A Buyer's Guide to Cloud Transcoding
For both live and on-demand video transcoding, there are compelling reasons to move to the cloud. Read this before making a decision.
Learn more about the companies mentioned in this article in the Sourcebook:
Is 2013 the year of encoding in the cloud? Actually, it might be, at least for live transcoding in the cloud, and even on-demand transcoding should see significant migration to cloud facilities.
Taking a step back, there are two distinct markets for cloud encoding: compressing on-demand files and transcoding live streams into multiple iterations for adaptive streaming. Of the two, the case for live transcoding is much stronger, because it solves the outbound bandwidth problem that plagues all live producers and reduces on-site encoding costs dramatically. Plus, because it's live, it doesn't share the biggest problem encountered by those encoding large mezzanine files in the cloud, which is upload time. Notwithstanding upload time, for many reasons, large and small streaming producers alike are increasingly moving to the cloud.
I'll cover both markets in this Buyer's Guide, and for each, I'll start with a look at the types of producers who should strongly consider encoding in the cloud over encoding internally.
Seven Questions to Ask When Considering Live Cloud Encoding
- Does the service output the necessary formats?
- Does the service supply the required features?
- Purchase or software as a service (SaaS)?
- How easy is the service to access?
- How's the quality?
- How's the price?
- How's the service?
Live Encoding and Transcoding
As mentioned, many live event producers share two major problems: getting encoding resources to the event and getting the stream out of the building. Live transcoding in the cloud solves both problems.
As an example, rather than attempting to encode five streams each for Flash and HLS on-site and then to transfer those streams to the streaming server, the producer would send one high-quality stream to the cloud transcoding facility. There, the stream would be re-encoded into the five files necessary for adaptive streaming and then packaged for Flash and HLS delivery. This reduces the equipment cost of the on-site encoders and the outbound bandwidth necessary to support the outgoing streams, a win-win for most live event producers.
SHOULD YOU USE THE CLOUD?
If you produce live events infrequently but need to support multiple-format adaptive streaming, cloud transcoding is a natural, particularly if you broadcast from sites with limited or unstable outbound bandwidth. At the other end of the spectrum, existing 24/7 live producers such as networks and other broadcasters probably already have the necessary encoding gear and outbound bandwidth, so the case may not be as strong. If you're just gearing up a 24/7 facility, however, note that at least one live transcoding vendor plans to offer special pricing to 24/7 broadcast facilities to make their services much more affordable.
The other consideration is quality. Intuitively, live transcoding should produce lower-quality streams than encoding from the original source, since you're working with previously encoded video at a relatively low data rate. The live transcoding vendors that I spoke with assured me that this isn't the case, in part because the codecs used in cloud encoding facilities are generally much more current than those used in existing hardware encoders.
Still, large facilities seeking premium quality with the budget for on-site encoding hardware should compare the quality of the two approaches before making the jump to live cloud transcoding. Conversely, if you don't have the capital expenditures budget and/or have limited outbound bandwidth at your event locations, cloud transcoding may be the only way to produce multiple-format adaptive streams, even if the quality suffers to some degree.
There will be a great deal of uniformity here, but this is worth checking up front, particularly if you're seeking to distribute beyond desktop and mobile devices into OTT and other markets.
Check early with your candidate vendors if you need features such as DRM or closed captions, since these won't be universally supported. One feature that's relatively unique is the ability to dynamically adjust the data rate of the outbound stream according to the changing bandwidth conditions, which can be very useful when operating under dynamic conditions. Note that you'll have to use a specialized encoder on-site to access this feature.
PURCHASE VS. SAAS
While all cloud transcoders are available as a service, some are also available for purchase and installation in your own private cloud or leased cloud facility. If you're interested in purchasing, get that question on the table early.
All services involve inputting network parameters for CDN or streaming server targets, which should be old hat to most live event producers. However, some services will require scripting, or accessing the service via an application programming interface (API). This won't be a problem for most large organizations, but it could be an absolute bar for some smaller producers. Before choosing a vendor, work through the live transcoding workflow, and make sure that it's within your technical capabilities.
All vendors claim top quality. If you find two or more vendors equal in all other parameters, you can compare their quality by running a short event on each and comparing the quality of the encoded streams.
While paying less is always good, simple and easy-to-estimate pricing is also important. Pricing is simplest when it solely involves factors such as the number of streams and the duration of the event. It gets more complicated when it involves the number of encoding instances or similar factors. Either way, the vendor should offer a pricing calculator that makes it simple to estimate the price of any event.
Make sure the service centers are staffed during the times you'll be hosting your live events, and check what it costs for real-time telephone support. Consider calling the support centers to get a feel for the technical level of the support personnel. Most are trained to solve simple problems, such as login or server address problems, but they may not be sufficiently knowledgeable for more complex issues such as the optimal encoding configuration for your source encoder.
Now let's shift our focus to the on-demand side.
Nine Questions to Ask When Considering On-Demand Cloud Encoding
- Purchase or SasS?
- Does the service output the necessary formats?
- How easy is the system to use?
- How extensive is the API, and how easy is it to use?
- What's the price, and how simple is the pricing model?
- How stable is the operation?
- What's the guaranteed speed and throughput?
- How's the quality and service?
- Does the service offer the necessary features?
As with live transcoding, let's focus first on who should consider cloud encoding and transcoding.
SHOULD YOU USE THE CLOUD?
The economics for on-demand cloud encoding are clear; rather than investing in your own encoding hardware and software and developing internal encoding expertise, you can save money by outsourcing both to a cloud facility. Perhaps you'll pay a few hours' worth of consulting time to identify the best encoding presets, but it's cheaper than employing your own compressionist, and you never have to worry about buying, upgrading, and supporting your own encoding hardware or software.
The 800-pound gorilla for cloud encoding is upload time: Mezzanine files are large, and their upload times can be quite lengthy. Of course, if the bulk of the files that you encode are uploaded to you, rather than generated internally as with user-generated content sites or service bureaus, upload time isn't an issue. Cloud encoding is therefore a natural. Ditto if you store your content in the cloud already.
Other sweet spots exist at both ends of the volume spectrum. If volume is low, upload time isn't so much of an issue, and it makes sense to offload encoding chores to a service. If volume is very high, encoding speed is typically critical, and even with the upload time, the ability to tap into hundreds of cloud instances may speed encoding overall.
Or, if encoding chores are limited, but episodic, like having to re-encode a huge library for a new target device or format, you can send a hard drive to Amazon and let them copy the files to an S3 bucket, solving the upload time dilemma. Obviously, as upload speeds increase, and connectivity prices decrease, the 800-pound gorilla will shrink in importance.
PURCHASE VS. SAAS, FORMATS
You'll have the same issues as mentioned previously: Are you looking to install your own encoder in a cloud instance or SaaS, and does the service output the required files and formats?
EASE OF USE
As with live transcoding, the ease of use varies dramatically by vendor. Some enable you to create presets in familiar desktop applications and upload files à la YouTube; some require that you support their APIs for any operation at all. It's obviously best when the technical expertise required to work with an encoding service matches your internal capabilities.
Most high-volume users of cloud encoding services automate operation via the cloud encoder's API, which differ in scope and ease of use. Review the API to make sure that it includes all critical elements, such as notifications for all relevant encoding events, and precise error codes that detail any problems to make them easy to resolve. Have your technical personnel review the API to estimate how long it will take to integrate, and to identify any potential feature gaps.
Again, simpler is better. Most of the larger vendors are moving toward simplified fees based on the size of files imported and created, with performance aspects such as encoding speed dictated by service-level agreements and volume commitments. Other vendors look at the actual CPU instances that are created for a particular job, which forces the client to make the trade-off between encoding speed and encoding cost.
Though 99.9999% uptime probably doesn't matter for smaller clients, large media companies need guaranteed throughput and therefore redundancy. For this reason, some of the larger cloud encoding vendors operate on Amazon Cloud and Rackspace, as well as having their own private encoding facilities.
Regarding the robustness of the encoding facility itself, if you have a library of problem files that have crashed your existing encoding programs, try uploading them to your candidate cloud facilities to see how they are handled. Errors happen, but look for clear error messages and notification capabilities that detail what's happening and why.
Speed and Throughput
One key benefit of the cloud is its ability to scale capacity to meet demand, and then shut off capacity when not needed. If you're a news organization or UGC site with an emphasis on encoding speed, scrutinize your service-level agreements to ascertain the throughput commitments the cloud encoding service is willing to make. This may include guarantees relating to queue time and encoding time.
If speed is critical to your operation, try uploading several large files to different suppliers and timing the upload, queue, and encoding process. Check each service two or three times at various hours during the day and you should get a feel for comparative performance.
QUALITY AND SERVICE
See the previous descriptions.
Closed captioning and support for DRM are obviously as critical for VOD as live. Beyond this, if a substantial component of your video library exists in the cloud, you may require basic editing features such as trimming, splitting, and concatenating video files, along with text and image overlay. You'd also want these features available via the API so you can automate them.
Another group of features centers around translation. If you're a multinational organization distributing videos worldwide, you may need translation assistance to create voiceovers in other languages or via captioning. The ability to integrate these services into the encoding workflow is much more efficient than handling it as a pre- or post-process.
Yet another group of features centers around format support. Virtually all cloud services can accept H.264 files, but what about Avid DNxHD or Apple ProRes? Even fewer will accept QuickTime reference movies, which may be the most efficient way to upload multiple iterations of the same basic content.
Finally, if you need both live and on-demand encoding services, it's usually simplest and cheapest to get them both from the same vendor. So if you're in the market for both types of services, check to see if your candidate vendors support them both.
This article appears in the forthcoming 2013 Streaming Media Industry Sourcebook.
Cloud services change the value proposition for online video creators and distributors. Dive deep into the cloud to learn about the savings and features it offers.