4 ms·

A serious point here: warehouse scale computing beats p2p in a dsl (or even feasible fttp) environments. Essentially the model of the internet we have is broadc

by sgt101 12y ago

A serious point here: warehouse scale computing beats p2p in a dsl (or even feasible fttp) environments. Essentially the model of the internet we have is broadcast with decoration from the upstream channel. To change this services using upstream that make money for the pipe providers are required. Otherwise, no upstream. Freeloading cannot work. Repeat CANNOT. Because you need fellas with highschool educations to look after the fibre and the boxes. Downstream will remove the boxes and put them into warehouse scale facilities. So - Haxxors, get yourselves to the internet of things if you wish to see democracy preserved. Freedom is in peril. Defend it with all your might.

zAy0LfpBZLC8mAC 12y ago
Why do you think that for FTTP? I do see the problem with ADSL as it is currently implemented (as analyzed very well by kragen in the mail linked to in a sibling post to this), but why would the same problem apply to FTTP? Also, I don't get how you get to freeloading - upstream bandwidth is as much part of the product that the customer is paying for as the downstream bandwidth is.
- sgt101 12y ago
  Downstream bandwidth is mostly delivered using caches. Upstream bandwidth is not delivered using caches. So to provide the backhaul for upstream bandwidth we will do deep reach into the datacentre <-> home and interconnect in the data centre. Therefore fttp will not provide p2p connectivity unless there is a need for local connectivity because deepreach and datacentre networking is rendered non viable by sheer traffic volume. But the traffic will need to be geographically localised and/or the services will have to provide the economic support for backhaul uplift. I think we can probably do petabits on single mode fibre bundles now with the right boxes but the trick is the right boxes, which are spensive and take lots of electricity to run. So even if you can pony up for an FTTP roll out then the use cases are going to be constrained by the network architecture that is used to support it. Now, for most people this is not a problem as the up channel is basically irrelevant in terms of bandwidth in the current consumer internet, but if you are the sort of person who imagines a decentralised and democratic consumer internet which is not mediated by massive supernational companies, or if you are the sort of person who is interested in where the value is captured in the industry value chain I think this does matter. Sharing other peoples content will not fund a democratic internet even if governments fund and regulate fttp. We need a new class of applications that generate revenue sufficient to pay for the infrastructure that will facilitate them. If we get that then the infrastructure will also facilitate a shift away from the data centre (not it's elimination though as it will be the way that the current usecases are delivered) If you have any ideas please write them up - before the industry dies!
  - zAy0LfpBZLC8mAC 12y ago
    I'm sorry, but I still don't get what you are trying to say. For one, caches are not exactly located on customer premises, either, right? They tend to be located in places that have easy access to power and are well-connected with fibres, and so far transmission speeds seem to be mostly going up. Also, well, yes, social networks tend to be geographically localised, so I would very much expect p2p application traffic to also be geographically localized!? Widely consumed content that's distributed in a p2p fashion obviously is just as amenable to caching as widely consumed content that's distributed by a central service, if it's distributed through some kind of content-addressed network. If that's cheaper for the ISP, they could just put caching nodes into their data centers. Finally, I completely don't get why you think my communication should "generate revenue". My telephone calls don't generate any revenue either, do they? I simply pay someone for moving my bits around, that should be sufficient motivation for them to take care of moving my bits around.
    - sgt101 12y ago
      I think that caches work where you have a vast imbalance of requests for a particular bit of content - they seem to me to define centralization. In terms of generate revenue - I mean provide a service that people are willing to pay for at a level that will fund a network that enables the use case and decentralization. Current services are tending to deep reach from the datacentre, because that's cheap. If all services that people are willing to pay for are suited to that model that's what everyone will get. To have a different infrastructure model you need a service that people will pay for that is not well served by deepreach from the datacentre. It's not the moving of the bits that is at issue, it's the how the bits are moved, where and who by.
      - zAy0LfpBZLC8mAC 12y ago
        "I think that caches work where you have a vast imbalance of requests for a particular bit of content - they seem to me to define centralization." Hu? Caches for geographical distribution are actually a form of decentralization! The problem with the existing ones is that they are mostly under centralized control, but there is no technical reason why that has to be, and an ISP's caching web proxy (which weren't that uncommon back in the day) is a simple example of caching infrastructure under decentralized control that could help an ISP reducing the load on their upstream connection for often-requested content. You don't need youtube to copy a video from your machine to the cache at some viewer's ISP, that's just the way it is commonly done right now because in that way youtube can make money - or, if you want to decentralize further, you could copy to a cache on the machine of some customer of said ISP in some geographic region: The more requests there are for some content, the more likely it is that there is a copy in your neighbourhood that doesn't need to be transferred from the original source.
kragen 12y ago
You may be interested that one week before, I wrote a post exploring your point (you are correct!) in detail: https://www.mail-archive.com/kragen-tol@canonical.org/msg00265.html https://www.mail-archive.com/kragen-tol@canonical.org/msg002... Summarized as, "Peer-to-peer overlay networks are inefficient on ADSL networks. ADSL networks are almost twice as efficient as SDSL networks. Better alternatives require redesigning the physical layer."
fidotron 12y ago
Actually the game is lost already because cellular modems are being embedded in products directly, skipping any ability you have to firewall them, all in the name of convenience. The fact is we rely too much on the cloud to connect two adjacent devices anyway. The whole notion that much of this data has to go upstream at all is the problem. (Shameless plug) This is what led me to write this: http://montrealrampage.com/king-ludd-19-manifesto-for-a-friendly-home/ http://montrealrampage.com/king-ludd-19-manifesto-for-a-frie...

by sgt101 12y ago

zAy0LfpBZLC8mAC 12y ago

Why do you think that for FTTP? I do see the problem with ADSL as it is currently implemented (as analyzed very well by kragen in the mail linked to in a sibling post to this), but why would the same problem apply to FTTP? Also, I don't get how you get to freeloading - upstream bandwidth is as much part of the product that the customer is paying for as the downstream bandwidth is.

sgt101 12y ago
Downstream bandwidth is mostly delivered using caches. Upstream bandwidth is not delivered using caches. So to provide the backhaul for upstream bandwidth we will do deep reach into the datacentre <-> home and interconnect in the data centre. Therefore fttp will not provide p2p connectivity unless there is a need for local connectivity because deepreach and datacentre networking is rendered non viable by sheer traffic volume. But the traffic will need to be geographically localised and/or the services will have to provide the economic support for backhaul uplift. I think we can probably do petabits on single mode fibre bundles now with the right boxes but the trick is the right boxes, which are spensive and take lots of electricity to run. So even if you can pony up for an FTTP roll out then the use cases are going to be constrained by the network architecture that is used to support it. Now, for most people this is not a problem as the up channel is basically irrelevant in terms of bandwidth in the current consumer internet, but if you are the sort of person who imagines a decentralised and democratic consumer internet which is not mediated by massive supernational companies, or if you are the sort of person who is interested in where the value is captured in the industry value chain I think this does matter. Sharing other peoples content will not fund a democratic internet even if governments fund and regulate fttp. We need a new class of applications that generate revenue sufficient to pay for the infrastructure that will facilitate them. If we get that then the infrastructure will also facilitate a shift away from the data centre (not it's elimination though as it will be the way that the current usecases are delivered) If you have any ideas please write them up - before the industry dies!
- zAy0LfpBZLC8mAC 12y ago
  I'm sorry, but I still don't get what you are trying to say. For one, caches are not exactly located on customer premises, either, right? They tend to be located in places that have easy access to power and are well-connected with fibres, and so far transmission speeds seem to be mostly going up. Also, well, yes, social networks tend to be geographically localised, so I would very much expect p2p application traffic to also be geographically localized!? Widely consumed content that's distributed in a p2p fashion obviously is just as amenable to caching as widely consumed content that's distributed by a central service, if it's distributed through some kind of content-addressed network. If that's cheaper for the ISP, they could just put caching nodes into their data centers. Finally, I completely don't get why you think my communication should "generate revenue". My telephone calls don't generate any revenue either, do they? I simply pay someone for moving my bits around, that should be sufficient motivation for them to take care of moving my bits around.
  - sgt101 12y ago
    I think that caches work where you have a vast imbalance of requests for a particular bit of content - they seem to me to define centralization. In terms of generate revenue - I mean provide a service that people are willing to pay for at a level that will fund a network that enables the use case and decentralization. Current services are tending to deep reach from the datacentre, because that's cheap. If all services that people are willing to pay for are suited to that model that's what everyone will get. To have a different infrastructure model you need a service that people will pay for that is not well served by deepreach from the datacentre. It's not the moving of the bits that is at issue, it's the how the bits are moved, where and who by.
    - zAy0LfpBZLC8mAC 12y ago
      "I think that caches work where you have a vast imbalance of requests for a particular bit of content - they seem to me to define centralization." Hu? Caches for geographical distribution are actually a form of decentralization! The problem with the existing ones is that they are mostly under centralized control, but there is no technical reason why that has to be, and an ISP's caching web proxy (which weren't that uncommon back in the day) is a simple example of caching infrastructure under decentralized control that could help an ISP reducing the load on their upstream connection for often-requested content. You don't need youtube to copy a video from your machine to the cache at some viewer's ISP, that's just the way it is commonly done right now because in that way youtube can make money - or, if you want to decentralize further, you could copy to a cache on the machine of some customer of said ISP in some geographic region: The more requests there are for some content, the more likely it is that there is a copy in your neighbourhood that doesn't need to be transferred from the original source.

kragen 12y ago

You may be interested that one week before, I wrote a post exploring your point (you are correct!) in detail: https://www.mail-archive.com/kragen-tol@canonical.org/msg00265.html https://www.mail-archive.com/kragen-tol@canonical.org/msg002... Summarized as, "Peer-to-peer overlay networks are inefficient on ADSL networks. ADSL networks are almost twice as efficient as SDSL networks. Better alternatives require redesigning the physical layer."

fidotron 12y ago

Actually the game is lost already because cellular modems are being embedded in products directly, skipping any ability you have to firewall them, all in the name of convenience. The fact is we rely too much on the cloud to connect two adjacent devices anyway. The whole notion that much of this data has to go upstream at all is the problem. (Shameless plug) This is what led me to write this: http://montrealrampage.com/king-ludd-19-manifesto-for-a-friendly-home/ http://montrealrampage.com/king-ludd-19-manifesto-for-a-frie...