Nothing was more welcome after day 1 of EBU BroadThinking 2013 than a good night of sleep, in order to reset the tech hype counters and make some mental room for two new sessions on broadcasters’ CDNs and the latest advances of hybrid platforms. What would be the best broadcasters’ CDN architectures today, what would be their smartest (green) evolutions, would CDN-Federation standards finally bring interop reality over hopes, what would be the most advanced deployments and future of HbbTV, how it compares with YouView in the UK : day 2 agenda was looking quite attractive – and indeed the presentations were packed with valuable informations and experience feedbacks.
So here is the recap of the most interesting DAY 2 presentations and a short report on some insightful demos that were playing on the EBU floor right to the conference.
Session “WHICH CDN MODEL FITS BROADCASTERS’ DEMANDS?”
As a massive hammer blow on first part of the day, Martin Boronski – CTO of M6Web – came with a remarkable exposé “Building your own CDN : a broadcaster’s perspective” on why & how one of the most important French TV channels build its own CDN for OTT distribution. As a context information, everyone must know that M6 is a RTL Group subsidiary, developing video distribution activities for live, catchup & vod on a wide range of devices (PC/Set-top-boxes/Smartphones/Tablets/Connected TVs/Xbox) and distribution networks (IPTV/OTT), with 1 billion viewed videos in 4 years and 50 million per month these days with very popular services like the multi-devices M6 Replay one. This is a really massive audience and it can explain why building its own CDN for on-demand contents was a major decision to take. Martin exposed a bunch of drawbacks of commercial CDNs (insufficient connectivity with local IPSs, lack of local support, lack of support for HTTP ABR delivery) as one of the main drivers for change at the time the decision was made, completed by a big opportunity of cost reduction due to IP transit cost decrease and use of open-source solutions, as well as the need to stay in line with market trends pushing to DRMization of the contents and ABR streaming instead of proprietary protocols. So M6 builded its own CDN in a single POP, using commodity-hardware caching servers, Isilon storage, Juniper routers and open source software like Nginx (for long-form contents) and Varnish (for short-form contents) : nothing really rocket-science here, but a reasonable assembly of mature and affordable technologies, handled by a highly-skilled sysadmins team.
Apart from the technological independence on the server front, the project had the ambition to bring down the number of network intermediates up until the end-user and to guarantee two network paths for each ISP – a redundancy feature which commercial CDNs could not necessarily provide. The simplification of the delivery chain resulted into the disappearance of Global CDNs (at least for all catchup/VOD distribution) and Tier1 intermediates, as well as a 30% cost savings compared to a global CDN (IP transit representing 44% of the overall budget). That’s the part where whispers start to say “wow!” in the conference audience, all the more that Martin made clear that they had no down-time for the last two years on the CDN, with bursts at 40Gbps… As an echo of some of my day 1 topics, Martin then talked about their ongoing plans to deploy Unified Streaming Platform solution on their origin servers, in order to simplify the production chain which is complex today due to a high number of target devices and induced ABR/DRM packagings for each program – and their plans to setup performing multi-CDN switching for both QOS and failover purposes. Martin specified that the idea of placing M6’s caches inside ISPs’ datacenters for lightening the peering costs was floating around as a natural evolution, but that the reluctance of ISPs is strong as long as the actor on the other side has not the weight of a Netflix. That was a speech highly appreciated by the audience, with a lot of questions and interest raised, and a perfect introduction for the next presentation.
Next on stage with her “CDN topologies and the live multicast use case” presentation was Nouvedita Nouvel from BroadPeak, a very innovative French company (and spinoff from Technicolor) providing various CDN solutions for operators and content providers. Starting with an educational approach on IPTV and OTT delivery business models and a comparison matrix of global CDN vs operator CDN, Nivedita came up with an intermediary conclusion stating that “an ideal set-up would combine operator CDNs for regions of focus and global CDN for extending the reach”. Focusing next on the live web TV use case, and mapping the inherent scalability/QoS/dimensioning dangers that it presents when facing a massive audience served in unicast, Nivedita outlined two major advantages that operators can benefit – and that global CDNs can’t prevail of : a managed network with multicast capabilities and the control of CPEs deployed in the home network. Those advantages, combined with the use of the nanoCDN technology which was shown first during last IBC, allow them to turn the CPE into a CDN edge right inside the last meter (not even the last mile) of the home network !
The nanoCDN technology is made of mainly two different components: the Transcaster server which is running in the head-end, taking a standard unicast live ABR stream (like an HLS stream in push mode, secured by DRM if needed) as input and converting it into a multicast channel, so that the live streams are conveyed in multicast down to the home gateways inside the ISP managed network. On the home gateway, nanoCDN is present as an embedded software component which converts back the multicast stream into unicast – therefore allowing all compatible devices in the home network to read the stream without any end-user application modification (the only requirements is to apply an additional routing management layer over stream URLs to handle also cases were nanoCDN is not used – with BroadPeak umbrellaCDN for example). In the head-end, the operator can use BroadPeak’s Mediator server to monitor live stream popularity over the network and trigger their automatic ingestion through the Transcaster servers when they reach a popularity level where provisioning a multicast group makes more sense than delivering all the streams in unicast. The beauty of the system is that through the intelligence of the nanoCDN home gateway component, the end-user device then uses the multicast-to-unicast translated stream most of the time, and can fallback directly on the unicast delivery mode when the stream is paused or trick-moded. At the end of the day, the operator sees that his OPEX costs don’t rise with the stream popularity, as all the heavy load is absorbed by the native capabilities of multicast routing. That’s a heavy budget difference and all actors in the workflow chain (ISP, content providers) shall be interested as it is both a source of reduction of required peering capability, of backbone use and an major workflow optimization by reuse of existing resources already encoding live ABR stream in HLS or other formats. That’s what I’m calling it a major technology disruption, and I wish all the best to BroadPeak with this groundbreaking offer ! NanoCDN pilot deployments started with French ISPs, so it seems it’s going a very successful way…
As a follow-up of the EBU CDN Workshop in last October and especially the presentations (freely downloadable with a simple EBU web account) of Marcin Pilarski [Orange Labs] and Oskar van Deventer [TNO] about the standards and ongoing experiments in CDN Federation technologies, the BroadThinking presentation from François Le Faucheur [Cisco] about “Initial Experiences with CDN Federation” was a good mean of getting the latest updates on this touchy topic. Touchy because federating CDNs implies changing the economics of this business ecosystem, and also touchy from a technical point of view because the standard currently being defined is quite complex and potentially difficult to implement. After an introduction about the possible CDN strategies available for Media companies, François synthetized the business case for CDN Federation (CDN Federation being defined as “A collection of CDNs operated by autonomous members and interconnected via open interfaces so they can act as a multi-footprint logical CDN”) with prime-CDN and sub-CDNs intelligently providing the contents using the best downstream CDN to reach user – “best” potentially reflecting performance, costs and peering agreements. Jumping to the ongoing experience from CDN Federation Pilot, Francois summarized all the efforts already deployed by a consortium of network service providers (8 participants like Cisco, Bouygues Telecom, SFR, Swisscom… in phase 2, 14 participants in phase 3) to test the CDN Federation concept at large scale and contribute to the standardization through the IETF CDNI initiative : phase 3 is finishing this May with Advanced capabilities (flexible CDN selection, delivery and content adaptation for mobile access…) and expanded ecosystem/geography track record. Basic (Phase 1) and enhanced (phase 2) steps of the trial have already allowed the validation of an important number of features for on-demand, live and web delivery : ABR VOD and live streaming in all ABR formats, time-shifting services, HTTP/DNS based redirection services, access-control policies through signed URLs, multi-CDN reporting etc… Advanced features still need external API integration as they are not integrated into IETF CDNI reference standard. Phase 2 saw also the validation of an important number of business hypothesis, thus providing good hope that delivery actors will embrace it naturally (despite the needed technical efforts). On the standardization front, François Le Faucheur and Richard Woundy [Comcast] are co-chairing since June 2011 the CDNI Working group inside IETF, and they have already 6 documents adopted as basis of the final CDNI Interface specifications which are to come out in second half of 2013. The IETF has formalized a number of Models and Interfaces allowing to specifying key interactions (Control, Logging, Request routing, Distribution…) between upstream and downstream CDNs.
Couple of personal comments on this topic now: while the initiative seems to create a growing momentum in the industry, CDN Federation is still in its infancy due to incomplete CDNI standards and a perceived high level of difficulty for many medium or small actors potentially interested by such evolutions. To make an IT analogy I would say CDNI compares with CORBA while proprietary services like Cedexis, BroadPeak umbrellaCDN or MediaMelon look more like agile WS-* sort of standards with pragmatic and quick time-to-market approaches. While it’s not an uncommon situation to see such precursors grab preemptive market shares, the feedback of today’s multi-CDN policies that are in production is not written in advance (for example it seems that client-side multi-sourcing à la Conviva is gaining wider interest from the broadcasters, over server-centric routing decisions policies) and it could also impact the popularity and success of CDN Federation once it will have been fully standardized and productized. Anyway, François’s presentation was a great update on this expert topic.
Session “NEXT GENERATION OF HYBRID PLATFORMS”
Jeff Hunter’s “YouView – hybrid in the UK” presentation was a good occasion for many participants to get a glimpse of the original initiative which is trying to federate content providers and ISPs in providing a a mix of SD/HD DTT, live DVR, seven day catchup and other on-demand contents through a common platform. YouView is somehow close to Google TV’s approach with the transversal search feature over all channels’ contents, although the content provider is retaining control over presentation of content and can use his own player integrated in the YouView UI. The centralization of all contents is a major difference over HbbTV which only provides the common technical socle for building services. Youview’s one is based on Adobe AIR, meaning more fluid interface and a user experience HTML interfaces can hardly provide.
As regards the traction YouView has gained so far, it seems that 100.000 compatible devices are out in the wild, with a general positive feedback. As regards the major next steps for the project, Jeff spotted the delivery of channels over IP (Multicast, unicast DASH) on top of the DTT existing delivery, more companion screen features for managing recordings, and an exploration of support for HTML based services. My 2 cents : by many ways YouView has taken more advanced/risky technical choices and it may be the secret of producing a sexier service than the ones we see in continental Europe over HbbTV, but at the end of the day my feeling is that what people like most are services that match their urban stressful lives and bring up new value-add over “usual” services like live/catchup/vod (even blended and aggregated), see for example the Salto start-over service deployed over HbbTV by TDF for France Télévisions – a simple but really disrupting service offer…
Peter MacAvok [EBU] then delivered his “New developments with the HbbTV 2.0 spec” presentation, which quickly recapped the history of the project and outlined the scope of HbbTV 1.5 with MPEG-DASH, Common Encryption and multi-DRM support. There is currently no official Test Suite for HbbTV 1.5 features (only for 1.0 features with old test suite v.0.9 of August 2012) but some providers like Digital TV Labs have already included HbbTV 1.5 features into their testing product Ligada iSuite for HbbTV (it’s needed for TNT 2.0 service development in France, which is based on HbbTV 1.5). As regards the HbbTV 2.0 specification, it seems that the consortium has switched over a timeboxed approach to release the specs, with a deadline at end of 2013 which means that everything not fully defined by then will be discarded. This is somehow concerning because the general directions Peter presented (HTML5 support, second screen synchronization, better access services support, improved existing 1.5 features) are all required for the sake of the standard evolution, given the long time that each generation of TV needs to be launched on the market with the appropriate norm level support. Hopefully for 2.0 no hardware evolution will be required (no HEVC mentionned – coincidence or volontary oblivion ?) and the release could make it through quicker than the 1.5 generation which is one year late on the market.
We don’t know exactly which new features are the most at risk but, from my point of view, if for any reason HbbTV 2.0 doesn’t include all the second-screen and sync features submitted by HBB-NEXT, it will be a major fail. Let’s cross our fingers…
Finally, in an echo to Peter’s presentation, Thierry Fautier from Harmonic and Xavier Heusghem from NAGRA shared a field experience on large-scale deployment of HbbTV 1.5 services as they are currently doing in Spain for Albertis Telecom, through their “MPEG DASH coming to life with HBBTV 1.5 and Cloud Services” joined presentation. Taking advantage of MPEG-DASH, ISO BM File Format,Common Encryption and Multi-DRM, the joined Harmonic-Nagra solution standardizes the management of multiple DRMs for which the companies made a specification proposal. On top of these DASH/CENC/DRM aspects, the Nagra Multiscreen Cloud Service offers a global management opportunity of such HbbTV 1.5 services in the cloud, with support also for non-HbbTV compliant devices and support for various business models (AVOD/SVOD/TVOD, no EST). Clearly a new competitor for well-established OTT services providers having integrated HbbTV in their offer, like SyncTV or Tvinci.
And what about the demos ? What was hot & therefore is good for a NAB booth standup ?
The most jaw-dropping demo was certainly the one from InterDigital Communications, where a camera mounted on the screen analyzed the environment and adaptivey displayed the best suited video version of the movie, based on analysis of user behaviour and viewing conditions, including user distance to the display, display density, contrast, and ambient illuminance – all that without perceptible impact on video quality. A very impressive demo now needing commercial outputs…
[02/21/2014 update : they now have packaged SDKs for this, read whitepaper]
DASH players : here come the CENC. Both BuyDRM and Castlabs (unsure about Qualcomm on it) were showing off their Android players supporting Common Encryption and PlayReady DRM (not several DRMs in one stream, for the moment – no real customer demand for it yet). Same kind of player was running on a new generation Panasonic Connect TV in HbbTV 1.5 with Dolby Digital+ audio track.
I missed the Fraunhofer FOKUS demo but the description in the EBU lineup looks good and the website is quite interesting on DASH and DRM interoperability : “Fraunhofer FOKUS solutions Famium and DASH Transcoder form a platform for the consumption and delivery of protected and unprotected adaptive bit-rate content. In this demonstration these components are used to present use cases that go beyond just playback. The DASH client Famium is based on the DASH-IF reference client and has been enhancedwith multi-screen and synchronization capabilities.”
And finally… HEVC : Harmonic is there ! Thierry Fautier had a shiny smile on his face when he showed me his little jewel : a 720p50 2Mbps HEVC encoded tennis match playing beautifully on a Nexus 10 tablet via a Squid Design player (4 hours of autonomy). He had an equivalent in 720p25 playing on an iPad 3 equipped with a Docomo player.
And now some conclusion words after this rich 2-days event !
The event itself is truly a high-quality one, with mostly highly valuable presentations (which you can download from the EBU event website if you are a registered EBU member). The companion tech demos are nice and the spot is an overall good one for networking with fellow engineers and broadcasters. The Q&A sessions after the presentations launched true debates and approaches confrontations in a fragmented European broadcast paysage, and I think next year event would even be better if a timeslot was reserved for pure debate. Check regularly the EBU website for registration : next year it will be on March 26/27 and we may cross each other there!
On the substance, I remain after the event a bit worried about the HbbTV consortium’s capacity to keep a good pace in bringing up the new specs and to make the good sacrifice choices if needed. Standardization is indeed a tough goal to reach given the contradictory industry interests at stake – but this is precisely why we need strongly financed institutions like the EBU, the ETSI or the IRT to make this happen in the best possible schedule and feature set, even if the committees are hard to run !