Peer-To-peer Architecture Could Save Streaming Video

I am the founder of the company providing the pluginless peer-to-peer video streaming service the article is talking about (www.streamroot.io).

>-- residential/mobile internet service has asymmetric upload/download speeds.

Indeed, but it is not so much of an issue: 1. the average video stream bitrate today is 1.5 Mbps, which is closer to the upstream limits than the downstream limits. 2. Our solution is hybrid, so if you can't get all the data from other peers, you just get the rest from the CDN, so even if someone has a 100 kbps uplink, he still contributes to the swarm. 3. With fiber there are more and more people having 100+ Mbps uplink, and these people can serve dozens of others peers, and this balances the asymmetry partially.

>-- ISPs are suspicious of heavy uploads originating from residential homes and could be flagged as "servers" in violation of ISP's TOS or reclassified as "business" thereby increasing the billing amount.

This must be different depending your country and ISP, but the upside comparing to some full P2P service like Bittorent is that you share video segments only while you are on the streamer webpage (as soon as you close your tab the p2p stops), and also all the communications are encrypted with DTLS provided by WebRTC, so it is not so easy for ISPs to figure out what is exchanged. And finally, we actually help the ISPs by connecting the peers that use the same ISPs, and are on the same sub-networks first, so they have less peering issues.

>-- too much latency for live mega events (World Cup, etc) because of asymmetry mentioned above. (Your neighbor tweets that Spain scored the winning goal before your P2P stream received the last packets showing that it happened.) Latency for bittorrent is ok, but for live megaevents, no.

I think that this where our technology really differentiates with the old peer-to-peer systems : we don't add any additional latency to the stream, but just use the latency already generated by HTTP streaming protocols like HLS or MPEG-DASH. So with or without our system, you will still have a 20-30 seconds latency with the encoder, as it is already the case with all the professional live streams like the ones used during Superbowl or the World Cup.

As for CDNs cooperating to broadcast a live stream, it is not really the spirit of the CDN market right now, instead what happens is that the Broadcaster can contract one or several backup CDNs for their biggest events, and use them in case the primary CDNs breaks down. This is what happened for the Superbowl stream : NBC had Akamai as their primary CDN, and Level3 as a backup. In the end they didn't have to use the backup, but still paid both CDNs for the event !

If its free or just for fun, then yes you can go this route. But people paying for an experience are, in our experience, intolerant of any interrupt or quality drop.

> -- ISPs are suspicious of heavy uploads originating from residential homes and could be flagged as "servers" in violation of ISP's TOS or reclassified as "business tier pricing" thereby increasing the billing amount.

Yes, but ISPs would love to join with p2p to reduce their upstream bills. There are 2 problems currently. First, there is no single protocol or system so maintaining one cache box for each system is crazy (YouTube/Google, Netflix, HBO, Amazon, Hulu... all proprietary). Second, even though there could be a payoff in the long run with savings, there's no immediate payoff for setup.

This could all be solved with a standardized protocol and system. But sadly the big bodies, in particular IEEE, are a bureaucratic mess and filled with industry sponsored members.

Internet is going backwards. Something's gotta give.

This is exactly what we are doing :) We have have a hybrid p2p model that fetches video segments from others peers as long as it can, and switches back to CDN on urgency cases. And our tracker uses geoIp data, ISP and some network topology elements to connect the best candidates together