Skip to content

What is Flink OTT? 🤔

The Shortest Answer (a.k.a Buzzword-Bingo)

Flink OTT is a WebRTC-based, software-defined, server-side-managed Peer-To-Peer video delivery network that enables end-user devices to share live video data to offload server infrastructure by 75% while providing better QoE and infrastructure reliability.

The Short Answer (Read this if familiar with the topic)

Flink OTT is a software-defined P2P Video Delivery network (CDN) made for the OTT broadcasting industry, based on WebRTC. The software helps large-scale OTT broadcasters to overcome the challenges of scalable and reliable live video delivery. To achieve this goal, Flink builds a distributed delivery network between connected end-user devices that are streaming video content at the same time. This Peer-To-Peer network is then used to share video content between all viewers and build an ad- hoc video delivery network with unlimited scale and capacity. To keep control of this network, Flink comes with a centralized network management service, which dynamically connects devices based on network and bandwidth information. It then instructs peers to deliver data to each other, creating a zero-overhead, zero-delay delivery network on top of any existing (multi) CDN infrastructure.

Flink OTT is used by broadcasting companies around the world to reduce their CDN expenses by 75% on average and 90% at peak times. At the same time, Flink increases user session time due to 20% higher bitrates and 30% less rebuffering. Thus, Flink OTT solves the problem of reliable and scalable infrastructure for live video streaming.

To build streaming infrastructures today, broadcasters have to rely on content delivery networks that work on client-server network architecture. This approach is inefficient, expensive and hardly scalable when it comes to live video streaming. Flink OTT extends the existing client-server network with a user-to-user network to offload the streaming servers and solidify the entire infrastructure.

The Full Answer (For those who came here just by accident)

Flink OTT - Intro 01

Spoiler: If you know this situation, you will probably like Flink OTT. But let's start with the actual problem.

The Media Broadcasting Industry

The Media Broadcasting Industry is a globally spread community of television channels, content owners, network providers, agencies, consulting firms, technology providers (video processing, content delivery, end-user-experience or “Quality of Experience”, Security, app development), End-To-End Applications, production companies and online video platforms (OVPs). The entire industry is highly consumer-driven and has a total estimated market value of 50 billion USD. While this estimation was done in 2017, current statistics forecast a growth rate of 167% to a total evaluation of 83.4 billion USD. In comparison, the same evaluation for 2010 resulted in market value of 6.1 billion USD. This shows a previous market growth of 820% between 2010 and 2017.

An introduction to how video can be consumed

To consume broadcast video data, end consumers have to decide between four possible distribution channels. These are Sattelite, Cable, IPTV, and OTT. Sattelite and Cable are, until now, the most common and best-established channels. Also, they are typically available in most households and due to their longtime existence, they are known to most people. IPTV was the first approach to stream video data over an IP based network. The first services launched in the mid-2000s and strongly evolved into a strongly supported market section. Players like Deutsche Telekom (TV Entertain), Vodafone and other network owners or Internet Service Providers (ISPs or Telcos) use their wide-spread network infrastructure to distribute video data via their own networks. However, to provide an IPTV service, a company has to either own or rent a dedicated network in which content can be delivered to end consumers. A newer approach (and the one with the strongest growth) is using the public internet as media distribution channels. This “over-the-top” (short “OTT”) delivery is used by what most people know as “streaming platforms” like Netflix, Amazon, YouTube, etc. These Online Video Platforms (OVPs) are not relying on a dedicated network that subscribers have to use to consume content. As soon as a subscriber has public internet access, streaming media is possible.

  • Satellite TV. Television from providers like DirecTV delivered via radio waves. Cable TV: Television from providers like Time Warner Cable, delivered via a coaxial cable connection.

  • IPTV (Internet Protocol Television). Television from providers like Prism TV, delivered over the Internet via a privately-managed network.

  • OTT Streaming: Television from third-party services like Netflix and YouTube, delivered over the open Internet.

OTT as an emerging market

The immense success of the latest OTT applications has drawn a lot of attention from traditional media companies (television channels, broadcasting companies, news channels, movie producers). Today, the OTT industry has become a strong market section of its own as is often explained as “the future of television” and “a new entertainment era”.

The success can mostly be derived from the pure fact, that companies and consumers are no longer bound to traditional watching habits and get much more degrees of freedom at both ends. End users can decide when they want to stream their favorite content while content providers can use the full power of web tracking and internet advertisement to maximize personalization and advertisement. Without the need of a physical network connection or a dedicated IPTV device, content can also be accessed from anywhere anytime. With the two main monetization models AVOD (advertisement-supported VOD) and SVOD (subscription-based VOD), end-users get an easy way to opt-in for these new, much more flexible media entertainment platforms. Also, for a much lower price.

With a new approach, there come new problems

Despite the overwhelming success and pros of choosing OTT as main media distribution channels, companies are also facing some major challenges on this path. To understand why OTT brings a lot of new problems into the typical broadcasting process, one has to get an understanding of how content is delivered to end consumers for each distribution method.

In a nutshell, all 3 methods Cable, Satellite, and IPTV make it “easy” to deliver content to subscribers while OTT is making it very hard. This becomes much clearer when thinking about how content is delivered.

Media delivery via satellite (also often referred to as “Over-Air”) is a signal that can be accessed by every subscriber within signal reach. Media delivery via cable is based on an existing, dedicated cable network where every subscriber has her analog connection to the network. Media delivery via IPTV is similar to cable delivery with the main difference that content is purely digital and the main distribution protocol is IP.

What all these methods have in common is that their performance does not depend on the number of subscribers within the network. All three methods provide a “broadcast from one point to everyone who is listening” method. This method of posting information into a network is also called “Multicast” (one-to-all). Once sent, the network itself manages the distribution of a message to all its users. This is where OTT takes a step in the opposite direction. Within the public internet, Multicasting is impossible (if not, every user would be able to easily send messages to every other user, completely flooding and congesting the network). The internet is based on “Unicast” communication (one-to-one). It has only limited disadvantages, but the major ones are the cost and complexity in dealing with large numbers of concurrent users. For the TV industry, which is used to broadcast scale viewers, this can be an issue.

A unicast TV channel creates a single, unique stream linking the server to an end viewers’ set-top box. So 100k concurrent viewers of a live stream will require 100k unique, concurrent IP streams. The key thing in terms of cost and scalability is that there are few economies of scale. Each viewer of a unicast channel takes up additional bandwidth. For example, if you have only one viewer playing a 100-kilobits per second (Kbps) stream, only 100 Kbps is being used. If you have 10 viewers all playing 100Kbps channel, then those viewers as a group are taking up 1,000 Kbps. The problem is that TV channels tend to like big audiences and a reasonably small linear channel can challenge the upper limits of unicast.

The three main problems when using OTT - Cost, Quality & Scale

  • Cost. To cope with the tremendous amount of bandwidth and infrastructure required for broadcasters to establish Unicast connections to their subscribers, they use so-called Content Delivery Networks (CDNs). CDN providers own huge network infrastructure, data centers and other Points of Presence (PoPs). With a lot of infrastructures but without own content, CDNs become the missing piece in the OTT broadcasting puzzle for broadcasters. However, CDNs price based on traffic delivered through their network. This transforms the scaling problem of connectivity to an equally difficult scaling problem of cost.

  • Quality. The internet is a place of “best effort”. By design, it does not provide any guarantee for quality, delay or even successful delivery of any data. Coming from controlled networks and over-air transmission, OTT seems highly unstable and unreliable (which it ist) to deliver large data in real-time. It would also be a shock for TV viewers having to step back from the usual HD / 4K TV signal to stuttering, quality-switching SD signal that arrives 20 seconds later than any usual TV. Nearly every research shows that keeping the level of video quality as close to TV-grade quality is crucial for an OTT service. Viewers expect the same level of quality for video watched online as they do with traditional pay-TV services like satellite or cable. While CDNs can help with data delivery itself, they still rely on the best-effort based internet structure which does not allow Multicast to work.

  • Scale. Modern Content Delivery Networks are serving huge amounts of traffic through the internet. Akamai, the world’s leading CDN provider server up to 30% of the internet’s traffic alone. However, while a CDN’s growth relies on the expansion of physical infrastructure, it can only hardly keep up with the current growth of OTT subscriptions. Also, CDNs can not help when it comes to the expansion of mobile networks and countries not as far developed as Europe or the Northern US. Central America, Africa and parts of South America can barely provide the necessary resources today and will face great difficulties when expansion and quick growth are required.

As initially described in the Introduction, there are three main problems:

  • Cost
  • Quality
  • Scale

All these problems have their origin in switching from a Multicast distribution type to Unicast distribution.

In a nutshell, what Strive’s technology “Flink” does is making Multicast possible inside an OTT network. Flink uses end devices that stream data via unicast from a content server and builds direct unicast connections to other devices streaming the same data. Flink then uses these direct user-to-user connections (also known as Peer-To-Peer connections) to exchange data. The way how end users are connecting to each other and what amount of data is exchanged via these connections is monitored and proactively managed by a centralized backend server (the “Strive Network Manager”). Thus, Flink extends the traditional OTT unicast network with an overlaying Peer-To-Peer Delivery Network (short P2PDN) which can take care of up to 90% of the entire network traffic.

  • How To Solve Cost. Since the amount of traffic a content provider produces within a content delivery network is equal to the amount of money to be paid to that CDN, reducing CDN traffic is crucial to reduce video delivery costs. The pricing model of CDNs is either usage-based or commitment-based on the amount of traffic, measured in Terabyte (or sometimes Gigabyte). Typically, a provider would estimate its traffic for the next year and negotiate a “Cent per Gigabyte” pricing with the CDN provider. With a pricing average of 0.0386 USD per GB (more in “Sample Pricing Calculation”) in Europe and Northern America (APAC, Middle East, LATAM and Africa are much more expensive), there is a direct connection between CDN traffic and CDN cost. With Strive, up to 90% of this cost can be completely removed by shifting a proper amount of traffic inside the P2PDN instead of the CDN. Especially at peak times (for example an important football match, the Superbowl or a live concert), the number of viewers often reaches unexpected heights. The fact that the resulting traffic is very hard to predict is crucial for every CDNs business model. Strive, on the other hand, does not charge for traffic. This disruptive decision is a key argument for using Strive from a broadcaster’s perspective. Instead of traffic, Strive charges for the maximum concurrent users at any given time. While this number is still scaling quite fast, it is much more predictable than the amount of traffic delivered. Also, traffic depends on the delivered video quality. Even today, SD streaming is a common standard in video streaming because 1080p and 4K would double or quadruple the amount of traffic while the number of concurrent users does not change. In conclusion, Strive’s business model is still scalable (and profitable) and also solves cost-efficiency from a broadcaster’s perspective.

  • How To Solve Quality. Good or bad quality of experience, or QoE (an expression for the amount of customer satisfaction with the streaming experience and general video quality), has multiple factors of which one is the last-mile delivery. Both OTT business models, subscription-based or advertisement-based, are strongly dependant on customer satisfaction and retention. At the same time, studies show that consumers become less patient every year. For example, 43% of global OTT users name rebuffering (stalling video signal and the infamous “spinner on the screen”) as their main frustration issue and 49% of all users will stop a stream after just one incident. As soon as users stop watching a stream, ad revenue decreases. As soon as they choose another provider, subscription revenue decreases. These numbers are consistent with the current trend in the broadcasting industry of reducing latency to increase customer satisfaction. While Strive is not reducing latency in video streaming, it strongly optimizes last-mile delivery. As stated before, this is also an important factor in the “QoE equation”. By providing more than just one edge server that users can stream content from, Strive provides up to 10 additional sources (other end-devices) for each user. This highly increases the probability of finding a faster and/or better connection for a single device, thus increasing the average quality of experience.

  • How To Solve Scale. The efficiency of the P2PDN strongly depends on the region, the number of concurrent users, their density within the network and available bandwidth within the last mile (thus also on the network topology). The Network Manager ensures to monitor the available device-to-server bandwidth and continuously compare it to the available device-to-device bandwidth in real-time. Based on this information, the Network Manager will decide whether not to stream content from the OTT network or the P2PDN. This process ensures a perfect optimization policy, never harming the potential of the OTT network. The case scenario when using Flink is the best case scenario when not using Flink, thus providing a 100% risk-free usage of Flink within an OTT network.

Since the Network Manager keeps an open connection to each end-device, it ensures maximum flow control within the P2PDN. The software for the Network Manager is written in the modern, backend-focused programming language Golang. This language is designed and developed by Google and is one of the most growing programming languages from 2016 until today. We use Golang to build the Strive Network Manager and manage over 100,000 concurrent users with a single machine. Since the Network Manager software is easily scalable in a horizontal way, managing even a million concurrent users is not a problem for the Strive Network Manager.

Summary

Well, this was quite an introduction. Having read all the above, you might almost call yourself an industry expert 😉. Well, maybe not. But you should at least now have a better understanding of why Flink OTT is a crucial piece in the puzzle of the next generation of video delivery services.

If from this point you have any open questions, please feel free to contact us anytime via email or through our website.