Upload
vongoc
View
218
Download
3
Embed Size (px)
Citation preview
“Supporting Wireless Video Growth and Trends”
4G Americas Whitepaper April 2013
IEEE CVT meeting, September 2013
Kamakshi Sridhar, Ph.D
Director, Wireless CTO
Alcatel-Lucent
Whitepaper jointly led with
Gautam Talagery
Director, Mobile Broadband, Ericsson
Contributors and Reviewers
Whitepaper benefited from
active participation and
contributions from operators
(TEF, Rogers, Lime), vendors
(ALU, Ericsson), device
manufacturer (QCOM) leading
to very good discussions
addressing different
perspectives.
Javier Lorca - Telefonica
Tony Lutz – Qualcomm
Dave Robinson – Alcatel-Lucent
Harish Viswanathan – Alcatel-
Lucent
Barry Pratt – Rogers
Cleverston Miller – Lime
Ram Venkatramani – previously
from Openwave Mobility
Karri Kuoppamaki – T-Mobile
Gautam Talagery – Ericsson
Kamakshi Sridhar - ALU
4G Americas Whitepaper
“Supporting Wireless Video Growth and Trends”
4G Americas Whitepaper April 2013
• http://www.4gamericas.org/documents/4G%20Americas%20-
Supporting%20Mobile%20Video%20Growth%20and%20Trends%20April%20
2013.pdf
www.PresentationPro.com
Goals
• Drivers and trends
• Types of video
– Short/long form,
– HTTP-Adaptive Streaming and HTTP – Progressive Download
RF and core challenges and solutions:
• Spectral efficiency, Varying channel, Mobility
• Signaling and battery issues
• Backhaul limitations
• HAS, TCP interactions, and role of client optimizations
• Video QoE metrics
• Impact of Codecs
• Architectures: HET-NET, CDN and eMBMS
• Role of Video playout buffers
• LTE QCI, and GBR
Fact: Mobile video is growing driven by smartphones and tablets, and new apps. What enables the successful, mass adoption of mobile video?
White paper examines
issues specific to video
that affect its delivery
over wireless networks.
We provide recommendations to app developers, and on bit rates for various codecs.
Market trends in Video Delivery/Market analysis
Streaming traffic during live events is high – (2009 statistics)
You-Tube and Real Time Entertainment traffic is growing Video/Audio streaming is growing
Key trends from Sandvine report, Ooyala, BBC statistics summarized to illustrate growth of mobile video.
Types of video
Video content can be classified as:
• Short form content, Long form content (based on duration)
• Operator video and OTT video (delivered by the service provider or not)
• Best Effort video and Guaranteed Bit Rate video (based on quality)
• Video for real time communication and for entertainment (based on usage)
Major video delivery technologies:
• Non-HTTP Real time streaming protocols (RTP/RTSP delivered video). To stream live video content to proprietary clients.
• HTTP Progressive Download (HTTP-PD): Access before download
• HTTP Adaptive Streaming (HAS): Video divided into chunks.
MPEG DASH (Dynamic Adaptive Streaming over HTTP) is a standard for
adaptive streaming over HTTP that replaces existing proprietary technologies
A unified standard is helpful to content publishers, who could produce one
set of files that play on all DASH-compatible devices.
Key requirements of mobile video delivery
Video flows are typically long duration flows.
Video QoE must be sustained over long durations, over RF, core.
For good video QoE, RF, backhaul and core challenges must be addressed e2e.
• Video must adapt to changes in link bandwidth.
• Video quality during handoffs must be managed
• Video quality as the cell becomes increasingly congested must be maintained.
• Must maintain the video quality (data rate, jitter specifications), fairness
– Video places demanding requirements on bit rate, jitter, and latency
• Must differentiate video quality according to subscription level
• Need to provide scalable solutions with large number of video flows
…. and competing traffic flows
• Need to consider impact of video and audio codec performance
• Must identify proper metrics to assess video QoE.
• Need to address impact of user device screen size, client on video QoE.
Impact of mobility, backhaul and
wireless core on video delivery
Data and throughput consumed by streaming
applications is High, particularly for Tablets
Paper examines: RF impact on mobile video.
– Spectral efficiency of 3G/4G cellular
technologies
– Mobility, Handovers, retransmissions, varying channel quality
Paper addresses
• Air interface latency:
– Challenge for interactive wireless - high-
quality videoconferencing or online gaming
– Progressive download video not so
sensitive
• Signaling challenges for Video - limited
• Adaptive streaming - best for battery
consumption – and network load
• Backhaul congestion impact, and role of traffic
shaping on video QoE.
• Encryption, Content rights management
Application Throughput
(Mbps)
MByte/Hour Hrs./day GB/month
0.5 0.9
1 1.7
2 3.5
4 6.9
0.5 1.4
1 2.7
2 5.4
4 10.8
0.5 3.4
1 6.8
2 13.5
4 27
0.5 6.8
1 13.5
2 27
4 54
0.5 13.5
1 27
2 54
4 108
Stereo Music 0.1 58
Small Screen Video (e.g. Feature
Phone)
0.2 90
Medium Screen Video (e.g.
Smartphone)
0.5 225
Medium Screen Video (e.g. Tablet) 1 450
Larger Screen Video (e.g. High
Quality on Tablet)
2 900
Video codecs and HTTP Adaptive Streaming (HAS)
• Popular video codecs include H.264, VP6, VP8, etc.
Codec (compression/decompression) algorithms play a key role.
• New/Upcoming codecs (HEVC): HEVC is a successor to the current state of
the art video coder - H.264/AVC video coding standard.
– With large consumption rates, HEVC would result in significant reduction
in infrastructure costs due to lower bandwidth required.
• HAS works by breaking the video into chunks, and each chunk is encoded
and sent out at an appropriate bit rate based on the channel quality as
estimated by the client. Without buffering at the client end (like in PD), HAS
is able to provide adequate QoE at the user end, without freezing video frame
• All HAS solutions have the same underlying architecture - server side
contains segmented video sequences and manifest file that describes content
properties and lists segment URLs. Segments are offered at different bit rate
levels which allow switching between bitrates when needed.
HTTP Adaptive Streaming (HAS)
• The paper also describes the BEHAVIOR OF VIDEO HAS CHUNKS WITH
TCP/IP particularly with poorly designed HAS clients.
• Issues such as Unfairness and Instability: Under Utilised TCP throughput
due to the interactions between HAS rate determining algorithm and TCP
congestion control resulting in downward spiral are described.
• We note that the important role of well designed clients and how they can
overcome these issues with HAS/TCP.
• Through subjective testing, it has been shown that HAS delivers an
acceptable video quality over a range of typical network conditions.
• A well-designed HAS client should be able to work well and share fairly with
other HTTP flows without knowing about the other devices that share the
network. The network fairly provides as much data downloading capacity as it
can to each end device. The HAS client on the device then decides which bit
rates to download at each point in time to provide the best quality experience
to the end user (i.e., no stalls or skips and a playback quality that pretty much
matches the current download rate of the HAS client).
Video Quality of Experience:
metrics and monitoring
Typical video QoE problems: Blurring, blocking, stall, skips, encoding dependencies
Must consider the impact of:
•Device size, viewing distance
•Playout issues, and varying image quality due to adaptive streaming
The algorithm used in the buffering or playout of video from a source can influence:
•Network bandwidth use; battery usage; QoE with buffered or delayed video
Metrics for characterizing video QoE – MOS: average and standard deviation
•Three classes of monitoring Video QoE as: Full/Reduced/Non Reference
•Main factors that influence end user QoE of streamed video are identified, including device and media factors and network factors.
•3GPP 26.247 parameters for HAS content.
•Work by the NGMN P-SERQU maps video chunk quality to a predicted MOS
Recommendation:
•Monitoring of Video QoE through monitoring the quality of HTTP chunks rather than individual radio and TCP parameters is preferred option.
Architectures for mobile video delivery
• Cellular (HET-NET) and WiFi offload architectures:
Macro/pico/femtos, role of WiFi
• LTE broadcast architecture –eMBMS
eMBMS can support live streaming of sports, high attach rate content such as
breaking news, and background file delivery.
• Content Delivery Networks
Distributed caches containing copies of data, to avoid server bottlenecks
Home
MMEMME
SGWSGW PGWPGW
SGSNSGSN GGSNGGSN
S11
Gx
S5/S8
SGiRxHSSHSS
S6a
S1-MME
LTE RAN Macro/
Pico/FemtoS1-U
4G PicoMacros
GnIu
RNC/SC GW
OperatorIP servicesOperator
IP services
PCRFPCRF
Home
3G RAN Macro/
Metro/Femto3G MetroMacros
To PCRF
WiFi APs ANDSFserver
4G Femto
3G Femto
CDN
eMBMS
HetNet
The main method of video optimization at the base
station is by controlling the data rate delivered to the
device.
LTE provides powerful QoS mechanisms.
End-to-end QoS support in LTE for video delivery
must consider:
• Role of LTE QCI and LTE guaranteed bit rate
bearers
• Parameter settings for HAS: GBR, MBR,
• Differentiation and fairness with GBR service
• Limitations of GBR bearer
• Adapting Video usage behavior through yield
management techniques.
LTE QCIs: 3GPP TS 23.203
Optimization and Mitigation techniques
QCI Resource Type
Priority Packet Delay
Budget (NOTE 1)
Packet Error Loss Rate
(NOTE 2)
Example Services
1
2 100 ms 10-2
Conversational Voice
2
GBR
4 150 ms 10-3
Conversational Video (Live Streaming)
3
3 50 ms 10-3
Real Time Gaming
4
5 300 ms 10-6
Non-Conversational Video (Buffered Streaming)
5
1 100 ms 10-6
IMS Signalling
6
6
300 ms
10-6
Video (Buffered Streaming) TCP-based (e.g., www, e-mal, chat, ftp, p2p file sharing, progressive video, etc.)
7
Non-GBR 7
100 ms
10-3
Voice, Video (Live Streaming) Interactive Gaming
8
8
300 ms
10-6
Video (Buffered Streaming) TCP-based (e.g., www, e-mail, chat, ftp, p2p file
9
9 sharing, progressive video, etc.)
Non-RF specific optimizations for video delivery
• Transcoding – Adapting encoding of video stream to available bandwidth at the device.
• Transrating - Changing resolution, fidelity, frame rate without full decode / encode
• Rate capping – An intermediate node adjusts the sending rate of a progressively downloaded video
• Scalable Video Codecs (SVC): The content encoded using a base layer and one or more enhancement layers
• Content caching in the network … useful but does not solve RAN congestion.
• Time shifting or side loading (over WiFi) supported by content caching in the UE – downloading, caching video content to the device during off-peak times or when the device is connected to WiFi.
• Congestion Triggered Optimization (for HTTP-PD) - Video Optimization engine dynamically starts optimizing (transrating/transcoding) video traffic during periods of congestion.
• Microbursting - High video data rate, within a short time to fill up the client buffers
Video delivery with CDNs
Content capture and ingestion.
• Video is compressed and transmitted to content ingestion sites
• Coding the video into multiple versions for different codec technologies, container formats, resolutions, and content variances (e.g. trailers or short clips, embedded captions) etc.
• Application of Digital Rights Management technology adopted.
• Access policies, restrictions control access to the published contents and advertisement placements.
Content distribution to local video headends follows the content ingestion:
• Video created sent over high speed networks to different video headends.
• Video delivery begins in video head end with supporting caching CDN nodes.
• Transcoding content into format appropriate for a particular device.
• The CDN server processes the request from end device, extracts device information from a subscription data base and delivers the video, transcoding if necessary.
• The delivery server maintains the session with the end device streaming the video as the end device consumes the content.
• The video server may additional invoke the APIs provided by the mobile core to establish QoS for the session.
Recommendations on bit rates for various codecs and screen sizes/device types
• The choice of bit rate for the audio and video streams is a compromise between
quality and bandwidth used for streaming. The device screen size and viewing
distance are a critical aspect in determining the bandwidth needed for encoding the
video.
– Scenes with high motion or scenes with a lot of detail require more bandwidth.
– Chaotic motion is the hardest to encode and requires higher bandwidth
• Compression efficiency of the video codec impacts the bandwidth needed to encode a
video quality with a given fidelity.
• The Content Provider may encode their content in multiple forms
Video Quality
Level
Video (kbps) Total (kbps) Comments
VQ1 200 296 Smartphone
VQ2 400 496 Tablet / Smartphone
VQ3 800 896 Tablet / Smartphone
VQ4 1400 1496 SD 480P TV screen through games console and
STBs, Connected TVs, PCs and Tablets
VQ5 3150 3246 HD 72P TV screen through games console and
STBs, Connected TVs, PCs and tablets
VQ6 7100 7196 Full HD 1080P, TV screen through games console
and STB, Connected TVs, PCs with General
Processor Units.
Video Quality
Level
Video (kbps) Total (kbps) Comments
VQ1 200 296 Smartphone
VQ2 400 496 Tablet / Smartphone
VQ3 800 896 Tablet / Smartphone
VQ4 1400 1496 SD 480P TV screen through games console and
STBs, Connected TVs, PCs and Tablets
VQ5 3150 3246 HD 72P TV screen through games console and
STBs, Connected TVs, PCs and tablets
VQ6 7100 7196 Full HD 1080P, TV screen through games console
and STB, Connected TVs, PCs with General
Processor Units.
Recommendations for video application developers
• Both the video delivery algorithm and the network configuration must be optimized.
– Video application developers need to consider the challenges in operating over a network with mobile devices.
– Applications must be adaptable to changing access capabilities, and access types.
• Utilize APIs between the Application Layer and Access
– Drive the creation of APIs to demand QoS from the network
– Mobile operators could provide the development space and environment for realistic simulations
• Chunk size: Ensure that video servers deliver chunks of video in small regular intervals ~10 seconds.
– The most optimal value of chunk size depends on, amongst other things, the eNodeB timer value settings. Video sources with significant inactivity periods between chunks of data can waste signaling resources and device battery.
• Ensure that polling requests from the devices to the network are bundled to maximize use of seized resources.
Summary & Conclusions
• For Operators
– Architecture choices and improvements in both access and content.
– Work smartly with available and forecasted mobile broadband.
• For Vendors
– Areas for cooperation, integration in content generation and delivery
– Improvement and optimization in devices and networks.
• For Application developers and architects
– Optimize design of video applications to minimize effect on mobile
networks
• For Industry analysts
– The state of video on mobile broadband is described, trends identified.
• For Device manufacturers
– Evolution in video processing technology and codecs.