The Transformation of Transcoding
Transcoding or Protocol Conversion?
The biggest issues facing today’s attempts to deliver a single video to every device—the mantra of media delivery, whether it be streaming or progressive download—is the sheer proliferation of devices. What once was a twofold decision (laptop/desktop versus set-top box) is now a 50-fold headache.
Between the myriad delivery protocols, the plethora of video screen sizes, and the varying bandwidths needed to address each, the problem is as much about threading one’s way through the device labyrinth as it is about choosing the best content to deliver.
Consider, for instance, the number of devices we used for testing: Besides nine web-based profiles, we used 23 mobile-based profiles, encompassing numerous devices from Apple’s iPhone and iPad to Google’s Nexus One, Motorola’s Droid, the Palm Pre, and Samsung’s Omnia. These devices represent the primary mobile operating systems—iOS, Android, webOS, and Windows Mobile—on the market today.
Even within the operating system categories, device sizes, screen sizes, and pixel density all vary widely, meaning that each device needed profiles to properly compare content output.
Fortunately, with the exception of the Windows Mobile devices, each device’s operating system supported the MP4 container format, with the AAC or MP3 audio codecs and H.264 video codec. The upcoming Windows Phone 7 operating system will also support MP4/AAC/H.264 within the year.
The benefit of the common thread of the MP4 container format is the ability to offload a bit of the transcoding workflow, in return for increasing the server workflow.
“What’s that?” you’re probably thinking. “I understand transcoding workflows, but what are protocol workflows?”
The reasoning behind protocols has to do with two of the biggest shifts in transcoding to date: the industry’s settling—well, almost settling—on a single codec and the re-emergence of multibitrate transcoding. Early transcoding solutions, in a trend that continues with simple transcoding tools, focused on conversion of a single bitrate video, encoded in a particular codec, to a video with the same bitrate but with a different codec.
Also known as transmuxing, given the need to multiplex files for easier delivery, protocol conversion began in the broadcast world but is finding its way into streaming and progressive download delivery. A move is underway to use standard MP4 files to deliver to any and all of the devices and operating systems listed above.
To do so, the generic MP4 file is converted—at time of delivery—into multiple protocols, from Adobe’s RTMP to the more generic RTP/RTSP and on to the ubiquitous HTTP protocol. Protocol conversion by the media delivery server eliminates the need to convert a single video file into HTTP fragments for each of the three protocols, simplifying the transcoding process as well as the content management headache that comes with delivery to a multitude of devices.
Speed or Quality? Why Not Both?
The testing involved a series of clips and output formats that were chosen for their ability to tax different parts of the transcoding process, from random movement to deinterlacing to varying origin sizes. For general purpose computing solutions that rely on a CPU, the newer Nehalem and Westmere Intel chips are adequate when it comes to providing speed and quality for single encodes to a limited number of output formats.
But what about those instances in which the required umber of outputs rises significantly? To test this out, we ran each test in a series of incrementally rising output: Starting with nine web and IPTV outputs, we then tested 23 mobile outputs and, finally, all of the outputs at the same time, for a total of 32 outputs at one time.
For this kind of heavy lifting, the initial results indicate that specialized solutions, such as those that combine CPU- and ASIC-based processing, can best marry the benefits of general-purpose processing and high-throughput specialty silicon. RGB Networks’ recent acquisition of RipCode, for instance, rounded out the company’s carrier-class video gateways—ideal for MPEG-2 and H.264—with RipCode’s ability to provide CPU-based transcoding for the yet-to be-standardized screen sizes and bitrates inherent in the Wild West of mobile video delivery. Optibase, recently acquired by VITEC Multimedia, appears to be headed toward a similar complementary model to VITEC’s current product offerings.
2010 proved to be a banner year for companies looking to bolster their strengths or move into new markets.
Companies and Suppliers Mentioned