duration, t 250 us 250 us
Line data rate, LDR = N / S / t 8.0 Mbps 27.4 Mbps
Net data rate, NDR = LDR x K / N 7.5 Mbps 24 Mbps
Interleaver depth, D 32 352
Size of required memory, B = (N–1) x (D–1) 7874 bytes 27378 bytes
Interleaving delay, ID = B / LDR 7.87 ms 7.97 ms
Block size (Protection period), PP = N x D / LDR 8.16 ms 8.10 ms
Correctable error burst length, BL = D x T 256 bytes 1760 bytes
Impulse noise protection, INP = floor( BL / (N / S)) 1 2

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 65
 Example: Interleaving of RS Codewords

Data bytes
Parity bytes

Original RS Codewords
1 1 2 2 3 3 4 4

1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
Interleaved RS Codewords

 Interleaving
 Spreads a bursty error among multiple codewords
 Allows the decoder to repair the error with fewer parity bytes
 Introduces delay
 Renders the whole block useless upon a decoding failure

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 66
 Fast vs. Interleaved Path

 Assumptions
One impulse noise arrives every 15 seconds
2% of these impulses cause an error
 Conditional Probability of (DMT error | There is an error)
85%  One DMT in error
12%  Two DMTs in error
3%  Three or more DMTs in error
 Fast Path
ADSL/ADSL2+: One (maybe two) IP packet loss in every 750 seconds
 Interleaved Path (Interleaving delay: 8 ms)
ADSL: Up to 7 IP packet losses (at the net rate) in every 5000 seconds
ADSL2+: Up to 19 IP packet losses in every 25000 seconds

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 67
 First-Line of Defense in Loss Repair
1-D/2-D Parity Forward Error Correction
L  Source Block Size: D x L
 1-D Column FEC (for Bursty Losses)
1 2 3 R1 Each column produces a single packet
Overhead = 1 / D

FEC Packets
4 5 6 R2 L-packet duration should be larger than the
(target) burst duration
D XOR
7 8 9 R3  1-D Row FEC (for Random Losses)
Each row produces a single packet
10 11 12 R4
Overhead = 1 / L

XOR  2-D Column + Row FEC

Overhead = (D+L)/(DxL)

C1 C2 C3

FEC Packets

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 68
 First-Line of Defense in Loss Repair
Example: 2-D Parity Forward Error Correction

0 1 2 3 4 5 R0

6 7 8 9 10 11 R1

12 13 14 15 16 17
18 19 20 21 22 23 R3

24 25 26 27 28 29 R4

30 31 32 33 34 35 R5

C0 C1 C2 C3 C4 C5

All nine missing data packets are successfully recovered

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 69
 First-Line of Defense in Loss Repair
1-D/2-D Parity Forward Error Correction
Retransmission Server
(w/ Feedback Target) STB

STB

DSLAM
Transport & STB
Access
Distribution Network
STB
RTP Source Aggregation
Router

Channel 1 DSLAM
FEC for Channel 1 STB

 Each TV channel may be associated with one or more FEC streams

FEC streams may have different repair capabilities
IP STBs may join the respective multicast sessions to receive FEC stream(s)

 General Remarks
 FEC scales extremely well with upfront planning, easily repairs spatially correlated losses
 Longer outages require larger overhead or larger block sizes (More delay)
 FEC requires encoding/decoding operations

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 70
 Second-Line of Defense in Loss Repair
RTP Retransmissions
Retransmission Server
(w/ Feedback Target) STB

STB

DSLAM
Transport & STB
Access
Distribution Network
STB
RTP Source Aggregation
Router

Channel 1 DSLAM
Feedback and retransmissions for Channel 1 STB

 There is a (logical) feedback target for each TV channel on the retransmission server
If optional FEC cannot repair missing packets, IP STB sends an RTCP NACK to report missing packets
Retransmission server pulls the requested packets out of the cache and retransmits them
The retransmission is on a separate unicast RTP session

 General Remarks
 Retransmission recovers only the lost packets, so no bandwidth is wasted
 Retransmission adds a delay of destination-to-source-to-destination

 Protocol suite comprises RFC 3550, 4585, 4588 and RTCP SSM

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 71
 Further Reading

 Articles
“IPTV service assurance,” IEEE Communications Magazine, Sept. 2006
“DSL spectrum management standard,” IEEE Communications Magazine, Nov.
2002
“Error control for IPTV over xDSL networks,” IEEE CCNC 2008

 Related Specifications
[Link]
[Link]
[Link]
[Link]
“Asymmetric digital subscriber line (ADSL) transceivers,” ITU-T Rec. G.992.1,
1999

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 72
 Improving Viewer Quality of Experience

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 TV Viewers Love Zapping
Results are Based on 227K+ Users in NA

Min 1
Mean 726
Std 814
90th Percentile 1672
95th Percentile 2250
99th Percentile 3798
Max 24186

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 74
 Zappings are Correlated in Temporal Domain
On a Sunday between 8:00 – 9:00 PM

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 75
 Delay Elements in Multicast MPEG2-TS Video

 Multicast Switching Delay

IGMP joins and leaves
Route establishment (Generally well-bounded)
 Reference Information Latency
PSI (PAT/CAT/PMT) acquisition delay
CAS (ECM) delay
RAP acquisition delay
 Buffering Delays
Loss-repair, de-jittering, application buffering
MPEG decoder buffering

Reference information latency and buffering delays are

more critical in MPEG-based AV applications

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 76
 Typical Zapping Times on DSL IPTV

Unit Time Total Time

IP STB sends IGMP Leave < 100 ms

IP STB sends IGMP Join < 100 ms

DSLAM gets IGMP Leave < 100 ms

DSLAM gets IGMP Join < 100 ms ~ 200 ms

DSLAM switches streams 50 ms ~ 250 ms

Latency on DSL line ~ 10 ms ~ 260 ms

IP STB receives PAT/PMT ~ 150 ms ~ 400 ms

Buffering

De-jittering buffer ~ 150 ms ~ 550 ms

Wait for CA < 50 ms ~ 600 ms

Wait for I-frame 0–3s 0.5 – 3.5 s

MPEG decoding buffer 1–2s 1.5 – 5.5 s

Decoding < 50 ms 1.5 – 5.5 s

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 77
 A Typical Multicast Join

Time the IP STB needs to wait

to start processing multicast data
Time
M u l t i c a s t D a t a
RAP (1) Join RAP

IP STB

RAPs might be far away from each other

RAP data might be large in size and non-contiguous

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 78
 Concurrent Multicast Join and Retransmission

Data the IP STB needs to get

from the retransmission server
Time
M u l t i c a s t D a t a
RAP (1) Join RAP

(1) Retransmission request

Retransmission
Server
IP STB (2) Unicast retransmission

If the residual bandwidth remaining from the multicast stream is small,

retransmission may not be able to provide any acceleration

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 79
 Retransmission Followed by Multicast Join

Data the IP STB needs to get

from the retransmission server
Time
M u l t i c a s t D a t a
RAP (3) Join RAP

(1) Retransmission request

Retransmission
Server
IP STB (2) Unicast retransmission

More data are retransmitted due to deferred multicast join

However, IP STB ultimately achieves a faster acquisition

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 80
 Proposed Solution
Unicast-Based Rapid Acquisition
 IP STB says to the retransmission server:
“I have no synch with the stream. Send me a repair burst that will get me back on the track
with the multicast session”

 Retransmission server may need to

Parse data from earlier in the stream than it is needed for retransmission
Burst faster than real time
Coordinate the time for multicast join and ending the burst
 This solution
Is applicable to any RTP-encapsulated multicast stream
Uses the existing toolkit for repairing packet losses in multicast sessions
RFC 3550 (RTP/RTCP)
RFC 4585 (RTP/AVPF)
RFC 4588 (RTP Retransmissions)
RTCP SSM (RTCP Extensions for SSM – with the RFC Editor)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 81
 Unicast-Based Rapid Acquisition
[Link]

Leave Ch. X

Acquisition request for Ch. Y

Retransmission
Downstream
Channels
Source

Server

IP STB
Decoder priming, join time, burst description

Unicast Burst Join Ch. Y

Merge & Discard duplicate

Multicast Stream

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 82
 How to Prime the MPEG Decoder?
[Link]
Retransmission Server
MPEG2-TS Parser

RAP Cache for Channel #8

PAT PAT PAT TSRAP #N TSRAP #1
for Channel #8
MPEG2-TS

PMT PMT PAT PMT CA CA PAT PMT CA CA

CA CA CA CA

CA CA CA CA

V V V MPEG Data Cache

for Channel #8

 Transport Stream Random Access Point (TSRAP) may include

PAT: Program Association Table
PMT: Program Map Table
PCR: Program Clock Reference used to initialize the decoder and STB clocks
SEQ: Sequence Header (MPEG2 video)
SPS: Sequence Parameter Set (H.264 video)
PPS: Picture Parameter Set (H.264 video)
ECM: Entitlement Control Messages

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 83
 Experimental Setup

 Comparison
One IP STB with non-accelerated channel changes
One IP STB with accelerated channel changes

 Video Streams
Encoded with AVC at 2 Mbps and 30 fps
One stream with 15 frames per GoP (Short-GoP)
One stream with 60 frames per GoP (Long-GoP)

 Transport
1356-byte RTP packets (7 TS packets plus RTP/UDP/IPv4 headers)
20% additional bandwidth consumption for bursting
500 ms loss-repair buffer in each IP STB

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 84
 Short-GoP Results

~65% Reduction

Min Mean Std 95th 99th Max

Non-accelerated 1323 2785 645 3788 4101 4140

Accelerated 501 1009 260 1345 1457 1965

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 85
 Long-GoP Results

~65% Reduction

Min Mean Std 95th 99th Max

Non-accelerated 1831 3005 575 3920 4201 4300

Accelerated 536 1013 265 1377 1521 1937

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 86
 When Acceleration is Disabled

Channel change times depend on

- The time of zapping request
- The frequency of the TSRAP points
 Inconsistent zapping times

TSRAP Acquisition Delay: Time for IP STB to receive all TS-related information
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 87
 When Acceleration is Enabled

Loss-repair buffer size  Loss-repair buffer size + TSRAP period

Channel change times DO NOT depend on

- The time of zapping request
- The frequency of the TSRAP points
No correlation  Consistent zapping times

No correlation

Age of TSRAP: Denotes how far TSRAP is behind multicast session when the burst starts
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 88
 Companion-Stream Approach
The Basic Idea

Low-quality
Multicast Stream
I I I I

Join Splice and leave the multicast session

High-quality
Multicast Stream I I

Quality Issues Continue with Full

due to Drift High Quality
Zapping

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 89
 VQE QoS/QoE Monitoring

North-bound Interfaces
 VQE-S collects RTCP reports and outputs
them to the management application
 Management application
Collects raw data from exporter Web
Interface
Organizes database
Conducts data analysis, trends
TCP Export Interface
Create alerts

 Management application supports

standards-based north-bound interfaces
 Reports and analysis can be granular to
Regions, edge routers
DSLAMs, access lines RTCP RTCP
Reports Reports
Home gateways
Set-tops

 Set-tops can support RTCP reporting and

TR-069 (or TR-135) concurrently

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 90
 RTCP Sender/Receiver/Extended Reports

 RTCP Sender Reports provide info on data sent recently

Wallclock time and the corresponding RTP timestamp
Total number of packets/bytes sent
 RTCP Receiver Reports summarize the reception quality
Timestamp of (and delay from) the last received sender report
Highest sequence number seen so far
Number and fraction of the lost RTP packets
Estimate of the interarrival jitter
 RTCP Extended Reports (XR) provide
Detailed transport-level stats and application-specific information about the RTP transport
Several advantages over traditional and proprietary monitoring solutions
 RTCP XR framework is easily extensible to report on
Packet-level loss events, loss patterns, mean time between losses, loss durations, etc.
 Correlation engines identify, characterize and isolate the problems
Audiovisual reception quality
Effectiveness of the loss-repair methods
 Loss-repair methods can be adapted and improved depending on the network conditions
Effectiveness of channel change acceleration

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 91
 RTCP Extended Reports (XR)
[Link]

V=2 P RC PT=RR=201 Length

SSRC of Packet Sender
SSRC of Distribution Source
Fraction Lost Cumulative Number of Packets Lost
Extended Highest Sequence Number Received
Interarrival Jitter
Last SR (LSR)
Delay since Last SR (DLSR)
V=2 P SC PT=SDES=202 Length
SSRC/CSRC_1
CNAME=1 Length Canonical Name (MAC Address)
…
V=2 P Rsvd. PT=XR=207 Length
SSRC
BT Type Specific Block Length
Type-specific Block Contents
…

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 92
 RTCP XR Example: Loss RLE Reports
[Link]

RTP Receiver

Application
Post-repair
Pre-repair
Received/recovered
Loss-repair
Buffer

Buffer
packets
Source and repair data Methods
#1, #2, …, #K

Pre-repair Loss RLE Post-repair Loss RLE

(RFC 3611)

The difference tells us the aggregated

performance of the loss-repair methods

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 93
 Open Source Implementation for VQE Clients

 Client-side implementation is available as open source:

Documentation

[Link]

FTP Access

[Link]

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 94
 VQE – Summary

 Designed with both video and network considerations

Scalability
CAC and QoS
Multicast
High availability
 An open and standards-based solution
Highly extensible
Better interoperability
 Offers hybrid loss-repair and rapid channel change solutions
Improves customer satisfaction
Expands the IPTV coverage area
 Provides end-to-end monitoring capability of individual IP STBs
Reduces costly help-desk calls and truck rolls
Helps isolate the source of the problem

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 95
 Main IPTV Standards Organizations
SDO Alignment is a Challenge for Interoperability

 ITU-T IPTV Global Standards Initiative (GSI)

Many study groups (9, 11, 13, 16…) developing IPTV standards

 ATIS IPTV Interoperability Forum (IIF)

Architecture, DRM, metadata and interoperability

 ETSI TISPAN
IMS, IPTV architecture and QoS/resource management

 Digital Video Broadcast (DVB) Project

Developing standards for IPTV based on MPEG and IETF protocols

 Internet Engineering Task Force (IETF)

Developing base protocols for IPTV transport and control planes

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 96
 IPTV-Related Standards Organizations

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 97
 Further Reading

 Articles
“Reducing channel-change times with the real-time transport protocol,” IEEE
Internet Computing, May/June 2009
“On the scalability of RTCP-based network tomography for IPTV services,” IEEE
CCNC 2010
“On the use of RTP for monitoring and fault isolation in IPTV,” IEEE Network, to
appear

 Related Specifications
[Link]
[Link]

 Industry Tests
Light Reading: IPTV & Digital Video QoE: Test & Measurement Update
[Link]

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 98
 Over-the-Top Services and Internet Video

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 Studios are Going Directly to Consumers
The Lines are Blurring between TV and the Web

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 100
 NBC’s Beijing Olympics Success
More than 1100 Years of Video Streamed in 17 Days

Video Streams: 75.5 Million

10.8M for Athens & Torino games combined (+601%)
Unique Users: 51.9 Million
25.2M for Athens and Torino games combined (+106%)
Page Views: 1.24 Billion
561.1M for Athens and Torino games combined (+122%)

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 101
 Increasing W3C Applications on TV

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 102
 Netflix
The Power of Recommendation

Content 41% of DVD spending is on films with < $30M box-office

72M DVDs with 100K titles - Licensing fees are based on box-office revenues

10M DVDs processed per week Top-rented movies (2007) were not top 20 box-office hits

Shipped 1 billionth movie in 02/07 - Subscribers chose specialty films against all new releases

Shipped 2 billionth movie in 04/09

XBOX
1M subscribers by 03/09
12K streaming content choices
1.5B content minutes by 06/09
Revenue
Quarterly revenue of $400M
32% gross margin Plans
10M subscribers (4% churn) 1 DVD out at-a-time (2 DVDs a month) for $4.99
1 DVD out at-a-time (unlimited) for $8.99
- Most popular
- Highest gross margin plan
~1% of subscribers change plan after signup

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 103
 YouTube

Summary Traffic Volume & Distribution

 #1 UGC site worldwide  ~25B videos/month
 Complemented with professional  250M+ unique viewers
content from CBS, Disney, BBC, etc.
 ~2 minutes/video
 Devices
 40% video share in many markets
[Link] (PC)
[Link] (Mobile)  Upload of ~20M videos per month
[Link]/xl (TV)
 Bitrates of 200-2000 Kbps
 Ad-funded business model
 Peak rate: ~1 Tbps
 Revenue of $100M (2008), not yet
 Volume: ~125 PB/month
profitable
 CDN: GoogleNet

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 104
 BBC iPlayer

Summary Traffic Volume & Distribution

 Only available in the UK  ~60M videos/month
 Catch-up TV (400 hours/week) & live  2M+ unique viewers
TV
 ~22 minutes/video
 Devices
 Streaming:download = 10:1
PC and Mac
Mobile phones, iPod and iPhone  Bitrates of 160-3200 Kbps
TV with Wii, PS3
 Peak rate: 75-100 Gbps @ 10 pm
 Gets a public license fund, i.e., 60 Gbps for streaming
consumers are funding the service 15 Gbps for download
1.5 Gbps for iPhone

 Volume: 7 PB/month (April 2009)

 CDN: Akamai, Limelight, Level3

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 105
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 106
 iPlayer Monthly Summary
October 2009

 Total Requests
53.2M for TV programs
Only 4% of the requests were for live streams
Only 3% were for download
26.1M for radio programs
67% of the requests were for live streams

 Devices
85% from computers, 7% from iPhone/iPod Touch and 6% from PS3

 Viewer Profile
Male/female  1
Majority is under 55 (Younger than the typical TV viewer or radio listener)

 Viewing Time
For TV, iPlayer is used at roughly the same time of day as linear TV viewing
There is proportionally more daytime and late-peak use
For radio, iPlayer is far more popular in daytime than traditional radio listening

Source: BBC iStats

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 107
 iPlayer – Monthly Online Requests (2009)

90
80
Monthly Requests (Million)

70
60
50
40
30
20
10
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct
TV Radio
Source: BBC iStats
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 108
 iPlayer – Average Daily Users (2009)

1.2

1
User Agents (Million)

0.8

0.6

0.4

0.2

0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct
Users of TV Only Users of Both Users of Radio Only
Source: BBC iStats
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 109
 Hulu

Summary Traffic Volume & Distribution

 Only available in the US  ~400M videos/month
 Catch-up TV (1000+ shows) & movies  40M+ unique viewers
(500+) from NBC, FOX, ABC (Hulu
 ~7 minutes/video
owners) and other >100 content
partners  Bitrates of 480-2500 Kbps

 Devices  Peak rate: 160 Gbps

Primarily PC and Mac
 Volume: 15 PB/month
iPhone (expected soon)
TV support via installing software on PC  CDN: Akamai
 Ad-funded business model
 Revenue of $65M (2008), not yet
profitable

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 110
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 111
 Internet Video in the US
July 2009

Videos Viewed (x1000) Share (%)

Google Sites 8,953,948 41.9

Viacom Digital 812,343 3.8

Microsoft Sites 630,631 3

Fox Interactive Media 558,500 2.6

Hulu 457,010 2.1

Turner Network 390,848 1.8

Yahoo! Sites 374,746 1.8

Disney Online 169,756 0.8

CBS Interactive 150,165 0.7

ABC Television 137,800 0.6

Total Internet 21,371,342 100

Source: comScore
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 112
 Internet Video in the US
July 2009

Unique Viewers Average Videos per

(x1000) Viewer
Google Sites 120,852 74.1

Microsoft Sites 64,540 9.8

Fox Interactive Media 51,830 10.8

Yahoo! Sites 47,363 7.9

Viacom Digital 42,415 19.2

Hulu 38,132 12

Turner Network 35,676 11

CBS Interactive 30,736 4.9

AOL LLC 24,161 5.2

[Link] 20,517 4

Total Internet 158,384 134.9

Source: comScore
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 113
 Many Internet Video Competitors are Emerging

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 114
 Internet Video Delivery – CDN Market

 Internet video CDN revenues are growing ~30% each year

It will reach $0.6B in 2009

 Most Internet video providers outsource delivery

A few very large content providers build their own CDN infrastructure
E.g., Google, Microsoft, Amazon
Only a few network service providers will compete in the global CDN market

 Traffic growth is driven by

More users
More content
Longer viewing at higher qualities

 CDN providers are diversifying into value-added services

E.g., Akamai and Limelight are getting into advertising business

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 115
 Internet Video CDN Services – Main Providers

 Akamai
Leading CDN provider with 40K servers in 1500 global sites and on-net caches in 900 ISPs
Has a CDN capacity of ~3 Tbps
Average traffic is ~1 Tbps and historic peak is 2 Tbps

2009 video CDN revenue estimate is $250M

 Limelight Networks
Second CDN provider with a heavy focus on video
Has a global IP backbone with 60 caching sites in 20 cities
Has a CDN capacity of ~3 Tbps
2009 video CDN revenue estimate is $150M

 Level3
Leading global IP transit provider, now selling CDN as a complementary service
(Acquired CDN provider Savvis)
Has a few 10’s of global caching sites
Has a CDN capacity of ~1.5 Tbps
2009 video CDN revenue estimate is $75M

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 116
 Media Delivery over the Internet

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 Cisco Content Delivery System
The Network is the Platform
Published
Content
 Extensible Architecture
Internet Independent scalability of storage, caching
Content and streaming
Non-stop service availability
Programming
Content Convergence of live and on-demand content
Acquirers

 Distributed Network
Management
Multi-protocol centralized ingest
Caching Popularity-based multi-tier caching
Nodes Service Routing
Multi-protocol decentralized streaming

 Service Routing Functionality

Streamers
Service routing at the edge or headend
Global and local load balancing

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 118
 Session Shifting
Pause and Resume Across Devices

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 119
 Virtual Video Infrastructure (VVI)

Persistent Storage
1000
(National Level)

Core Bandwidth (Gbps)

w/o VVI
750
w/ VVI

CORE 500
Caching Nodes
(Regional Level) 250

0
20K Hours 100K Hours
Content Size

Based on
Streamers
(Market Level)
EDGE
% of HD content is 10%
# of markets is 80
Average # of streams per market is 10K
3.9 Tbps
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 120
 Progressive Download vs. Streaming

HTTP Request

HTTP Response

Progressive Download

Play, stop, pause, seek

Media Packets

Streaming
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 121
 Progressive Download vs. Streaming

Progressive Download Streaming

Server Type Web server Streaming server
Apache, Microsoft IIS Windows Media, Adobe Flash,
Apple QuickTime, Cisco CDS-IS
URL Format [Link] rtsp://…, rmtp://…

Protocols HTTP RTSP, RTMP, RTP

Bandwidth Usage Less efficient More efficient
Downloaded as fast as Streamed as requested at the
bandwidth allows encoding rate
Client Buffering Yes Yes
Monitoring and User Tracking No Yes

Multicast Support No Yes

Play, stop, pause, fast seek, fast
Client Controls Play, stop, pause
forward, rewind
Client-side Storage Yes No

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 122
 Adaptive Streaming

 A Hybrid of Progressive Download and Streaming

Imitates streaming via short progressive downloads
Clients download progressively in small chunks so bandwidth is not wasted
Clients stream so we can monitor consumption and track them

 Use of HTTP
Provides easy NAT traversal
Leverages existing HTTP caching infrastructure (Cheaper CDN costs)

 Adaptation to Dynamic Conditions and Device Capabilities

Adapts to dynamic Internet conditions and home-network contentions
Adapts to display resolution, CPU and memory resources

 Improved Quality of Experience

Enables faster start-up, seeking and quicker buffer fills
Reduces skips, freezes and stutters

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 123
 Adaptive Streaming with One TCP Connection
Adaptation
Logic

Content

Movie A – 200 Kbps

Movie A – 400 Kbps

... Start quickly
Movie A – 1.2 Mbps
Improve quality
...
Movie A – 2.2 Mbps Conditions are good
...

Movie K – 200 Kbps

Slow down
...
Movie K – 400 Kbps
Chunks Improve quality
...

...
Movie K – 1.2 Mbps
...
Movie K – 2.2 Mbps
Get HD

Time (s)
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 124
 Adaptive Streaming with Multiple TCP Connections
Similar to Driving on a Multi-Lane Highway

TCP 1

TCP 2

TCP 3

Using multiple concurrent TCP connections helps mitigate the HOL blocking
and allows to fetch multiple (sub-)fragments in parallel

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 125
 Smooth Streaming – File Format

Multi-Bitrate Encoding
- Content is encoded at multiple bitrates
- Storage is one file (MP4) per bitrate
Fragments
- Each GoP is stored in a Movie Fragment box
- Fragments are independently decodable
- Fragments are sent on the wire
Source: [Link]
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 126
 Leveraging HTTP Caches

Time
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 127
 Digital Entertainment Content Ecosystem (DECE)
Buy Once, Play Anywhere

 DECE is a U.S. limited liability company but open to international

participation
 DECE Goals
Create the best consumer experience for digital content distribution
Progressive download, burning and streaming
An open market – Choice of interoperable devices and online retailers
Develop and license specifications for ecosystems
Define container format for progressive download

 DECE builds on existing industry standards

H.264 codec
Support for multiple DRM

Source: DECE LLC ([Link]

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 128
 DECE Ecosystem Architecture
Content Owners
Enhanced value of content
Supply chain optimization
Competitive marketplace
Solution to interoperability
DECE
Coordinator Mass market for digital content

Digital Service Provider Online Retailers Digital Service Provider

A / CDN Enhanced value of service B / CDN
Consumers
Retailer LASP Lower barrier to entry Retailer
Purchasing LASP not a format
an experience,
A A Competitive marketplace B B
Choice of digital retailers
Increased device reach
Authorized sharing within household
Greater choice of interoperable devices

Domain Domain Domain

Device A Device
DD B Device C
Streaming
DECE Domain/Account Device

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 129
 DECE Participants (Partial)
Almost 50 Members by January 2010

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 130
 Further Reading

 Microsoft Smooth Streaming

[Link]

 Adobe Dynamic Streaming & Live DVR

[Link]
[Link]

 Move Adaptive Stream

[Link]

 Apple HTTP Live Streaming

[Link]

 Also see:
Swarmcast and Octoshape
Widevine Adaptive Streaming
Vidiator Dynamic Bitrate Adaptation

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 131
 Targeted (Advanced) Advertising

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public
 Targeted Advertising
Ads Delivered Based on Consumer Profile

Amiya Jain Jason Allen

California Arizona

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 133
 Broadcast vs. Cable Spot Insertion

Broadcast Networks Local Affiliate Service Provider

Video Auto.
Server Schedule
System

Ad

Broadcast Program Production MPEG Aggregation

(Baseband) Switcher Encoder & Turnaround

Cable Networks Service Provider

Ad
Server Schedule

Ad SCTE 30

Broadcast Program MPEG

(MPEG) – SCTE 35 Splicer

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 134
 What Operators Want

Leverage the Advanced Content Delivery Infrastructure

• Grow high margin revenue via advertising

Derive Incremental Value from Existing Ad Inventory

• Higher CPMs through targeting

Generate Value for Content Providers

• Deliver addressability and access to non-linear tiers
• Transaction revenue and/or inventory splits

Achieve Parity with the Internet But with the Dynamism of TV

• Dynamic, addressable, measurable

Cross-Platform Campaigns and Delivery

• TV, broadband and mobile

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 135
 The Advertising Marketplace
Trading Ad Inventory

 Inventory means the ad insertion slots available for sale to media buyers
 These slots are defined by
Content producers
Aggregators/distributors

BUY SELL

Media Content Aggregator

Advertiser Agency Consumer
Buyer Owner Distributor

Ad Content Placement Capability

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 136
 The Advertising Workflow
Campaign Management and Execution

 Advertisers plan campaigns with media buyers

 Media buyers contract with content distributors (broadcast/cable
networks) to execute their campaigns

MANAGEMENT EXECUTION

Ad Insertion
Plan Contract Manage Reporting
Decision Playout

Campaign Stewardship and Optimization

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 137
 Information Used by Campaign Manager

 Placement Opportunities
Temporal and/or spatially constrained advertising instances defined by content
owners and/or distributors

 Context
Information describing the content, e.g., title, genre and rating

 Audience
The available viewer and device information, including
Demographic
Geographic
Behavioral

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 138
 Targeting Approaches
Changing Ads from Interruptions to Information

 Contextual
Associating ads with the content being viewed
Placement based on title, genre, network, day time, location
Offline demographic correlation

 Database
Associating subscribers with service tiers via SMS integration
Deriving demographic/psychographic attributes against third-party databases

 Behavioral
Recommendation engines, affinity correlation, filtering
Deep packet inspection
Inference engines based on click-stream analysis

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 139
 End-to-End Advanced Advertising

Media Buyers

Campaign Proof of play

Placement
Manager
Op.

Ad Ad Data Capture
Context Decision Insertion & Aggregation

Behavioral
Audience
Analysis Data warehousing
Profiling

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 140
 The SCTE 130 Standard

 SCTE 130 defines a set of logical services that implement extensible

XML-based messaging communication interfaces
 SCTE 130
Promotes easy integration between advertising subsystems
Enables interoperable, multi-vendor advertising solutions

 SCTE 130 offers a private interface that integrates with the

vendor/provider-specific ecosystem components
 SCTE 130 control plane is IP-based
SCTE 130 can be extended to other content delivery platforms

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 141
 The SCTE 130 Standard

Inventory
System Back Office
Content
Store

Placement Op. Content Subscriber

Information Information Information
SCTE 130-5 SCTE 130-4 SCTE 130-6

Placement Ad Ad Campaign
Execution Management SCTE 130-3 Decision Management

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 142
 Further Reading

 SCTE Standards
SCTE 30: Digital Program Insertion Splicing API
SCTE 35: Digital Program Insertion Cueing Message for Cable
SCTE 130: Digital Program Insertion – Advertising Systems Interfaces
URL: [Link]

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 143
 Acknowledgments

A big thanks goes to

My colleagues at Cisco
Friends at IETF, DVB and other SDOs

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 144
 It's not our fault our generation has short attention spans,
Dad. We watch an appalling amount of TV.

Don't you ever, EVER talk

that way about television.

IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 145
IEEE CCNC 2010 © 2010 Cisco Systems, Inc. All rights reserved. Cisco Public 146