Adaptive Streaming in an OTT World

Sunday, August 12, 2012

Article Image
One aspect of preparing an ABR package is to prepare the half-dozen or so streams corresponding to each of the package's layers. Using conventional hardware, transcoding for an ABR package with a half-dozen layers will typically take about six times longer than transcoding a single fixed-bitrate file of the same content.
It's often said that change is disruptive, and that's certainly true for the most significant media distribution development in recent memory. Jim Duval, Director of New Products and Strategy at Telestream examines the advent of new viewing devices for content delivery. 

With the extension of video viewing from standard TVs to smart phones, tablets, connected TVs, DVRs, and game consoles, over-the-top (OTT) media has become a major factor in content distribution, making today's media landscape radically different from that of just a few years ago. What used to be all but impossible—the delivery of consistently acceptable image and sound quality despite often uncertain and fluctuating bandwidth—is now taken for granted by viewers on these devices. Adaptive bit-rate (ABR) streaming is the key technology to making it all work.

The emergence of ABR—Apple HTTP Adaptive Streaming, Microsoft Smooth Streaming, and Adobe Dynamic Streaming, and MPEG DASH—coincides with the significant investment that content owners and distributors have made in recent years to extend media distribution beyond the confines of traditional broadcast and cable television. It would be a mistake to think of the content preparation side of that investment solely in terms of discrete devices for transcoding video into the correct format for various outlets. Instead, efficient high-volume content preparation demands a comprehensive workflow approach that addresses archiving, digital asset management, transcoding, and delivery in an intelligent workflow. ABR has direct and significant impacts on this workflow.

With ABR growing in importance, it's imperative for content owners and others who prepare media for distribution to fully understand how ABR files differ from conventional streaming files, the implications of those differences for conventional media preparation processes, and how those processes may be successfully adapted to create highly-efficient, cost-effective workflows addressing both ABR and conventional deliverables.

Understanding adaptive streaming
In conventional streaming, once a connection is established between an end user and a media file, that file streams at a fixed bitrate and displays at its inherent resolution (e.g. 640 ? 480). ABR streaming instead tailors streams to the resolution of the playback device and the available bandwidth of the connection. Connect via a tablet over a strong Wi-Fi signal, and you'll get a larger picture and higher bitrate. Connect via a smart phone in your car and you'll get a smaller (lower resolution) picture at a lower bitrate. If your connection improves as you drive, the system will adapt on the fly to the changed conditions, increasing the bitrate at which the content is streamed.

To enable this flexibility, an ABR file isn't really an individual file at all, but rather a package of files. An ABR package includes a manifest file, which holds the stream metadata, and a set of multiple "layers," each made up of the media data for a different target bitrate. To enable switching between layers as conditions change during streaming, the content for each layer is fragmented into files of only a few seconds in duration.

One aspect of preparing an ABR package is to prepare the half-dozen or so streams corresponding to each of the package's layers. Using conventional hardware, transcoding for an ABR package with a half-dozen layers will typically take about six times longer than transcoding a single fixed-bitrate file of the same content. Few facilities are currently equipped to increase production by a factor of six without bringing production to a standstill.

One solution to this capacity gap would be to maintain a conventional approach to transcoding while investing heavily in new machines to run the transcode processes and in expanded data storage capacity, thereby adding a host of associated ongoing costs in areas such as cooling, energy, and real estate. While ABR provides content owners with access to a new, fast-growing market, it does not offer the multi-fold increase in revenues that would be needed to justify this approach. 

An alternative solution is to take a new, more efficient approach to the design of the transcoding systems themselves. At Telestream we've done that by developing an exclusive new technology, called Lightspeed™, that implements parallel video processing and transcoding algorithms to accelerate video processing and H.264 encoding on parallel GPUs and also on multicore CPUs. The result is the highest possible image quality at the fastest possible speed. Solutions incorporating Lightspeed provide the boost in processing power required to address the ABR throughput dilemma, and they do so without a corresponding increase in hardware, operation, and maintenance costs.

The workflow context
While highly-efficient transcoding technology is a big part of the ABR solution, it's far from the only factor to consider in planning an effective media preparation system for high-volume use. A content preparation system must also address the assembly into an output clip of material drawn from multiple source clips (e.g. provider logo, main content, provider promo), ideally working from a playlist that specifies the exact content (source files, offsets, durations, etc.) for each finished clip. Transcoded files must be QCed. The components, both transcoded and externally provided, that make up the deliverable must be assembled into format-compliant packages for delivery to target destinations, in some cases with DRM or other encryption applied. Job status must be tracked through the workflow, and the quality and compliance of the final result must be verified before the materials are handed off, after which receipt must be confirmed.

With such a complex overall process it's long been evident that high-volume media preparation can only be achieved cost-effectively with a comprehensive, unified solution that addresses the entire journey of content from source to destination. Any task that can only be done manually, or that must be done more than once, is an obstacle to maximum productivity. So the imperative for vendors serving high-volume content providers has been to design these inefficiencies out of the picture, maximizing quality, throughput, and control while minimizing labor. Telestream's Vantage systems, which bring transcoding, media capture, metadata processing, and analysis together into a single managed process, are a prime example of this approach.

So how does the advent of ABR change this picture? The fact that the output of an ABR preparation process is a package rather than a single file does have a substantial impact on the requirements of individual workflow steps. An ABR-ready solution must be able not only to generate all of the component files that make up each package, but also to view that set of files as a single job, to track that job through the workflow, and to hand that job over to downstream recipients as a format-compliant package rather than an undifferentiated mass of separate files.

Zoom out to look at the big picture, however, and it's clear that the capabilities required of ABR-capable systems overlap almost entirely with those of existing non-ABR solutions. And thus it makes far more sense to adapt existing systems to handle ABR than to develop, test, and deploy new workflows specifically for ABR:

    • In most situations, the same content will be processed into both ABR and non-ABR outputs. Industry experience has shown that when transcoding for multiple outputs it is faster to access a given source file just once and to transcode in parallel than to access the source multiple times to perform separate serial transcodes.
    • Most non-transcoding steps in the workflow will also apply to both ABR and non-ABR outputs. Again, it's more efficient to perform these steps just once than to perform them for non-ABR deliverables and again for ABR. 
    • Most of the technology required to perform the needed tasks, to automate the workflow, and to track jobs through the process is the same for both ABR and non-ABR content. Purchasing, operating, and maintaining separate systems to handle these tasks for ABR is inherently less cost-effective than adapting existing systems and scaling them to meet combined ABR/non-ABR demand.

The bottom line is that most of the capabilities needed to effectively handle ABR—source-file playlists, job tracking, status reporting, and the handoff of output materials to external systems—are already available in field-proven high-throughput systems for non-ABR content, which means that there is no compelling reason to develop an entirely new framework for ABR processing. Instead, at Telestream we've taken our advanced field-proven systems, analyzed the impact of ABR on every step, and created unified solutions that maintain all the advantages of full-featured video transcoding, workflow automation, and system management while also being uniquely adept at simultaneously handling ABR and non-ABR outputs.

The result of Telestream's ABR efforts is a unified system that streamlines the entire process of source file decoding, video processing, parallelized H.264 encoding, packaging, encryption, quality control, and delivery. Combining our industry-leading expertise in workflow automation and management with the power of Lightspeed™ technology, Telestream Vantage is the ultimate high-throughput solution to the challenge of content delivery for multiple screens.

Article Search


   cmip equinix XStream cmip  cmip 
BPL Broadcast Limited, 3rd Floor, Armstrong House, 38 Market Square, Uxbridge, Middlesex, UB8 1LH, United Kingdom | +44 (0) 1895 454 411 |  e:  | Copyright © 2014