Best Practices for Premium Video Streaming, Part 2: Content Preparation
Online audiences demand flawless playback experiences, particularly when it comes to premium content. This requires online content providers to carefully plan and execute during the content preparation and ingest phases. The ramifications of any missteps early in the over-the-top (OTT) video distribution chain can result in countless issues during playback. With the cost of acquiring content rights on the rise and audience expectations for quality of experience increasing, the stakes are higher than ever.
Preparing video content for delivery in a cost-efficient manner at the highest quality levels requires adherence to best practices for transcoding, packaging, manifest preparation, ingest, and storage. It’s also essential for implementation of personalization, navigation, dynamic advertising, security, and other functions that contribute to meeting audience expectations and driving revenue.
In this second installment of “Best Practices for Premium Video Streaming,” we’ll explore content preparation, the interplay of origins with CDNs, and steps to maximize quality of experience.
Transcoding & Bitrate Profile Creation for OTT Video
Fundamentally, the transcoding process requires that content be prepared at the highest quality levels and optimized prior to delivery. This entails preparation of streaming segmentation, bitrates, bit depth, and framing policies for live as well as on-demand use cases across all internet-connected viewing devices. The quality of the video renditions produced during transcoding are critical to maintaining quality downstream during delivery and playback. However, it’s not enough to create the highest quality video renditions in all cases. Depending on the situation, video quality can be too high, resulting in wasted bandwidth; or it can be too low, resulting in pixelated playback.
Optimizing Bitrate Ladders for Optimal Video Playback
To meet playback quality expectations, great care should be taken during the transcoding process to select the optimal bitrate ladder for a given piece of content. Some content providers opt for a one-size-fits-all approach, creating a similar bitrate ladder for their entire VOD catalog, but this can result in unnecessary storage and delivery costs and result in suboptimal playback quality.
The best practice for video on demand calls for context-aware encoding, which establishes an optimal bitrate ladder for each title in a catalog. Each scene is produced in multiple quality levels and the bitrate adapts as needed. Using this approach, the perceived quality during playback is the same, but requires less bandwidth. Increasingly, content providers are using machine learning to derive optimal bitrate ladder selections.
The best practice for live streaming scenarios remains the creation of a single bitrate ladder for everyone, which can be trimmed for certain viewing scenarios. Trimming the number of available renditions can be handled further down the line via manifest manipulation or through advanced player logic.
Addressing the Next-Generation Codec Conundrum
While H.264/AVC has long been the dominant codec used in video streaming, the advent of 4K ultra high-definition (UHD) has encouraged content providers to look at emerging, alternative codecs. That includes H.265/HEVC, which consistently delivers a peak 50% improvement in compression efficiency over AVC.
With industry confidence in the superiority of 4K UHD stream quality and built-in support from the latest Apple smartphones and TVs, HEVC has gained traction in advance of wide-scale UHD availability. HEVC is also gaining on the encoder side as content providers and distributors replace or expand encoding assets.
With that said, the emergence of other next-generation codecs indicate we may not reach the industry consensus with HEVC that we had for AVC—at least not for some time. Royalty costs remain an obstacle for HEVC, while high-performing options like Google’s royalty-free VP9 and the Alliance for Open Media’s AV1 are both supported by a range of devices, web browsers, and industry leaders, giving HEVC some stiff competition.
All of this uncertainty points to a need to pay closer attention when choosing codecs. Since not all codecs support all devices, it is advisable to consider multiple codecs when the economics make sense. The rule of thumb is to consider implementing new codecs at the point where the savings on delivery costs reach the incremental cost of storage and encoding. The economics of any use case should always be the deciding factor in choosing the right mixture of codecs to support optimized delivery to all target devices.
Manifest Preparation and Packaging for OTT Video: HLS, DASH & CMAF
For a given piece of content, each set of encoded renditions must be packaged with a manifest file that allows targeted clients to acquire and render the content using the recommended streaming formats that work with their players.
The process must also accommodate the complexity that characterizes preparation of adaptive bitrate (ABR) manifest files. This includes creating subsets of manifest files that utilize compilations of metadata for directing audio track captioning and linguistic subtitling; selecting digital rights management modes; descriptors associating advanced features with specific content segments; and placement options for dynamic advertising.
Thanks to accelerating market adoption of Dynamic Adaptive Streaming over HTTP (MPEG-DASH) and Apple’s HTTP Live Streaming (HLS)—along with vanishing support for Microsoft’s Smooth or Adobe’s HTTP Dynamic Streaming (HDS)—best practice is for distributors is to utilize HLS and/or DASH streaming formats for the vast majority of use cases. However, there are still some corner cases that require the use of Smooth.
The master manifest file—identified as the Master Playlist in HLS and the Media Presentation Description (MPD) in DASH—provides the player information about the audio and video codec, available bitrate profiles, segment sizes and sequencing, details relating to captioning, subtitles, and advertising.
All of these elements must be presented and synchronized to ensure precise, smooth playback on client devices. Best practices stemming from common use of fMP4 containers with HLS and DASH and the need to maximize CDN efficiency is to utilize the emerging common media file format (CMAF).
The Emergence of CMAF
The Common Media Application Framework (CMAF) makes it possible to encode a video in multiple bitrate profiles with uniform segmentation utilizing the fMP4 container for streaming over either HLS or DASH. As an Application Format (AF) in the ISO lexicon, CMAF provides a lightweight framework that does not introduce new processes, but rather serves to combine existing formats and standards in a new way. With formal standardization achieved in 2017, CMAF needs to be taken into account as distributors adapt their workflows to accommodate best practices for 2018 and beyond.
For on-demand scenarios, the best practice entails using DASH and/or HLS, fMP4 containers and CMAF. Content providers can utilize one set of audio and video files packaged in CMAF along with two manifests (one for HLS and another for DASH) that reference the file. This can help lower content preparation and storage costs while providing better CDN efficiency thanks to increased cache hit ratios.
In the case of live streaming, the best practice calls for the use of DASH and/or HLS, fMP4 containers, and CMAF. In on-demand use cases, CMAF enables content providers to utilize a single set of live audio/video files with two manifests to reference the file. However, in the case of live streaming, CMAF can help content providers achieve lower content preparation and ingest costs. Best practices include consolidating live ingest feeds into origin servers to ISO only, instead of both TS and ISO, which can cut bandwidth in half.
CMAF also has an optional chunked encoding mode for live streaming that, when combined with chunked transfer support through an origin and CDN, can cut end-to-end latency down to a few seconds and also allow latency to be decoupled from segment duration. This option provides an avenue for content providers to achieve lower latency without lowering segment duration, which comes with increasing quality and overhead penalties.
Hosted vs. In-House Origin Infrastructures
Following transcoding and packaging, ingestion of the packaged content onto origin servers that can be accessed by CDNs is a critical next step. To support live or 24x7 linear video services reaching mass audiences, best practice includes using an encoder that pushes content to origin equipped to handle massive call volumes for content over CDN infrastructures. Best practices dictate that the origin service be able to dynamically find the best points of entry based on network conditions, audience locations and other factors; and also support optimal modes of transport to minimize latency with no loss in quality. For on-demand scenarios, it is advised to utilize origin services that offer highly scalable infrastructures working in tandem with distributor workflows to optimize storage for high-performance video use cases.
Providers using their own origin infrastructures must have sufficient capacity operating in pull mode to handle all calls from all CDNs, as well as separate backup facilities in the event of an origin failure at the primary origin facilities. Switchover to backup also requires automated response capabilities, 24/7 monitoring, and high-performance transport connectivity.
Maximizing Quality Assurance for OTT Video
Encoding and packaging for OTT distribution require absolute assurance that those processes are consistently performing as required. This is not only fundamental to achieving end-to-end performance goals—it’s also essential to holding CDN operators accountable when it comes to meeting SLA commitments.
Commercial-grade reliability in the ingestion process begins with having sufficient resources to handle peak loads in primary and backup facilities that have no mutual dependency on the same power resources. Each set of facilities should have secondary routes into targeted origin locations to prevent network-related interruptions. Completely automated failover mechanisms are also essential to ensuring seamless continuity in the event of failure.
Performance monitoring and analytics tools must be employed to provide the comprehensive visibility needed to: identify issues before they cause disruption; compare input and output quality; confirm that latency and quality expectations are met for each video program’s rendition sets; and verify that workloads are properly apportioned to avoid overloading on transcoders.
A growing OTT video market with ongoing technical advances that continually raise the performance benchmark make sustaining competitive edge a challenge. High-value video providers can remain competitive by adhering to these content preparation best practices inclusive of the right expertise, technologies and facilities to meet increasing performance requirements and quality expectations.
[This is a vendor-contributed article from Akamai. Streaming Media accepts articles from vendors based solely on their value to our readers.]
Advances in monitoring and analytics anchor superior OTT user experiences. Here's a road map for pre-event testing that will ensure high-quality performance across the board.
Video playback is where the rubber meets the road. Make sure you're using a player that gives you all the functionality you need to successfully deliver your high-quality content to a fragmented device universe.
Steps for ensuring CDN performance meets audience expectations for OTT streaming
In this first installment in a series of articles looking at best practices for delivering premium video content, we look at managing the first mile.