Improving QoE for OTT: A Review of Netflix’s Techniques
OTT video consumption is exploding across the globe, with the latest stats from Parks Associates showing that more than 310 million connected households will have at least one OTT service by 2024. What’s especially grown in size? The number of live OTT services available. Today’s consumers are watching live sports and other major events through their video streaming service, and they expect a high-quality experience, similar to broadcast. This blog will examine why benchmarking is vitally important to the success of OTT services and how Netflix has set the gold standard.
Benchmarking With Netflix
Quality of Experience (QoE) is the magic word that will make your subscribers stay watching your content or leave if it is not meeting their expectations. QoE is defined by a mix of many parameters but the most important are: start-up time, no rebuffering and low end-to-end latency (for live content). For VOD assets only the first two apply.
Netflix is considered today to have the best user experience for pure OTT delivery for VOD content. If a consumer has a good experience with Netflix and a poor experience with a different OTT provider on the same network, the other service will not be successful. The situation today is the result of a long and well-organized strategy to develop best-in-class technologies for offline video encoding on top of what is natively offered by the OTT standards or recommendations. Netflix has a full arsenal of weapons at its disposal to deliver a superior QoE for OTT. Let’s take a look at all of the different video optimizations that Netflix has deployed.
Per-tile/per-chunk encoding: On the encoding side, Netflix was the first to deploy “per tile” and then “per chunk encoding” in 2015. These encoding techniques, which are partially based on the same concepts as Harmonic’s content-aware encoding (CAE), have been shown to deliver a consistent savings of 40 % vs CBR for UHD ABR. Netflix is using HEVC only for UHD and is preparing to deploy AV1 for all of its devices in the 2019-2020 timeframe. With this evolved approach for offline encoding, Netflix can provide the best quality at a given bitrate for every scene of content. This gives the industry a target but it’s important to keep in mind this was made possible for VOD asset encoding, not for live. Doing the same for live is an interesting challenge that will require the usage of additional tools like machine learning to make it possible in a reasonable CPU processing envelope.
Open Connect: At the edge caching point of the workflow, Netflix is using its Open Connect program, deployed in 2012, to boost QoE for OTT. Open Connect puts Netflix-owned caches in ISP networks. Based on information shared by Netflix, the full catalog can fit on one server. With redundancy, this equates to two servers per ISP PoP. Based on this, it’s safe to say that Netflix is not just an OTT service provider anymore; rather they practically have a managed network for distribution to the home.
Client optimization: Netflix has not published anything publically about its client optimization technology, but we know they have a strong client team that develops best-in-class applications ensuring a consistent user experience on all the consumption platforms. Many client-side optimizations are possible and although they are not disclosed, they make Netflix a frontrunner.
Network monitoring: This is the last but not least point of optimization/benchmarking when you want to deliver premium content to millions of subscribers. Netflix has not published much on how it monitors its network, but this is expected to be a major subject of research, as in the U.S. alone it serves 58 million subscribers.
While standard OTT services are far from having all of the tools to match the QoE provided by Netflix, advanced OTT services can not only compete with Netflix but also provide an experience that is close to broadcast for live services. To do that, they need to employ similar video optimization techniques like CAE, use newer codecs such as HEVC and AV1, traffic prioritization, client acceleration and network monitoring. Doing all of this for live content is the next endeavor the streaming industry should make to build successful and profitable OTT live services.
About Patrick Gendron
Patrick Gendron is Director of Innovation at Harmonic for Digital Television applications. He joined Harmonic with the acquisition of Thomson Video Networks. Patrick recently moved from managing the Harmonic R&D Innovation team to the Marketing Innovation & Evangelism team and is Harmonic’s representative at DASH IF, DVB TM and Streaming Video Alliance. Previously, Patrick held senior program and engineering management positions in the digital television headend domain, with international R&D management activities, at Grass Valley and Nextream. He started his career as a research engineer at the Laboratoires Electronique de Rennes (Thomson CSF) where he developed new technologies for professional video transmission over optical fiber (long-haul, single-mode links). As digital technology was maturing for television applications, he moved to Thomson Broadband Systems in a project management role for a number of first-generation digital TV products such as satellite modulators and contribution MPEG codecs. Patrick is a graduate in Computer Science and Telecommunications from the Ecole Supérieure d’Electricité (Supélec).