Livepeer provides a fascinating example of how data centre infrastructure that lays outside traditional CDNs, traditional cloud companies and telcos is organized and orchestrated to provide a streaming network for use by smaller media owners looking to distribute video at lower costs. Aimed at video providers targeting less than 100,000 concurrent users for live events, the Livepeer network is an alternative to CDNs at this scale of operation and will result in average savings of 75% vs traditional CDNs, according to Eric Tang, Founder at Livepeer.
Livepeer customers include companies distributing local, school or college sports, music events, religious events, educational lectures and more. While valuable to those watching it, the content may not be what the industry terms ‘premium’ and monetization potential reflects this. Without very cost-effective distribution, the economics for this TV may not work.
Some Livepeer customers will use the Livepeer network for 90% of their event needs and occasionally break the 100,000-user barrier, at which point a traditional CDN will be used, too. Livepeer provides an ‘as a service’ offering, so can take care of the handovers when this happens. There are also customers who may use traditional CDNs for premium sport content but Livepeer for their non-premium sport.
The key innovation is that while Livepeer owns some data centre capacity of its own, the network is built primarily upon compute and storage infrastructure owned by others, with Livepeer organizing this capacity to achieve the tasks required. The resulting decentralized network can be used for media processing tasks like transcoding as well as transport.
Media owners seeking video processing and distribution services pay Livepeer, which then pays the third-party infrastructure providers based on their contribution to each job. There is a sophisticated marketplace, backed by tokens acting as virtual money, where the infrastructure providers bid for work and pay Livepeer for the right to perform tasks. They are then paid by Livepeer at a higher rate for the completed work, thus providing the infrastructure partner with their profit. Most of the network partners are professional data centre operators.
The Livepeer orchestration software knows the price ceiling for a given task and narrows the list of possible infrastructure providers down to those who can compete below that ceiling. Then the algorithm assigns tasks to the best performing nodes among these suppliers – providing an incentive for each network contributor to maximize efficiency.
In its marketing, Livepeer says its decentralized architecture “combats exorbitant vendor fees and ensures efficiency, scalability and innovation.” The company says it is helping streaming companies experiment and then build and scale innovative streaming products at significantly lower financial risk.
“Meanwhile, established streaming companies that want to experiment with new engagement strategies, user-generated content or gaming services can trial and launch offers on-the-fly without incurring heavy video delivery costs.”
Livepeer provides a SaaS solution to customers covering the end-to-end requirements of live streaming TV with all the features a media owner would expect, from redundancy to detailed analytics (covering, for example, the number of concurrent viewers, total viewership minutes, buffering rates, time to first frame, etc., all of which can be broken down by media owner, device and OS, etc.).
Network security is backed with encryption (so content is safe from any network snooping, for example). Some of the infrastructure providers participate in the Livepeer network on a pseudo anonymous basis but others are known entities (including Livepeer itself, acting as a data centre operator). If a media owner has exceptional security concerns they can use only the known providers.
According to Tang, this fully transparent subset approach would increase costs by 30-40% but still leave a media owner with a 50% saving versus a traditional CDN. This figure includes premium SaaS features provided by Livepeer.
While Livepeer is mainly concerned with live streaming, the company does support live-to-VOD and pre-recorded VOD, including via partners. It lists support for social streaming, reality entertainment, live music, gaming and enterprise video, among other applications.
The Livepeer solution uses standard streaming, so there are no changes to how video enters or exits the network compared to a standard CDN. No changes are needed at the receive device and app.
Tang emphasizes that the Livepeer network is not designed as a replacement for high-scale CDNs, and the company does not compete in that space. In the market for lower-cost distribution of less-premium live content, his company’s network model provides a way to bring additional capacity to streaming. “We have created an efficient market where more people can participate, harnessing the democratization of infrastructure services.”
One Livepeer customer has used the network to stream DJ concerts from Time Square in New York. Another – God Behind Bars – streams religious content to a one-million strong prison population. Customers also include Fishtank, a 24/7 live streamed reality show that recently surpassed $3 million in revenues. At the more premium end, streaming service Myco uses Livepeer to deliver live cricket and Premier League football in Pakistan.
Livepeer talks about its network as open-sourced video infrastructure technology that media owners can build upon, modify and integrate into their services. There is a substantial focus towards building an open marketplace of AI video tools on the network – allowing developers and video engineers to start building and experimenting with AI video applications – including for real-time video.
Livepeer claims significant cost savings for real-time AI video tools running on the Livepeer network vs traditional cloud architectures. “The goal is that any content creator or streamer could plug in to the Livepeer network to easily and affordably use any mix of AI video tools. This could be anything from text-to-video generation, captioning, real-time animation or graphical data overlays in live sport.”
The company has been inviting its existing community and software developers and video engineers to build and use practical use-case driven AI workflows on the network.
Livepeer says it has raised over $50 million in funding since 2016 across four investment rounds.
