Stream engine: What goes on behind delivering contents on OTT platforms

Ever thought about what happens at the backend of your over-the-top (OTT) player every time you hit the play button to watch your favourite content whenever, wherever, and on whichever device you want? Don’t fret. Every video-streaming player wants us to believe that the only essential ingredients for binge-watching are great content and our ability to fight sleep. As Ali Hussein, CEO of Eros Now, puts it: “Amid content and marketing, tech is the underdog that ensures a sublime consumer experience.”

For a while now, the OTT arm of production and distribution company Eros International has been working on introducing voice tags at the backend of its content so that users can voice search for their favourite titles in different languages.

Like Eros Now, several streaming platforms, internationally and in India, are investing heavily in technology — almost spending one-fourth of their overall expenditure in it — to enhance user’s binge-watching experience.

“ZEE5 is betting big on the tech that allows viewers to check highlights of the most popular content as they browse the platform. It is collaborating with foreign tech companies that use machine learning to send personalised notifications and recommendations to each user based on their device, internet bandwidth, and content preferences,” says Tarun Katial, CEO of the streaming platform.



MX Player (a part of the Times Group that publishes ET Magazine) has a proprietary technology that allows a user to watch 10 hours worth of content using only a 1 GB data pack. Amazon Prime Video now provides the latest in video and sound tech — 4K viewing and Dolby Atmos — for select titles. Last week, Netflix was in the news — and received criticism from fans and creators alike — for experimenting with a speed-watching software that will allow users to watch videos at 1.5 times the normal speed. However, it is unlikely the billboard ads from these OTT platforms will promote any of these facets as against the latest content they have acquired or produced.

OTT players don’t often talk about the technology behind their platforms outside of tech conferences and hackathons. Because, like most millennial relationships, streaming tech is complicated. “Further, there are no technological benchmarks because the industry is yet to mature,” says Vishnu Mohta, cofounder of Hoichoi, an Indian OTT player that specialises in Bengali content. “Every player is in the catching-up phase.”



Kolkata-based Mohta, 35, had to do some catching up himself when he launched Hoichoi over two years ago. During this period, he realised that streaming tech involves more players than the cast of a typical Indian family drama. Each platform could be working with 30 tech suppliers at any given point.

For starters, all content across OTT platforms is stored in the cloud. The cloud is made up of hundreds of physical data centres spread across the world, rife with cables connecting heavy-duty equipment. These are set up by major tech giants such as Google, Amazon and Microsoft and are widely known as Google Cloud Platform, Amazon Web Services and Azure, respectively.

OTT players rent space in the cloud for storing their files. On AWS, for instance, the basic monthly fee for storing 1,000 GB of content per hour is $125. The rate changes depending on several factors, including the region of cloud storage. Why not store files internally, you ask? “Because OTT players own an ever-expanding content library. Setting up data servers for storage and delivery not only requires huge capex but also a high level of expertise. Why invest in that infrastructure when you can rent as per usage?” says Abhijit Bhide, senior vice-president of engineering at Vuclip, the parent company of global OTT platform Viu.


Source: Dacast, Star TV and respective companies
In fact, OTT firms outsource a lot of functions to specialised tech players even before the content is uploaded to the cloud. Every time a streaming platform creates or acquires content, it uses a third party encoding software to compress the original file that can run in terabytes (TBs) into a streaming-friendly lower file size that retains the visual quality of the content. This compressed file then goes through encryption, which protects the content from unauthorised streaming. Remember that one time you were trying all hacks to watch an episode of Game of Thrones on your streaming player abroad but couldn’t? You would have hit the wall of the player’s encryption code. Among other things, it prevents users from watching content in a region the player doesn’t have the licence to show it.

After encryption comes metadata. As the name suggests, metadata means data about the data on the files. This is stored in the form of a brief description of what the content is about, and all kinds of relevant tags associated with it. One may argue there is little that is technical about metadata but it is crucial for search optimisation. “Many torrent streaming sites have better metadata than OTT platforms, which is why they are able to lure uninformed users to their sites,” says CEO of a streaming company, who wishes to stay anonymous.

It is only after accomplishing these three sacrosanct steps — encoding, encryption and metadata — that the content is uploaded to the cloud. When a user logs into an OTT platform, its in-house or external recommendation engine comes into play. “There are over 400,000 assets any player has in the cloud at a given point of time. The recommendation engine uses machine learning to display from those the items a user is most likely to watch. This is based on the user’s profile and consumption pattern,” says Rubi Boim, founder of Talamoos, an Israel based real-time prediction platform.



Boim works with ZEE5 in India and 10 other streaming platforms in other countries. “We don’t go by tags because then you would only get suggestions for Tom Hanks movies if you’ve watched one recently. We learn from the consumption pattern of all users to apply personalisation to each individual,” he says.

Once the user decides what to watch and hits play, the third and final phase of streaming tech gets into action. This is the part where the video player asks its content delivery network or CDN to fetch the desired content from the cloud. CDN, as Viu’s Bhide describes it, is a network of “intelligent pipes” through which the data flows from the cloud to the end-user.

Much like cloud storage units, CDNs are also data servers except there are hundreds of thousands of them located across the world. They take encoded and encrypted files from the cloud to their server that is closest to the user’s location and eventually onto the video player. Their job is to significantly reduce the time taken for the content to reach the user.

Sidharth Pisharoti, regional vice-president of the media and carrier division of Akamai Technologies, a leading CDN player, finds a suitable analogy for the network’s role in streaming. “We build internet super-highways to take you to your destination without getting stuck in the World Wide Web’s equivalent of the Bengaluru traffic.”



With over 240,000 servers located around the world, Akamai is the CDN partner for many OTT players including Hotstar, Hoichoi and Viu. On most days, the second highest web traffic at the CDN comes from India, the country Akamai claims as one of its fastest-growing markets. In June 2019, Hotstar reported a concurrent viewership of 25.3 million for the live-streaming of India-New Zealand Cricket World Cup semi-final on its platform powered by Akamai’s CDN. Pisharoti explains why that is no mean feat. “On TV, you just pull the content from the satellite and onto your DTH service provider. On OTT, you don’t know whether the user will log in from his mobile or desktop or smart TV or whether they’ll be on the road, in their office, or at home. Changing internet bandwidth basis device and location make the last-mile delivery even more complicated because the size of the driveway you build keeps changing.”

All this happens within milliseconds of a user logging in and hitting the play button. And the process is still far from over. Once the CDN fetches the file, the OTT platform’s video player decodes and decrypts it. It optimises video quality as per the device and user’s internet bandwidth to ensure minimal buffering while streaming. For non-subscribers, the video player also stitches ads to the content at this stage. Even when a user is offline, the OTT engines are at play.

Optimove, an Israeli hyper-targeted marketing company, is one such engine working with Indian and international streaming companies to provide personalised communication to their users. “Personalised communication for OTT is a challenge, particularly in the Indian market where you are targeting over 200 million users who are consuming content in more than 12 languages. But eventually, it will have to be the norm,” says Sam Resnicow, director of strategic services at Optimove.



The complexities of streaming tech have opened up avenues for vendors across major tech hubs. In the last one year, Kaltura, a New York-based video SaaS company, has signed one streaming platform from the media and telecom vertical every six to eight weeks, says Gideon Gilboa, senior vice-president of the company. Kaltura counts OTT player Voot and direct-to-home service provider DishTV among its clientele from India.

Until two years ago, prediction platform Talamoos worked mainly with ecommerce portals and human resource companies, besides a few government agencies. Today, 60% of its revenue comes from OTT platforms. Streaming tech, in all its complex glory, is constantly evolving as the players discover more rooms for improvement.

Recently, Resnicow of Optimove received 12 emails from a streaming company late at night. The subject read: “We Miss You. Come Back.” Next day, the 28-year-old went back to his notepad to write a word of advice for clients: “If the user has a bad experience with your OTT, make sure the messaging acknowledges it and vows never to repeat it.”