Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

csc.kth.se

Post reply on HN
Print view View original

1–10 of 13 posts

Re: Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

#3

Post by sp332 »

xtacy wrote: post #2

Are ISPs okay with organizations using P2P to offload their bandwidth requirements to clients? http://arstechnica.com/web/news/2009/02/cnn-p2p-video-stream...

I'm not sure I understand - the data has to come from somewhere, and go somewhere. If both endpoints are on the same last-mile ISP, the ISP wins because there's less traffic to and from peers, so their peering costs are lower. Other than that, why would it matter if the data came from a central server or from some other PC?

Re: Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

#4

Post by xtacy »

sp332 wrote: post #3
xtacy wrote: post #2

Are ISPs okay with organizations using P2P to offload their bandwidth requirements to clients? http://arstechnica.com/web/news/2009/02/cnn-p2p-video-stream...

I'm not sure I understand - the data has to come from somewhere, and go somewhere. If both endpoints are on the same last-mile ISP, the ISP wins because there's less traffic to and from peers, so their peering costs are lower. Other than that, why would it matter if the data came from a central server or from some other PC?

I agree that if the end points are inside the ISP, it's good.

But peering is different from transit; peering is a delicate relationship where ISPs route traffic to each other for "free", as long as the demands are not too asymmetric. If this software results in a lot of traffic generated from a peering ISP, then it could be causes of worry for the receiving ISP.

Re: Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

#5

Post by sp332 »

xtacy wrote: post #4
sp332 wrote: post #3

Earlier quoted context omitted.

I'm not sure I understand - the data has to come from somewhere, and go somewhere. If both endpoints are on the same last-mile ISP, the ISP wins because there's less traffic to and from peers, so their peering costs are lower. Other than that, why would it matter if the data came from a central server or from some other PC?

I agree that if the end points are inside the ISP, it's good. But peering is different from transit; peering is a delicate relationship where ISPs route traffic to each other for "free", as long as the demands are not too asymmetric. If this software results in a lot of traffic generated from a peering ISP, then it could be causes of worry for the receiving ISP.

But the "receiving" ISP is only receiving the data because its users asked for it. It doesn't make a difference if the data comes from a Spotify server or from a PC. It's still the same amount of data being transferred.

Re: Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

#6

Post by guard-of-terra »

xtacy wrote: post #2

Are ISPs okay with organizations using P2P to offload their bandwidth requirements to clients? http://arstechnica.com/web/news/2009/02/cnn-p2p-video-stream...

How are ISPs the part of equation?

I mean, they would probably not okay with torrents and youtube, but it's not like anybody asked them for consent. They have to sell a service to clients, and clients want torrents and youtube and spotify, so they either deliver or get off market.

Of course, in areas where ISPs are either state-controlled or monopolistic, situation might be different. In this case, pity to the people stuck with monopolist.

Re: Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

#7

Post by nodata »

xtacy wrote: post #2

Are ISPs okay with organizations using P2P to offload their bandwidth requirements to clients? http://arstechnica.com/web/news/2009/02/cnn-p2p-video-stream...

What an odd question. ISPs sell a link to their customers, they don't need to be okay with it.

Re: Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

#8

Post by borism »

nodata wrote: post #7
xtacy wrote: post #2

Are ISPs okay with organizations using P2P to offload their bandwidth requirements to clients? http://arstechnica.com/web/news/2009/02/cnn-p2p-video-stream...

What an odd question. ISPs sell a link to their customers, they don't need to be okay with it.

the "link" (guaranteed amount of bandwidth) they sell to you as a home customer ends with the next hop.

Re: Spotify – Large Scale, Low Latency, P2P Music-on-Demand Streaming [pdf]

#9

Post by wmf »

sp332 wrote: post #5
xtacy wrote: post #4

Earlier quoted context omitted.

I agree that if the end points are inside the ISP, it's good. But peering is different from transit; peering is a delicate relationship where ISPs route traffic to each other for "free", as long as the demands are not too asymmetric. If this software results in a lot of traffic generated from a peering ISP, then it could be causes of worry for the receiving ISP.

But the "receiving" ISP is only receiving the data because its users asked for it. It doesn't make a difference if the data comes from a Spotify server or from a PC. It's still the same amount of data being transferred.

Receiving data over a transit connection may be more expensive than from a CDN. In DOCSIS networks, P2P also tends to introduce last-mile upstream congestion which is expensive to fix.
Post reply on HN