Archives

In 2020, over 80% of video streaming traffic still used the H.264 codec, which seems staggering when you consider we’ve had HEVC and VP9 around for quite a while and there are even more advanced options on the way (AV1, EVC, VVC, LCEVC). This isn’t news to anyone who is close to this stuff and has been widely discussed before, but we should be taking a frequent look at what continues to block progress in one of the most fundamental components of the streaming media stack.

I recently conducted a survey on behalf of compression experts V-Nova, and then moderated a webinar (which you can check out here) to discuss the results with Luis Vicente, CEO of sports OTT operator Eleven, and Karam Malhotra, Global VP Revenues for Asian top 10 video app SHAREit. The survey’s focus was on key barriers to improve the quality-of-experience operators are serving their customers. Over 200 senior business and technical decision makers responded from a broad range of industry verticals and the results provided some real food for thought. You can download all the results from the survey for free, at this link.

From the results of the data, it was enlightening to hear which components of their video delivery systems people think are most critical for QoE. CDN comes out top here and in my view that’s simply because it’s the easiest to measure and everyone’s first thing to look at. This is indicative in part of services tending to consider components like video codecs as a fixed constant because changing them is assumed to be a long-term project, which it doesn’t have to be. Interestingly, however, video codec came out close second in terms of relevance to QoE.

It stands to reason that deploying new compression tech and streaming equivalent quality at substantially lower bitrates improves QoE all round by accelerating start-up times, reducing buffering incidents, increasing the % of viewers that can receive higher quality ABR profiles and reducing the traffic load on those CDNs that appear to be taking most of the heat for QoE issues.

So, what are the barriers holding the industry back from moving video compression forward? As you can see below, the widely publicized royalty costs regarding uncertainties for codecs like HEVC have generated a lot of hesitancy.In addition, the increased operating costs of encoding and processing are an important barrier too. Furthermore, quite a few of the survey options on this one all point to the risk of incurring significant new costs from heavier compute requirements and the additional complexity that comes from duplicating workflows to serve different types of device. For many, royalty issues are so loud because in truth most people see unclear business benefits from deploying new video codecs and immediately incurring cost increases.

Despite these barriers to adopting new compression technologies, 33% of respondents reported that they are wanting to trial new options in the next 12 months, a clear acknowledgement of the need to find ways to progress somehow. This is possibly being driven by the clear need to differentiate in this evermore competitive entertainment space by launching more premium services like 4K and HDR, which 29% said they were intending to do in 2021.

All of these survey results brings us to the question that really prompted me to do this survey in the first place, which is what would actually convince companies to upgrade their video encoding and reap the available QoE benefits? What are the key factors and where do they need to be to tip the balance?

Respondents were asked to quantify what sort of compression efficiency benefit would be needed (most easily expressed as how much could you reduce the bitrate and maintain equivalent quality). The average came out at 31% which is pretty well aligned with the typical improvement between codec generations (e.g. AVC to HEVC or VP9 to AV1) and also the uplift typically provided by using LCEVC enhancement with AVC or HEVC.

Respondents were also asked what reduction of transcoding costs would be required to convince them to upgrade (that big barrier from earlier). That came out at a very similar average of about 30%. The combination of the two answers is interesting, suggesting that transcoding efficiency is considered roughly as important as compression quality.

I then explored what impact on operating profitability would be generated by an improvement in compression tech meeting these sort of criteria. More than half of respondents felt that optimizing compression technology in this way would provide more than a 25% improvement in profitability. This is surely material, especially if we consider that improved QoE should also produce a benefit to the top line from better user retention and engagement.

A key takeaway from all of this data is that 52% of respondents don’t know what the primary cause of losing viewers from their service is, a key point that warrants more discussion another time.

The general theme that emerged from this survey is one of hesitancy to upgrade video compression because of an array of perceived complexity and cost risks. At the same time, respondents do recognize that improved compression could produce material business benefits. It’s the first time I’ve seen such a clear picture of the criteria that must be met to enable significant progress and an acknowledgement that the benefits to profitability are substantial. It’s also the first time that I see such a clear highlight on the importance of transcoding efficiency, tightly connected to the processing requirements of video codecs.

Objectively it looks like MPEG-5 LCEVC is well positioned to tackle the barriers this survey highlighted. A blend of significant improvements in compression efficiency, simple licensing terms and crucial reductions in the operational processing cost of encoding and delivery seems to be what’s needed to unlock things. It’s going to be interesting to see how that all plays out.

If you want to see all the results from the survey, you can download them for free, at this link.

On Friday January 8th, Roku announced it had acquired Quibi Holdings LLC, the company that holds all of Quibi’s content distribution rights for their library of 75 original shows. All of Quibi’s content will be added to Roku’s free ad-supported channel and as part of the deal, Roku is not permitted to change the format of the shows and must keep all the episodes separate.

Normally for a deal like this we’d all be running calculations on how many ads Roku has to sell, at what CPMs, and with how many viewers, to see a positive return from the deal. But in this instance I would argue that we can value the deal for Quibi’s content with different metrics. The day Roku announced the deal their stock closed up $19.84 a share, adding $2.52B to Roku’s market cap – in one day. Roku’s stock might have gone up that day even without the news, so there is no way to know the exact impact, but Roku’s stock was clearly impacted in a positive way from the news.

The sale price for Quibi Holdings hasn’t been announced but I heard the deal size was “around” $50M, [not verified] while others are reporting the price was less than $100M. Even if Roku paid $100M for Quibi’s content, the return on their investment based on the increase in their market cap would be 25x what they spent. If Roku paid $50M for the content they got a 50x return for their investment, all for just making an announcement. Even if the rumors on wrong on what Roku paid for Quibi’s content and the price was higher, Roku still made back their investment many times over.

If you’ve followed Roku’s stock, you know just how much Wall Street is in love with it and how much it moves based on news. The stock saw positive benefits when they announced the deal with NBCU for support for PeacockTV and again with WarnerMedia for support of HBO Max. We don’t know exactly how much Roku’s stock may have already gone up without the news, but even if it simply doubled the raise that day, Roku sill got a 12x return on their investment. Whether they sell any ads across Quibi’s content and generate any ad revenue almost doesn’t even matter at this point.

If you were still on vacation last week to start the New Year, here’s a rundown of some of the interesting news that took place you might have missed:

• My hands-on review of Discovery+ launch: Massive content library, good streamlined UI, parental controls, downloads and additional smart TV support to come. 4K streams max out at 13.5Mbps, four concurrent streams per account. More details: https://www.streamingmediablog.com/2021/01/discovery-plus-launch.html
• Roku announced it has acquired Quibi Holdings, LLC, and will put that content in the free Roku Channel with advertising. I heard the deal size was “around” $50M, others are reporting the price was less than $100M. Roku’s stock closed up $19.84 on the news, adding $2.52B to Roku’s market cap, in one day. https://www.linkedin.com/posts/danrayburn_streamingmedia-rokutv-quibi-activity-6753338314524254208-d4ce
• fuboTV announced preliminary Q4 2020 revenue and subscriber growth with fuboTV Paid subs at year-end expected to exceed 545,000, an increase of more than 72% year-over-year. Prior guidance was 500,000-510,000 subs. https://ir.fubo.tv/news/news-details/2021/fuboTV-Announces-Preliminary-Fourth-Quarter-2020-Revenue-and-Subscriber-Growth/default.aspx
• In a first time experiment with CBS, Sunday’s NFL Wild Card game was broadcast on Nickelodeon with a format and interface that targeted kids. The adult announcers, which were joined by 14 and 15 year old Nick stars, were making less references to football and more to SpongeBob, Good Burger, homework and favorite ice cream flavors. More details: https://www.linkedin.com/posts/danrayburn_nflfootball-viacomcbs-nickelodeon-activity-6754158545140236288-zND_
• Netflix to raise pricing in the UK next month. Its standard monthly package will go up from £8.99 to £9.99 and its premium package will rise from £11.99 to £13.99. The basic plan will remain £5.99. https://www.bbc.com/news/technology-55563970
• Comscore announced a strategic investment by Charter Communications, Qurate Retail and Cerberus, which wipes out $204M in debt. As part of the deal Comscore gets a 10-year data license from Charter and becomes the preferred local measurement provider for Charter’s advertising business. More details on the deal here: https://www.linkedin.com/posts/danrayburn_comscore-announces-strategic-investment-by-activity-6753356616789184512-Ta2Z
• Some key takeaways from Discovery’s CEO interview on CNBC talking about the launch of Discovery+ in the US. He didn’t give out any subscriber projections but did highlight some other details here: https://www.linkedin.com/posts/danrayburn_discovery-ceo-on-2021-plans-for-discovery-activity-6751916742202671104-cTiv
• In 2019-20, Netflix invested $400M (£293M) in creating original programs and released more than 30 pieces of original Indian content. Amazon Prime now has customers in more than 4,300 towns and cities across India and has about 50 shows in development. More details: https://www.linkedin.com/posts/danrayburn_in-the-land-of-storytelling-netflix-and-activity-6751617896595644417-zeQX
• A behind-the-scenes look (video) at the workflow ESPN engineers implemented to enable the announcer to call the Sugar Bowl live from his home: https://twitter.com/ESPNPR/status/1345429498758639617?s=20
• Roku announced preliminary estimated data for Q4 ending Dec. 31, 2020: Roku ended the year with 51.2M active accounts, up by approximately 14M accounts in 2020. https://blog.roku.com/en-gb/50-million
• NBCUniversal and Charter Communications announced a renewal deal which includes Charter broadband and video customers getting an extended free trial of Peacock TV. The free deal hasn’t launched yet so we’ll know how long it lasts when it launches. https://ir.charter.com/news-releases/news-release-details/nbcuniversal-and-charter-communications-announce-full-portfolio

If you’re interested in working at Disney Streaming Services, they have an immediate opening for a Sr. Software Engineer – Solutions Architect. I can make a direct introduction to who your boss will be, so if you are qualified, contact me (dan@danrayburn.com) if interested. Full job details: https://jobs.disneycareers.com/job/new-york/sr-software-engineer-solutions-architect/391/17366699

Today, Discovery+ launched in the U.S. for $5 a month with “light” commercials, or $7 commercial free, allowing up to 4 concurrent streams per account and a 7-day free trial. The company is showing “up to 5 minutes” of commercials per one hour of content, hence the “light” branding. This is similar to what NBCU is doing with Peacock TV, with a lower ad load on streaming services when compared to cable TV, which averages 20 minutes of ads per hour of content. As expected, the company is doing a huge marketing push across all social platforms and on TV, promoting the service with their “stream what you love” branding.

Discovery+ is available on platforms and devices including Apple TV, Fire TV, Chromecast, Roku, iOS, Android TV, Xbox consoles One/X/S and Samsung smart TVs from 2017 and newer. It is not yet available on smart TVs from Vizio, Sony, LG or PlayStation consoles but support for other brands of smart TVs is expected this year. The biggest strength of Discovery+ is their massive content library of over 55,000 episodes, with 2,500 series, across brands including A&E, HGTV, Food Network, TLC, Lifetime, OWN, Travel Channel, Discovery Channel and Animal Planet. They also plan to offer more than 1,000 hours of Discovery+ original content this year. Discovery also announced today a multi-year carriage deal with Vodafone in which Discovery content will be made available to Vodafone mobile customers in 12 European markets. Discovery+ was already live in the UK on Sky and previously announced a deal with Verizon in the U.S., where some Verizon customers can get Discovery+ at no cost.

I’ve been hands on with the service for a short time and so far, have found the UI intuitive and easy to navigate. It has content categories as the top nav, but also breaks out channels, which then features content only from that one brand. Functionality like continue watching has worked flawlessly for me, along with starting back up in the right place when moving to a different device. The left side nav is where you go home, manage your account and search for content, along with making a favorites list. Browsing allows you to search by brand and under each brand you can see what’s trending. Not all content is offered in 4K, which isn’t surprising being there are some old shows in the catalog, but I found plenty of other shows had 4K quality. Discovery didn’t say what percentage of the entire catalog has 4K support, but for the content that is in 4K, the max bitrate is 13.5Mbps. None of the content at launch is available for download and they don’t yet have any parental controls you can enable, but both of those options are coming to the service this year.

Of course with any new OTT launch in the market you have those in the media and on Wall Street that are going to base the success of Discovery+ on the wrong metrics. We are already seeing Discovery+ compared to HBO Max, Netflix and Disney+, which makes no sense. Some are asking if Discovery+ can show similar success to Disney+ when it comes to subscriber numbers, but that’s the wrong metric to be using. Disney+ is very niche content, targeting mostly kids and family friendly content and had about 8,000 episodes and movies at launch. Discovery+ has much more breath of catalog and is targeting adults across a wide range of content. 90 Day Fiancé is the number one show on TV with the 15-49 demographic and that’s not the type of content you will see on Disney+.

On CNBC today, Discovery’s CEO said he expects Discovery+ to “be very very big” and will have “big scale”, but didn’t define what that means. The company is not giving out any initial projections on the number of subscribers they are anticipating, but did say that “over the next couple of quarters”, they will give out subscriber numbers and growth. I see some making the ridiculous statement that Discovery will need to hire firms to help Discovery “understand” the streaming market and to be able to compete with Disney+ and others. Streaming is a technology, it’s not a “service”. The service is content, which Discovery knows very well. The medium that’s delivering the content is simply streaming technology.

Discovery knows all about operating channels, producing and distributing content, creating original shows, doing integrated marketing across multiple channels and tracking viewership. Discovery has a 20% share of cable viewership and the company says “nearly 250 millions hours of Discovery content is watched on TV every day.” TLC, beat every one of the cable news nets in the third quarter, based on time watched and their portfolio is #1 for average time spent across all TV entertainment in the US. That’s a lot of viewership they can build on. Discovery is not new to the content game and doesn’t lack any expertise or strategy, with the company having hired execs from places like Hulu, Amazon, Microsoft and others. For some to suggest that Discovery is “late to the game”, doesn’t understand the D2C market, or lacks any kind of streaming expertise – that’s simply not accurate. I won’t make exact projections on the number of subs Discovery will sign up, we need to give the platform time to grow. But Discovery+ is a global service, with a deep catalog of content, big marketing tie-ups and we should expect them to have tens of millions of subs by the end of this year.

Here’s my presentation entitled, “Video CDN Trends: Latest Pricing, Customer Challenges and Growth Opportunities” from the Mile High Video workshop event this month.  Happy to answer any questions in the comments section or you can email me directly at dan@danrayburn.com. Apologies for the rough voice, had a cold. #mhv2020

Recently Cloudflare announced that their researchers have been working to improve internet privacy and security through a DNS protocol called Oblivious DNS over HTTPS, or ODoH. While the announcement suggests that it will improve internet privacy, their proposal can actually lead to significant privacy issues, if it’s adopted. In addition, as it is written today, ODoH is likely to seriously impact internet performance or force providers to invest in software and process changes because it strips information that ISPs and CDNs require to do effective and efficient mapping to ensure performance for their users.

The ​IETF​ is an internet standards body made up of ​an international community of network designers, operators, vendors, and researchers concerned with the evolution of the Internet architecture and the smooth operation of the internet.​ Community members bring proposals for internet standards to the IETF through published “drafts”, mailing lists, and typically through in-person meetings. Proposals are submitted, and require working groups to pick them up in order to progress them into standards.

At this point, ODoH is an early-stage draft proposal which has been discussed in the “dprive” (DNS Privacy) working group but not yet adopted there, meaning that it is in a formative stage. Common practices to encourage the IETF to adopt work in a working group include deploying production implementations of proposed technologies, and driving interest and support from vendors and businesses, which is what Cloudflare is doing with their announcement.

So why is there contention around the proposal? Cloudflare’s head of research Nick Sullivan has ​stated​ that ​”sweeping technical changes to the internet will inevitably also impact the technical community. Adopting these new protocols may have legal and policy implications.” Some of these legal and policy implications are detailed in a blog authored by Akamai Fellow Erik Nygren, back in 2018 on what​ ​encrypted DNS means for the Internet as a whole​.

ODoH is an extension of DoH, so let’s start with that. The protocol exists because DNS queries are sent in cleartext. This means anyone on the network path between a user’s device and the DNS resolver can see both the query that contains the website the user wants to visit, as well as the IP address that identifies their device. Both protocols are designed to increase user privacy by preventing queries from being intercepted, redirected, or modified between the client and resolver – something known as a middle (MiTM) attack. DoH encrypts communications from the client to its resolver, and ODoH takes this a step further to obscure the client from the resolver.

DoH itself is just a protocol for doing DNS lookups over HTTPS. Most of the contention comes from ways in which DoH resolvers might be discovered and configured. For example, when DoH is used at the application level, it can bypass name servers configured at the OS level. So a web browser can come with a list of DoH compatible resolvers already configured, and traffic from that client would then use those DNS settings.

In 2019 Mozilla ​turned on DoH by default for all Firefox users​, using Cloudflare as the server. This meant that the browser would prefer DoH via Cloudflare. This has been​ ​heavily criticized as an anti-privacy move,​ ​since Mozilla is essentially handing off all DNS resolutions to a single for-profit corporation. ISPs expressed concerns over their ability to perform lawful interception and content filtering (for example, legal requirements or parental controls), and many feel that since Cloudflare is an American company, this could not only be centralizing a large portion of the internet but also making it subject to law enforcement from a single government.

To explain this, switching from using an ISP’s local in-country DNS resolver to a DNS resolver that is out-of-country could make both privacy and performance worse rather than better, regardless of what communications transport is being used for the DNS. Consolidating DNS lookups to a few services also introduces new risks for enabling the correlation of user activity, and these services potentially become highly attractive targets for subpoenas and extra-legal attacks. This is all made substantially more challenging as many users lack a way to judge the level of trust they have with various DNS service providers, making it hard for them to make an informed choice.

So Cloudflare’s ODoH announcement is their way of asserting that they will provide users with an option not to send IP information to their DNS resolvers, which Cloudflare claims will ensure privacy​. ​But this introduces another challenge: that ​ODoH will impact performance because it introduces significant latency, and also strips information that is required to do effective and efficient CDN mapping. Cloudflare’s research​ paper about ODoH​ provides testing conditions that are not relevant to the real-world, and the performance impact numbers mask latency introduced by load times for real web applications.

There are a wide range of other options being explored by the IETF. Given the wide variety of use cases, there is not likely to be a single solution. Many of the leading proposals involve a mixture of sources for secure DNS resolver configuration, device policy, associated and designated resolvers, and user choice. Something like ODoH may fit into this for resolving names where performance may be traded off for possibly improving privacy, but at this point, the concerns and potential pitfalls seem to outweigh the benefits.