NAB 2012 OTT Advancements & Doubts

NAB 2012 closed its doors two weeks ago, so it’s a good time to draw an appraisal on various themes concerning OTT issues, and especially premium OTT issues, that have been handled through industry technology offer during the tradeshow. Here we’ll go from production to distribution and examine the salient NAB facts and products, at least the ones which have a potential influence on OTT workflows evolution in the coming months (or years ?).

Encoding workflow : standard integration takeover is not for this year…

As FIMS (Framework for Interoperable Media Service, a workflow standardization and interoperability effort from EBU and AMWA) began to fly over the radar at NAB 2011, many NAB attendees (including me) were eager to check one year after if the initiative was gaining traction among the media companies and the broadcast industry in general. Indeed, this initiative is a major step towards the end of vendor lock-ins for the media actors.
As a reminder, FIMS aims at providing standard interfaces for usual processing services used throughout the media workflow, and specifies how they shall be orchestrated as defined in the Services Oriented Architecture (SOA) principles of loose coupling : in version 1.0, the covered domains include capture (ingest, source device control, metadata generation, file import), transform (rewrap, encode/decode, crop/resize/burn-in) and transfer. The newly launched FIMS phase 2 is focused on MAM and QC issues.

First FIMS draft was issued in November 2010, so the project just recently came to a sufficient maturity level to authorize field implementations. This year the AMWA/EBU was hosting presentations about the workflow implementation done at Bloomberg, hoping that major new implementations would follow this early-adopter example. Given this relatively slow traction gained among broadcaster, many industry actors have adopted a wait and see attitude. Even if they implemented the FIMS service interface in their product, it’s still considered as a technology move for the future. As summarizes Kirk Marple, Radiantgrid Chief Software Architect : “It doesn’t make you win markets right now, but it can make you lose some if you don’t support it”. Some actors, currently less involved in FIMS, don’t hesitate to share their skepticism and stick to their roadmap and proprietary APIs, like John Kudrle (Telestream Principal Engineer) : “FIMS significantly restricts features compared to our encoder SDK native capabilities, so it’s not our priority to implement it until customers are demanding it explicitly”. I discussed this issue with Jean-Pierre Evain, Senior Project Manager at EBU, and he told me that “we (AMWA/EBU) don’t encounter such reluctance from the most implicated encoding industry actors but still can’t pretend that FIMS current implementation is perfect, so all remarks on the major missing features and uncovered use cases in the spec are welcome”. Standardization has still a long run to go but it’s going in the right direction…

Cube-Tec advanced QC workflow

Cube-Tec advanced QC workflow

As a concrete proof of the traction gained by FIMS in the industry, Cube-Tec was demonstrating its new BPMN 2.0 CubeWorkflow engine supporting multiple QC engines (like Digimetrix Aurora, Tektronix Cerify…) with a workflow demo focused on loudness control. TMD was also announcing its upcoming Unified Media Services FIMS-based toolkit which is to be released during next IBC, while VSN was presenting its FIMS-compliant Spider framework. Seeing these initiatives flowering after IBM and Sony engines is a good thing in the way that this competition may bring more accessible workflow tools to lower budget TV channels. The same level of choice for compatible transcoding engines would be nice to reach after the pioneer implementations from RadiantGrid or Sony (Vegas) : in this area we’re waiting for product advances from Elemental, Cisco, Telestream, Digital Rapids and Harmonic.

But this year they were focused on different strategies and features sets – none of them using BPMN as a workflow language or FIMS as a service interface norm – but nevertheless very interesting :
– Elemental focusing on new platform features like DASH, Common Encryption, loudness management, advanced subtitling and 4K H.264 encoding
– Cisco polishing Inlet Armada (renamed as Transcode Manager) integration into its Videoscape solution offer and developing its ABR packaging output capabilities over mutualized transcodings

Digital Rapids Kayak deployment

Digital Rapids Kayak deployment

– Telestream releasing an upgraded version of its Vantage workflow engine and integrating x264 in its encoding nodes
– Digital Rapids releasing its new Kayak workflow engine and putting it in action into the Cloud (managed in coordination with on-premises deployment if needed, with Transcode Manager 2.0)
– Harmonic focusing on DASH, loudness control and multi-screen packaging

While exploring the halls, I also found some interesting encoding products on Thomson booth where they showed their new Vibe VS7000 blade transcoding unit (supporting, of course, DASH) which workflow engine is close to Vantage or Kayak and roadmap intends to handle clustered file to file transcoding on top of live transcoding features, and on Allegro’s booth where the AL2400 live transcoder interestingly featured an embedded origin server, DASH support and TTML subtitling support.

DASH & DRMs : situation gets clearer, without Apple…

Live DASH @ Akamai

Live DASH @ Akamai

One major technology announcement during NAB was made by Microsoft : the upcoming support of DASH in all technologies of its Media Platform which is now deployed in the Azure Cloud. While the technical step is not huge for Microsoft as Smooth Streaming and DASH fragmented MP4 profile are close cousins, it provides more meaning in the political aspect of things. By joining Adobe (who has slowly begun to erase HDS from the map in benefit of the DASH standard) and many other big industry actors supporting DASH like Harmonic, Ericsson or Akamai, Microsoft clearly tries to marginalize HLS and Apple. Despite tremendous market demand for a HLS standardized DRM, Microsoft dropped its earlier idea of releasing an official PlayReady specification for HLS that would have unified implementations made by Discretix, Verimatrix and other DRM solution providers like Authentec (only the DASH specification will be released later this year). Microsoft is only providing bare Smooth Streaming/PlayReady iOS SDK and an Android Smooth Streaming/PlayReady porting kit (shown ported on Nexstreaming NexPlayer) , which sounds like a kind of minimalistic effort aiming at discouraging further HLS implementations in favor of DASH ones.

Even if DASH made the headlines on many NAB booths and is a sure shot in terms of industry common initiative, the HLS-compliant installed basis is so huge that the industry will still have to cope with HLS for a long time, that’s quite sure because the decoding devices need to add the DASH stack first and mainly because the DRM requirements have not yet been finalized nor implemented throughout the various clients : the junction point between DASH and the Common Encryption Format is just finalized but a long story remains to be written as regards DRM interoperability on top of the shared encryption scheme. Major concerns during the DASH Tuesday panel were focused on how the various browsers would handle the necessary DRM hooks allowing several different DRMs to be supported on each client, the devices being less a subject of questions as you can deploy custom apps on it. Google Chrome and Safari were especially targeted by these questions, as both Google and Apple are not invested in DASH, but a beginning of answer came just this week on Google’s side with public confirmation that the move to DASH was on Widevine’s roadmap. This does not provide a sure clue that other DRMs like PlayReady and Marlin could be supported inside Chrome but it’s going in the good direction. On Apple side, all actors were quite sure that Apple wouldn’t jump again in the DASH wagon, several months after its deep involvement in DASH specifications, and would stick to HLS which is under control for them. Combine this idea with the rumors about a new upcoming Apple DRM (not FairPlay) and the ongoing involvment of Apple in the Common Encryption Format and you get all the presumptions of a solo HLS-centric initiative of Apple. This won’t be good for interop but it’s the most likely scenario considering Apple’s devices installed basis as of today. Another concern was raised during the DASH panel : unlike FIMS which has been organized as a patent-free project, DASH has not the same status, so we’ll have to wait the DASH-PG to issue a general agreement on this topic and have all concerned actors sign it afterwards.

OSMF-based DASH player

OSMF-based DASH player

As regards the DASH demos visible on the tradeshow, the playback was generally handled on Android/iOS tablets via Authentec, NXP , Netview and VisualOn players or PC via an OSMF 1.6.1 Flash Player (visible on Akamai’s booth, with a player supporting targeted advertising). DASH encoding/packaging support was provided by encoders from Harmonic, Elemental, Envivio, RGB Networks or Ericsson. The Fraunhofer demo of DASH use for audio adaptive streaming was also intriguing. No demo of DASH running inside the newly specified HbbTV 1.5 environment was yet visible but that will surely be an IBC hit this year as HbbTV gains massive traction in Europe now. Despite (or because of) the limited client support of DASH as of now, all pilot projects are very interesting for the involved industry actors, so you won’t have much trouble to get them invest resources for free in your potential project, we’re in a good time for early adopters…

Authentec/Elemental DASH+PlayReady demo

Authentec/Elemental DASH+PlayReady demo

While DASH is fully open to codecs outside H.264(which is good for future HEVC/MVC projects), the DASH-264 profile will federate industry as a first milestone in its spreading, it’s quite clear. As regards the dominating DRM in the DASH scope, the NAB was just suggesting PlayReady for this role, as it’s the only DRM used by everyone in OTT deployments and Microsoft’s strategy is pushing hard towards DASH now. Widevine traction was very discrete : only Brightcove and Vimond platforms announced support for this DRM during the NAB, and Marlin visibility was near to none (only Arxan was showing a Marlin hardening offer and Verimatrix a Marlin integration in its VCAS suite). The only demo of early DASH+PlayReady solution was shown by Elemental with an Authentec player, thus confirming Elemental edge market positioning on premium OTT use cases. UltraViolet was not omnipresent (only seen on Akamai, Rovi and DTS booths – with Elemental hardware, and on Digital Rapids booth demonstrating CFF production with Transcode Manager 2.0) so it’s difficult to evaluate real industry faith in this service offer for the moment, all observers agreeing to say that Common File Format and Common Encryption are the real advances that will stay even if UltraViolet fails as a commercial service solution.

Delivery infrastructure : serve more with less hardware… and better

Smartcache Server principles

Smartcache Server principles

NPEngine

NPEngine

Smartcache Server

Smartcache Server

From all the products I’ve seen throughout NAB this year, the major breakthrough standed at Toshiba booth : the SmartCache Server. This very innovative server supporting HTTP adaptive bitrate and RTP streaming can replace 3 to 4 usual streaming servers (for 20K$ or so per SmartCache unit) as all the network content delivery is done by a dedicated hardware controller (called the NPEngine) on which are plugged 16 SSD drives (providing a 4 TBytes storage space, and 8 next year with new SSD generation). Toshiba’s benchmarked stated that the unit can deliver a 512Kbits/s stream to 64.000 simultaneous users, capping at 40Gigabit per sec. Yes you read it well : 64.000 concurrent sessions, 40Gb/s ! The beauty of this architecture is that the NPEngine is transparently integrated in the Linux kernel (so you don’t have to modify existing applications to make them work with the Toshiba server) and that the CPU power is available for anything else that delivering video content : for example you could use it to run the Codeshop Unified Streaming Platform software to package and DRMize the pre-encrypted video contents on the fly. While the server architecture is a pure delight, you still will have to run on it software with a good cache management mechanisms, as the 4Tbytes capacity is not so huge given the usual video content weight.

OTT Media Grinder

OTT Media Grinder

On the QOS/QOE front, the main new product showed during the NAB was the Pixelmetrix OTT Media Grinder, a system designed to validate OTT delivery infrastructures by emulating thousands of OTT clients with a fine grain control on their behavior (like changing requested bitrates). The collected data falls for calculations into the VideoMargin Metrics which provides 7 parameters to measure OTT delivery quality on network, HTTP and video layer performance levels. Unlike pure QOE systems manipulating end-user applications like Witbe‘s one, OTT Media Grinder doesn’t decode the video streams so it can’t provide indicators or alerts based on the video content quality or defects, but still it is capable of requesting the DRM license necessary to view a stream, so it means that your whole video servers and DRM servers platform availability will be monitored in real-time : pretty useful ! For the moment the system supports HLS+AES, Smooth Streaming+PlayReady support is coming shortly and DASH is on the roadmap for end of year.

Video in the Cloud : this Revolution will be Televised

Witnessing a major industry shift, NAB was hosting this year a new Cloud Computing Pavilion inside North Hall – few booths, few visitors (at least on Monday morning when the exhibition opened), but the Cloud was here and almost everywhere else, in North Hall for the management aspects and in South Hall for everything else, confirming the market traction for such type of solutions. Now, apart from live encoding, you can manage the whole production chain in the Cloud : pre-production/production (Aframe, ProductionMinds, Arvato), editorial collaboration (Conciety), MAM (Reelway Reel Cloud, Front Porch Digital LYNX), NLEditing (Quantel Qtube), backup&archive (Fujifilm Permivault), search (Green Button), BI (Decentrix), fast-transfer (Aspera) and of course transcode and deliver video – but this has been the DNA of video in the Cloud for some years now…

The most visible announcement this year was made by Microsoft with the availability of its Azure Media Services solution, which is actually a transfer of all existing Microsoft Media Platform technologies in the cloud, meaning Smooth Streaming (and soon DASH) video delivery services, IIS Transform Manager, Expression Encoder and of course PlayReady DRM delivery, all of it offered through a consistent set of REST APIs. The good thing is that Microsoft is attracting its usual Media Partners ecosystem on Azure Media Services, and even the unusual one like Wowza, so there’s a bunch of OTT-interesting processing services that you will be able to run on Azure, like : transcoding (Vanguard vCompress Pro, Ateme Titan, Digital Rapids Kayak, Dolby), delivery (Wowza), security (Civolution Watermaking, BuyDRM) and transfer (Aspera On-Demand). Sounds like there is not much holes in the platform in terms of features, apart from live encoding (not transcoding) which is something unlikely (?) in the cloud. This makes Azure Media Services a direct and powerful competitor to Amazon Web Services platform which suddenly appear not so specialized in video processing as we could think six monthes ago, but maybe it’s a false idea, as the historical video processing/transfer services were deployed here first : Zencoder.com, Encoding.com, Aspera On-Demand (on booth of which Amazon was present). New services are still popping up on Amazon EC2 platform : Sorenson Cloud, Elemental Cloud (soon publicly available, it seems, on EC2 GPU & CPU flavors – and as a hybrid on-premises/cloud offering for burst-out handling), Haivision HyperStream transcoding services and the Red Hat Virtual Storage appliance which gracefully synchronizes several EC2 storage machines.

As new cloud providers appear, like CloudSigma who wants to grab the media industry cloud hosting business or Zunicore with its GPU servers, and that new services like the cloud-broker CloudMediaHub tend to simplify cloud use and billing, there is a great chance that the Cloud Pavilion will occupy a much wider space in NAB 2013 & next ones, as video is less and less a video topic and a more and more computing topic. SOA is authorizing a whole new combination of services on-premises and in the cloud with a guaranteed management of business processes – that’s where FIMS is crucial again, in the cloud perspective – and it therefore provides renewed insights on perfect SLAs and uninterrupted production workflows like the media ecosystem badly needs…

Let’s check if these OTT industry trends do confirm during next IBC !


For a selection of NAB 2012 news and announcements, you can check my curation page on this topic.