The online video streaming revolution has hugely increased the use of bandwidth by individual consumers. At the same time it’s also resulting in huge bandwidth bills for streaming sites such as YouTube.
Thus far the demand for video continues to grow, and it is even expanding to live video. To keep video services from collapsing and to save bandwidth costs, it seems almost inevitable that content providers will have to look at P2P-based streaming solutions. Last year we reported that CNN had experimented with a P2P-based live stream, and the Tribler research team has already shown that it’s possible to use BitTorrent to stream live footage.
There are currently a few dozen people working on P2P-based live streaming, and they are soon to be joined by Bram Cohen, the inventor of BitTorrent. Last week he tweeted that he will beat Tribler’s solution in terms of delay. “Tribler’s live streaming benchmarks are a joke. I’m going for < 5sec delay," Bram wrote.
This comment did of course peak our interest, so we decided to get in touch with Bram Cohen to ask him what he’s up to exactly. He told us that his BitTorrent-powered live streaming implementation is still in an early stage of development, but he hopes to have a working version ready “sometime next year”.
“I think there’s a very large market for live [streaming] in general, and to date noone has proven that a p2p solution can meet the real-world requirements for being an acceptable live solution. I intend on changing that,” Bram told us.
There are still a lot of problems to solve though, before the first version becomes available to the public. Getting BitTorrent to work effectively with live streams requires several major adjustments.
“Doing live properly is a hard problem, and while I could have a working thing relatively quickly, I’m doing everything the ‘right’ way,” Bram told TorrentFreak. He further explained that everything has to be redone in order to make BitTorrent compatible with live streams, “including ditching TCP and using congestion control algorithms different from the ones we’ve made for UTP,” Bram said.
“I am fundamentally a technologist, and am simply not interested in working on something which doesn’t solve the fundamental problem it’s supposed to tackle, especially in a market where there have already been so many bad technologies which failed to succeed based on sales and marketing,” he added
In his tweet Bram Cohen focused on an extremely low latency of less than 5 seconds, so content will not have to buffer for minutes before the stream starts. According to some, such a low latency could mean that a lot of potential upload capacity would go lost. However, Bram disagrees on this, as he explained to TorrentFreak.
“Lower latency doesn’t require extra bandwidth, it just requires that everything be designed from the ground up for low latency. In terms of overhead, I’m shooting for making a swarm able to work with only 20% extra upload capacity, which is subtly different from having 20% extra overhead – because there’s noise in real networks, there needs to be some slop for when things get bad.”
“My actual extra bandwidth used will be less than 10%. This very important benchmark number is generally speaking not even mentioned for most p2p live streaming solutions, and I get the feeling that the developers don’t even know what the value is. I’m taking an approach of viewing all the important benchmarks (latency, extra bandwidth necessary, offload percentage) as central to the whole thing, and running realistic simulations constantly to get a good idea for what they are and help optimize them.”
“Oddly, most live p2p solutions don’t even make coherent claims as to what latency they can provide, and when they do it’s a delay which hardly qualifies as live. My offload of course goes over 99% on large swarms – without that it’s hardly p2p,” Bram said.
The big question is of course how BitTorrent’s inventor will try to solve this puzzle. Many researchers including the Tribler team are looking into P2P-powered live streams, and not all of them agree that the tit-for-tat algorithm based on reciprocity is suited for live streams. However, when we asked Bram whether he is looking into a new algorithm he was very clear.
“No, the low latency requirements basically preclude any competitive algorithms and I’m going with a cooperative approach. It does do a very good job of squeezing out every little bit of upload capacity all the peers have though, and doing it with the same ISP-friendly properties as UTP,” Bram said.
Quite a few ISPs are complaining bitterly about the strain BitTorrent puts on their network, so they will be delighted to hear that they are not being ignored in the development process. If done right, BitTorrent-powered live streams may accelerate the availability of live streams on the Internet.
Not only will existing broadcasters be able to stream their content at low costs, individual users will also be able to stream a live feed to tens of thousands of Internet users from their home connection without having to invest in bandwidth. More than ever the public will be in charge of distribution, while BitTorrent-powered TV moves one step closer to becoming reality.
