Streaming Media
Eschewing the term "P2P" for the friendlier "peer-assisted," Chicago startup Phenix claims it can offer unprecedented scale and unmatched latency
The technology hasn't yet been built to live stream next February's Super Bowl—reliably—to 110 million viewers , should the NFL's usual TV audience all switch online. Or has it? Streaming systems developer Phenix reckons its platform has the unique capability to deliver global, high-quality, synchronous viewing at broadcast scale today.
"We are good to go and at unprecedented in scale," says Dr. Stefan Birrer co-founder and CEO of the Chicago-based startup.
"We dream of streaming the Olympics opening ceremony to a billion people worldwide in real-time, a task that is not possible with other streaming technologies," says Birrer.
In a recent article for Streaming Media, executives from several streaming and CDN companies concluded that the internet, as it currently stands, is incapable of allowing for mass-scale live streaming.
No so, according to Birrer. Phenix claims to be the only company offering a genuine real-time streaming experience to customers at scale, and operating under the mantra that if it's live, it's too late.
While most services are either HTTP Live Streaming (HLS) or WebRTC, both of which trade scale for delay, Phenix's platform offers the total package—high quality, less than 1/2 second of latency, and a potentially infinite number of concurrent users.
Its suite of technologies, built from scratch, is called PCast, which includes algorithms designed to eliminate unnecessary idle server capacity and to provision resources only when demanded. A "Flash Crowd Elasticity" provisions resources "in seconds" to handle large crowds joining popular streams "without interruption to the platform;" and an Interactive Transport Protocol optimizes packet delivery for each unique connection environment (3G, 4G, LTE, Wi-Fi) and available bandwidth.
"Our platform uses a scale-out architecture, where capacity grows linearly with added resources, and load is distributed to nodes with available capacity," explains Birrer. "Our microservices layer is designed to deal with various load levels to prevent overloading components at any time. Each of our 13 points-of-presence (POPs) can handle several million concurrent users, and the system can scale even further using CNAME or anycast IP load balancing."
Birrer adds, "APIs allow our customers to build any type of application involving interactive broadcasting, whether that's one-to-many broadcast, a few-to-many groupcast, or group chat. These APIs are enterprise-ready, WebRTC standard-compliant, and support all mobile devices (iOS and Android), computers (desktop, laptop, tablet), and web browsers."
PCast runs on the company's own cloud servers, which run on the Google Cloud Platform, but Phenix is seeking other vendors so it's not reliant on Google (as good as it is).
On top of this server-assisted CDN-like platform, Phenix creates scale with a Proximity Multicast technology claimed to decrease operating bandwidth costs up to 80%. The company used to be called PhenixP2P, but being aware of peer-to-peer's negative connotations (illegal sharing for one), they prefer the term peer-assisted delivery and have dropped the "P2P" from the name.
Birrer knows a thing or two about P2P. He has an M.S. and Ph.D. in computer science from Northwestern University with a special focus on researching P2P streaming algorithms. Also on the Phenix team is lead scientist Fabian Bustamante, a professor of computer science at Northwestern.
"Using peer-assisted delivery, applications require less bandwidth from the backend and consequently reduce the total cost for delivering content," says Birrer. "Peers utilize nearby peers to offload some or all of the bandwidth demands from the back-end to the network. Our peer-assisted technology scales naturally with the number of participating peers."
Prior to setting up Phenix in 2013, Birrer worked for Chicago software and financial management company SempiTech where he designed and developed a social video chat application for Rabbit Inc and worked on applications for trading and messaging systems—experience that gave him the insight to develop a video streaming technology that would have a latency in the milliseconds.
"The internet's next wave is realtime. Back when we started Phenix, the delay was in the minutes," he says. "Quality of video has improved since, but lag is still a major issue. There needs to be a better way when the younger generation wants everything now, and not a live stream sports experience ruined because friends or neighbours gets the result a minute before you."
In a similar scenario to the Mayweather vs. McGregor Showtime stream, Birrer suggests some issues could have been allayed with a technology such as Phenix's, which is "aware" when the platform is being overloaded.
"If you project for 4 million users and you have 5 million coming on board in a few minutes, the platform must be able to react to this and scale capacity accordingly. Most systems are not aware of the limits. A fundamental approach to our platform is that with advanced monitoring and control algorithms, Proximity Multicast instantly adapts to changing network conditions."
Each POP is certified for 300,000 concurrent streams today, he says. This, though, could readily be scaled to 5 or even 8 million per data center if an event like the Super Bowl should come calling. In that instance, Phenix would set up 30 POPs (in the U.S), each scalable up to 5 million users, he says.
"Because we built our stack from the ground up, we have a very deep knowledge about it and we can do some options that many other can't. You can have two encoders doing the same thing in parallel and exchange one signal for another if one stream fails. The goal is end-to-end redundancy."
It has 10 paying customers, some for two years, including a "billion-dollar Asian company" and another 20 to 30 testing proof of concept. These include companies in social media, news, and esports, as well as a sports broadcaster in Europe. No names were provided.
Phenix raised $3.5 million during its Series A funding round last April, bringing the total raised above $5.5 million since launch.
Coming soon to Pcast is edge insertion of advertising to avoid ad blockers. The company is also looking to expand into international markets, notably China.
"Even though we already deliver a super-fast stream we are still looking to make it faster and cut out a couple hundred milliseconds. We want to build technology that can stream the Olympic opening ceremony globally. That is the type of ambition we have."
No comments:
Post a Comment