3 ms·

This could be huge. I note however that Peer to Peer is not mainstream. It either never was, or it ceased to be. Bittorent is popular, but it's not mainstrea

This could be huge. I note however that Peer to Peer is not mainstream. It either never was, or it ceased to be. Bittorent is popular, but it's not mainstream. If it was, YouTube would distribute torrents of it's videos (it could be done right now by writing a torrent client in JavaScript). The sad reality is, there is a long lasting, major roadblock: Asymmetric bandwidth.

glesica 13y ago
Which is one reason that Google Fiber is so interesting to me. People, even technical people, tend to see it as overkill, but a fat, symmetrical pipe into every house could be a huge game-changer.
vectorjohn 13y ago
I don't think that's necessarily a major roadblock. Depending on implementation, with peer to peer you may connect to many peers, and their combined bandwidth will be good. Another way of looking at it: if I used 100% of my upload traffic continuously, it would be a lot more total data sent than I generally download with normal use. So as long as a peer to peer system ensures QoS, the bandwidth is there. I think.
- loup-vaillant 13y ago
  Nope. On a peer to peer network, download equals upload. If you arrive late, sure, everyone will upload to you. But if you arrive in the middle of the viral explosion, you won't be the only one who wants to download. Peers will upload to each other, not just you. So, unless one of the peers happen to be a big fat server with monstrous upload, you won't download the video much faster than the median upload rate. The idea of using your uplink to the fullest sounds good, but to do that, you need to acquire some content in the first place. And that requires the upload bandwidth of other peers…
wmf 13y ago
There's a tradeoff between the extra complexity of engineering a P2P system and the ratio of marginal cost of resources between home and data centers. In 2001 when people were going gaga for P2P it was actually a good idea compared to the status quo at the time (remember that Web 2.0 hadn't been invented yet). Since then data center storage and bandwidth has gotten exponentially cheaper but broadband has stagnated, so client-server is now cheaper than P2P. For example, before they were driven out of business, cyberlockers delivered better service and had a more sustainable business model than BitTorrent. And that's not even taking into account the network effects of "big data".

glesica 13y ago

Which is one reason that Google Fiber is so interesting to me. People, even technical people, tend to see it as overkill, but a fat, symmetrical pipe into every house could be a huge game-changer.

vectorjohn 13y ago

I don't think that's necessarily a major roadblock. Depending on implementation, with peer to peer you may connect to many peers, and their combined bandwidth will be good. Another way of looking at it: if I used 100% of my upload traffic continuously, it would be a lot more total data sent than I generally download with normal use. So as long as a peer to peer system ensures QoS, the bandwidth is there. I think.

loup-vaillant 13y ago
Nope. On a peer to peer network, download equals upload. If you arrive late, sure, everyone will upload to you. But if you arrive in the middle of the viral explosion, you won't be the only one who wants to download. Peers will upload to each other, not just you. So, unless one of the peers happen to be a big fat server with monstrous upload, you won't download the video much faster than the median upload rate. The idea of using your uplink to the fullest sounds good, but to do that, you need to acquire some content in the first place. And that requires the upload bandwidth of other peers…

wmf 13y ago

There's a tradeoff between the extra complexity of engineering a P2P system and the ratio of marginal cost of resources between home and data centers. In 2001 when people were going gaga for P2P it was actually a good idea compared to the status quo at the time (remember that Web 2.0 hadn't been invented yet). Since then data center storage and bandwidth has gotten exponentially cheaper but broadband has stagnated, so client-server is now cheaper than P2P. For example, before they were driven out of business, cyberlockers delivered better service and had a more sustainable business model than BitTorrent. And that's not even taking into account the network effects of "big data".