Building Unicast IPTV services leveraging
OTT streaming technology and adaptive
Fraunhofer FOKUS & Zattoo
May 19th 2014
 Motivation
Internet delivered Video is at the tipping point and became a mass-market driver
for over the top (OTT) video solutions. The massive growth of consumed Internet
video (Figure 1) was enabled by widely available broadband access in households
providing sufficient and affordable bandwidth. Internet services became commodity. In Germany by end of 2012 55% of sold TV devices are hybrid1, meaning they
combine traditional DVB access with OTT services from the open Internet. Devices
for video consumption became more powerful, ubiquitous and easy to use –there
are more than 13 million SmartTVs available in German households, which is already more than 30% of the overall TV penetration2. Furthermore, open Web
standards and tools evolved massively. HTML5, JavaScript and CSS is omnipresent
and one of the key driver for the success of video enabled web applications and
services. YouTube alone accounts for more than one billion unique users visiting
the site per month, resulting in more than six billion hours of video watched3. While
new content is provided in high quality, client-sided limitations do not always allow
for consumption of the best available quality at any given time. Variations in available bandwidth can easily lead to buffer underruns and video freezes while new
content has to be fetched to resume playback. Alternatively, a lower quality setting
could be used to prevent such freezes, which might well lead to not using the best
possible quality most of the time.
One approach to vanquish these issues is the introduction of adaptive bitrate
streaming. Adaptive streaming focuses on selecting the most appropriate quality
media in real-time, based upon evaluation of criteria that might impede and limit
the given possibilities, such as available bandwidth and CPU capacity on the client
side. Therefore, video content needs to be provided in multiple bitrates and resolutions allowing media clients to request and playback the most appropriate quality
adaptively. Most relevant adaptive streaming solutions are the proprietary solutions
Apple HTTP Live Streaming (HLS), Adobe HTTP Dynamic Streaming (HDS), Microsoft Smooth Streaming and the standardized MPEG Dynamic Streaming over
A second approach, mainly for First Screen IP-based TV services (IPTV) in operator
managed networks, is to leverage direct network peering between a service provider or service operator and the network operator. With the network operator
managing the distributed content he receives through the direct network peering,
AGF - TV viewing data 2012
Digital TV viewing evolution
2 | 10
he makes sure that the IPTV service is prioritized and can be streamed at the maximum bitrates and quality available.
This Unicast IPTV approach has massive consequences: Traditional IPTV services are
built on multicast-enabled managed networks operated by large telecommunication or network providers. As experience has shown, such deployments were
bound to heavy investments in the network infrastructure, end devices as well as
operations to enable a managed IP-based multicast transmission of TV content to
households. New IP-based video services need to be highly flexible, cost effective and prepared to be available an all major platforms and end devices
without the need for specific hardware or network prerequisites. These
new Unicast IPTV services build on HTTP-based, adaptive video delivery via
direct network peering or the open Internet using unicast delivery.
Figure 1 – Consumer Internet Video 2011-2016 in PB/month (visual networking index 2013
by CISCO4)
The Cisco® Visual Networking Index (VNI) Global Mobile Data Traffic Forecast
3 | 10
 Setting the scene – a technical
Video streaming generally consists of an audio and video source fed into an encoder, transmitted to a media server, which then supplies connected clients with streaming data. In
principle, there are two different models to transport the audio and video packets: unicast
and multicast delivery:
Unicast offers a one-to-one relationship between the server and the client receiving the
stream – Multicast on the other hand offers a one-to-many relationship (see Figure 2). It
serves a single stream replicated throughout the network so that any client may tune in.
Thus, it seems that multicast should be great for live video delivery while unicast might be
the better choice for video on demand. That is theory – in practice there is no black and
white. As multicast is based on UDP, it is a non-connection oriented protocol
providing a best effort delivery of media packets without native support of packet
and flow control. Furthermore, multicast lack of TCP windowing and slow-start mechanisms may result in network congestion. Enhanced efficiency through the control of network traffic, reduction of server load or e.g. elimination of traffic redundancy by multicast
delivery is feasible but will in practice require additional multicast enabled hardware,
complex network management and does usually come with high costs.
By nature, the open Internet is not multicast ready since it is built for unicast delivery using connection oriented protocols as TCP. With the advent of HTTP based
streaming protocols, video delivery can be built into every network without any
need for specialized hard or software, complex backbone infrastructures and cost intensive network management. This even allows for IPTV services (Live TV, Catch-up, VOD)
targeting the First Screen. Prioritization of video traffic over other, non-time-critical traffic
can easily be achieved by standard network management mechanisms through e.g. load
balancing and routing rules. One of the main benefits of unicast delivery is the fact, that
there is absolutely no need for special hardware and protocol support at the client
side as e.g. multicast capable routers or set-top-boxes.
Adaptive streaming even allows compensating for changing network conditions while still
providing users with the decent quality. No bandwidth needs to be dedicated to specific
video streams as streams consists of small, ordered media segments instead of a continuous
stream. Adaptive streaming enabled clients splice these segments together to a video
stream as defined in the appropriate manifest for each content. Unicast is supported by
any IP-enabled client-device and the support of adaptive streaming through e.g. HLS, HDS,
Smooth Streaming or MPEG DASH is available on all major client platforms. This means
that all kinds of clients can be served with live or on demand streams using the
exact same technology. To recap, this is fundamental different to multicast delivery and
multicast enabled clients as special set-top-boxes. From a service perspective, there is an
additional benefit for unicast: as unicast by nature is built for individual consumption, it
perfectly builds the basis to provide personalized streaming services as on demand
assets, catch-up TV, and time shifted content as well as network recording capabilities.
4 | 10
Figure 2 – Unicast vs. Multicast
Adaptive streaming is the concept of adjusting the streamed media to the needs of the
current client by enabling the client to select a media representation with appropriate spatial resolution and bitrate for the current given network connection to ensure a continuous,
non-sticking media playback with best possible quality. Therefore, the media content needs
to be available in different quality levels constituting different available bandwidths. They
are called media representation and traditionally range from small spatial resolutions and
bitrates for mobile delivery to full high definition media representations. This allows to
distribute e.g. maximum resolution, bitrate and fast startup to IPTV Set-Top-Boxes that are
connected via Ethernet or via a stable home WiFi connection, while providing lower resolution and bitrate streams to second screens that are currently being used in a weak WiFi
or in a mobile (e.g. 3G) network.
To allow a smooth switching between the different media representations and a seamless
media playback at the client, the content is split into small segments of 2-10 seconds for
each quality level containing one or multiple fragments. All current adaptive streaming
technologies use media that is represented in a fragmented format to allow the client to
effectively fetch a new media representation after a switching event through e.g. increase
or decrease of the available bandwidth. Therefore, each representation is split into corresponding fragments that represent specific ranges of the media timeline. As a result, adaptive streaming technologies as HTTP Dynamic Streaming (HDS), HTTP Live Streaming (HLS),
Smooth Streaming and MPEG DASH (see Figure 3) provide mechanism to deliver live and
on demand video using HTTP unicast in an effective manner.
Furthermore, streaming over HTTP perfectly supports the possibility of caching media data using Web caches. Both the delivery of Live-TV and on demand media content
benefits from caching mechanism. Well placed Caches can significantly reduce load on the
backbone, especially during peak times when many users consume the same content.
These edge serves are equipped with maximum I/O performance to serve content to a
maximum of users by simultaneously discharge the original server´s load. State of the art
5 | 10
content delivery networks (CDNs) also follow this concept to hide load from origin servers,
balance network traffic between multiple caches and relay servers.
Besides the described effective unicast approach, there are several other solutions under
evaluation by CDN providers using e.g. multicast for inner-network content distribution.
These approaches might be useful in special settings to deliver live content, but also massively increase the complexity of the given system. Converting between multicast and
unicast, chunking/on the fly and translating of content on the edge servers requires additional hardware and ultimately increases costs.
Figure 3 – Scope of MPEG-DASH
6 | 10
 Building an E2E video platform
using unicast delivery
As shown in the previous sections, HTTP based adaptive streaming over unicast builds
the foundation for new IPTV and OTT video services. Figure 4 depicts a highly optimized system setup for such a Unicast IPTV / OTT video solution comprising of ingest servers, storage, encoders, transcoders, a video middleware as well as direct peerings with ISPs
and an external CDN to deliver live and on demand video via HTTP adaptive streaming over unicast to a wide range of clients. The ingest servers receive source signals
from e.g. satellite or via IP and feed the signals into the encoders where each stream is
encoded to H.264 in highest bitrate and subsequently segmented. The segments are packaged in a universal format and stored in highest quality for further processing, catchup and
individual recordings via network PVR functionality of clients. Translating of live and on
demand segments to smaller bitrates and spatial resolutions is done by the Transcoder farm
on the fly. In a subsequent step, the segments are packaged (muxed) in the appropriate
format relying on the adaptive streaming format that has been requested by the client.
Finally, the playlist or manifest files that describe the media representations are generated.
Additional steps are the encryption of media segments, access control and caching of fragments. To deliver streaming content to requesting clients, content delivery servers (origins)
are connected through direct peerings with ISP networks. Established CDNs can be easily
leveraged as backup routes or to compensate traffic peaks if needed. Additional simple
Web caches can be integrated in strategic spots of the ISP network to significantly
reduce load on the backbone. This does also apply for managed networks e.g. within
the network of connected ISPs.
7 | 10
Figure 4 – OTT video service overview
As all components are based on standard of-the-shelf hardware running the different functions as software components, cost-effective future extensions, scalability
and flexibility is one of the key benefits compared to traditional solutions. Caches
and edge servers handle the increasing demand for bandwidth. Despite the resulting increase of unicast traffic in the network, the load can still be managed effectively with the
same caching mechanisms. This fact is being favored by the development and introduction
of next generation networks, which are built for Tbps network throughput as well as the
introduction of next generation video codecs as e.g. H.265. Current codecs typically allow
for a video quality comparable to traditional DVB quality with 1.5 Mbps for standard definition or 3Mbps for high definition content – new codecs will likely cut that in half. Unicast
IPTV setups where the network operator leverages his own network and bandwidth, potentially in combination with caching servers, clearly allow for even higher bitrates, especially for First Screen IPTV services.
8 | 10
 Summary and Outlook
To sum up, unicast IPTV services (live and on-demand) guarantee the same high quality
user experience as traditional multicast-based architectures when leveraging appropriate
technology and techniques.
First Screen IPTV services delivered over unicast quality-wise benefit from direct network
peering with the network operator. Furthermore, the network operator is able to prioritize
the network traffic that is caused by the IPTV service.
HTTP Adaptive Streaming is the foundation for multi-screen video delivery. By taking advantage of existing technology all major devices (e.g. second screens) and platforms can
be reached – without any special hardware. Moreover, streaming efficiency and scalability
can be significantly increased using strategically positioned Web caches.
In combination the aforementioned techniques lead to the possibility of streaming at
maximum bitrate and quality, while lowering
complexity and costs compared to traditional
multicast-based architectures.
Figure 5 depicts a real data analysis provided
by Zattoo. Over a period of a week more than
80% of the users of the OTT distributed B2C
Zattoo service were able to consume the
streams at the best possible quality level. Optimization in the delivery architecture (Web
caches and peering), improvements in bitrate
switching algorithms and developments in
emerging standards such as MPEG DASH will
lead to even better numbers. Even today, in
B2B co-operations where Zattoo leverages direct network peering to hand over ready-todistribute unicast streams to network operators, close to 100% of users are able to
consume the maximum bitrate streams when
using the service within the operator’s network.
Figure 5 – Maximum bitrate graph
The following testimonials are provided by network operators that use Zattoo technology:
Andreas Cavegn (Product Management, Telecom Liechtenstein AG, Vaduz) on reasons for
using Unicast for IPTV:
„Telecom Liechtenstein already deployed a Unicast IPTV Solution in 2009 and is now in the
process of rolling out a new Unicast IPTV solution which will be fully hosted by a technology
partner. This flexibility is only realistic using unicast technology, as no special hard- and
Linear Live TV consumption is decreasing in Liechtenstein like everywhere else. Time Shifted
Live TV, Catch-up TV, and VOD are gaining importance. Therefore, even today Multicast
technology does not have any relevant advantages and we expect this trend towards
unicast to continue. Dependency on Multicast enabled devices would limit flexibility in the
product offering. With our Unicast solution, a wide range of devices can be supported and
9 | 10
With the expected usage and load on the backbone, it is perfectly feasible to support a
Unicast deployment in the Telecom Liechtenstein Network. Multicast would bring added
complexity to network infrastructure and operation, which – in a network of Telecom
Liechtenstein's size – can’t be justified.”
Romain Lonfat (Head of Interactive Television, netplus.ch S.A., Switzerland):
„The peering we have in place today e.g. with our technology partner Zattoo for our own
unicast Mobile-TV service, has several key advantages for us. Firstly, we get access to high
quality video streams without depending on external content delivery network providers
(CDNs). Secondly, we don’t have incremental cost for the consumption of TV content as
we leverage our own network. And last but not least, we ensure that the subscribers of
our Mobile-TV service who are connected to the internet directly through netplus.ch will
get high quality streams at any time of the day. This is the case even at peak times as the
peering will make sure that the data will only transit on our managed network on which
we have different priorities (QoS) depending on the services and the contents.
The overall quality thus does not depend on delivery anymore but we can set a level of
quality depending on the devices covered. Today, we provide e.g. HD channels with a
maximum profile at 3Mbit/s (ideal for PC and tablets) but we can easily increase that
quality and bitrate if we need to address TV sets.”
Matthias Grewe (Founder and CEO abox42 GmbH, Karlsruhe):
„As a STB provider to network operators and ISPs in Europe and beyond, we see a strong
trend for IPTV solutions leveraging the power and the advantages of Unicast distribution.
Especially for 2nd and 3rd tier operators, Unicast IPTV is highly attractive as it suddenly makes
major hurdles and sources of complexity disappear. Most importantly, with a unicast IPTV
there is no need to set up and operate extremely complex and costly Multicast networks.
Quality of service is not an issue in this setup. On the contrary, with the use of the own
managed network for Unicast distribution, a Unicast TV service clearly matches the quality
of service of Cable and Satellite TV. Plus, the network operator can even outsource the
entire TV service to an experienced TV service provider who then hosts the TV product and
service. This for the first time allows for a rapid and cost effective rollout of an advanced
multiscreen TV service for 2nd and 3rd tier network operators. At abox42, we are already
involved in successful projects of this type.”
10 | 10
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF