A Buyer's Guide to Non-Real-Time Hardware Transcoders
We started out the discussion on real-time or live transcoders by talking about whether live transcoding was needed, so that's a good point to pick up the conversation in this Buyer's Guide to non-real-time transcoders.
Live hardware transcoders, such as format converters, are split between server-based and stand-alone hardware-based solutions. Server-based solutions are able to handle everything -- including frame rate, output resolution, aspect ratio, and segmentation -- but require the additional latencies that streaming solutions have grown accustomed to.
By comparison, stand-alone hardware-based transcoding solutions can be used to either do live (real-time) conversion -- at a fairly high price point for nothing more than acting as a dedicated, single-purpose format convertor -- or non-real-time transcoding. The latter group, which convert files rather than live streams, is the one we will focus on for this Buyer's Guide.
Faster Than Real Time?
When we talk about non-real-time transcoding, it doesn't necessarily mean slower than real time: Just like there are integers on either side of zero, known as positive and negative integers, transcoding systems can exist on either side of the real-time equation.
Some systems can convert files in faster-than-real time, transcoding an hour-long file in much less than an hour's time -- including the insertion of advertising content.
Assess Your Codec and Format Needs
Even if your live encoder -- or even your live format converting transcoder -- doubles as a file-based transcoding solution, the number of formats or codecs may be limited by the encoder. For instance, an H.264 encoder may not support WebM file-based transcoding, since WebM isn't integral to the primary use of the H.264 encoder.
Check Those Profiles
Another significant question to ask is how many codecs and formats need to be converted at one time. After all, the number of codecs on the market may be shrinking, but the number of discrete and disparate devices is rapidly growing. The typical transcoding session may require upward of 30 different profiles, including varying aspect ratios, frame rates, and resolutions.
The question of how many transcodes also has bearing on the next topic, adaptive bitrate, which is taking over almost all aspects of the streaming world.
Adaptive Bitrate Requirements
A corollary question to ask is whether your workflow has the potential need for adaptive bitrate delivery (ABR). The answer these days is almost undoubtedly "Yes."
While traditionally thought of for live encoding, ABR in its various competing formats from Adobe, Apple, and Microsoft is fast becoming a necessity for any type of content. When we say any content, we also mean on-demand content that has been stored in a mezzanine file format. Server-based transcoding solutions -- as well as transmuxing and transrating, which we discuss in detail in another Buyer's Guide -- receive a request from the end-user's player, determine the necessary resolution and bandwidth requirements, and then transcode/ segment the mezzanine file to the appropriate output for the client device.
The good news is that ABR has been standardized in the form of Dynamic Adaptive Streaming over HTTP (DASH) thanks to the hard work of all of the competing ABR companies noted previously. This new MPEG standard is further being distilled down to a fragmented MPEG 4 file (fMP4) format, centered on H.264, called DASH 264. This subset should see ratification early this year.
One Platform or Many?
We mentioned WebM previously to bring up another decision point about hardware-based transcoders. Many hardware-based solutions will support a number of legacy codecs and formats but will be slower to implement new codecs and formats. The process normally involves reaching critical mass in terms of end users who are able to view content in a particular format, on a particular platform, and then confirming that both the codec and the platform (e.g., a web browser) have stabilized before committing into silicon a recent codec that's yet to be standardized.
Single, Pairs, or Bunches of Transcoders?
How important is a single-purpose transcoding device? In terms of deciding between server-based or
appliance-based transcoding solutions, the question of an appliance versus software on general-purpose hardware often comes up.
For those involved in broadcast or high-profile events, a single-purpose device that can be dedicated to a task is probably the most appealing. There's just something about an on-off switch and the instant response of a dedicated appliance that warms the heart of a broadcast engineer in the way that a server-based solution never can.
On the flip side, if time is of the essence in terms of multiple codecs, then it may behoove you to consider several transcoders, each one accessing a single mezzanine file so that all transcodes can be performed in parallel.
A few hardware-based transcoders offer the time-saving ability to decode the mezzanine file once, hold it in memory, and then use it for multiple similar-type transcodes. This is often seen in conjunction with ABR, where groups of pictures (GOP) alignment is critical between various bitrates, but some systems will also speed the overall workflow by assessing redundancies across profiles and sharing the decoded file with each matching -- or nearly similar -- profile. This smart approach to transcoding goes a long way toward significantly removing the burden from the operator.
Non-real-time transcoding solutions come in all shapes and sizes, from small appliances and single-rackmount (1RU) enclosures to sizable server options and even cloud-based transcoders that require no capital expenditure.
No transcoding solution covers all the possibilities, as cliché as that may sound, so it's good to shop around based on particular workflows. In addition, as new standards emerge, we also recommend that potential buyers balance the current options with the ability to expand the capabilities of their transcoder of choice.
As such, some software solutions on general purpose hardware may be a safer bet if the workflow is not fully established, allowing for tweaks and enhancements to the overall workflow before committing to the cost and rigidity of a specialized hardware transcoding solution.
Stay tuned to StreamingMedia.com throughout 2013 as we explore, test, and verify claims on new transcoding tools that emerge to challenge current transcoding champions.
Six Questions to Ask When Purchasing Hardware Encoders
- Is real time too slow?
- What formats and codecs are you working with?
- One transcode or many?
- Do you need ABR output?
- How many platforms will you be delivering to?
- Will you need a single transcoder, a pair, or more?
This article appears in the forthcoming 2013 Streaming Media Industry Sourcebook.