6 ms·

You're accepting as fact that rising bandwidth usage = rising costs. This is not how the technology works. A 10 gigabit switch a few years ago used to cost $1,

You're accepting as fact that rising bandwidth usage = rising costs. This is not how the technology works. A 10 gigabit switch a few years ago used to cost $1,000,000 and now it costs $10,000. CPUs, optics, everything is just printed plastic and metal boxes. There's nothing fundamentally expensive about any of it. Engineers and scientists can drive the price of bandwidth down to nearly zero. Comcast and AT&T simply want to extract as much money as they possibly can. None of it would be possible if they didn't bribe politicians and run massive disinformation campaigns against the population that convinces them that bandwidth is rare and expensive. People are defending the idea that we should be happy to have the scraps Comcast and AT&T see fit to give us. We should be demanding to exercise the full power and freedom that the technology allows. Saying you don't use it so you don't care is just shorted sighted and selfish. The poor kid in Chicago educating himself by watching YouTube videos is going to end up getting his parents water shut off with $200 overage fees.

rayiner 9y ago
Internet bandwidth usage is growing 25% per year: https://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/vni-hyperconnectivity-wp.html https://www.cisco.com/c/en/us/solutions/collateral/service-p.... The amount of performance you get out of $1 of Juniper or Cisco gear is not going up 25% per year. And it's not even just about upgrading to the latest switch or router. You've got a cable network where each node used to support up to thousands of users, and each user might be several layers of amps away from the node. To accommodate newer generations of DOCSIS, you've got to split nodes and eliminate the intermediate layers of amps: http://www.lightreading.com/cable/ccap-next-gen-nets/cable-nodes-becoming-a-choke-point/d/d-id/728754 http://www.lightreading.com/cable/ccap-next-gen-nets/cable-n.... Replacing 1 node with 10 isn't cheap, nor are the (often unionized) technicians that have to drive around neighborhoods burying fiber and installing nodes.
- staunch 9y ago
  1. Juniper and Cisco aren't nearly as important as they used to be and their importance is growing less every day. Networking gear is headed towards commodity pricing. 2. Network technology almost certainly improves by 25%+ per year in terms of price-to-performance. Generational upgrades will make it uneven but that's not an unrealistic number to hit on average. 3. The "last mile" is the hard part but sticking with cable (DOCSIS) technology is not a requirement. There's no reason we couldn't have a massive federal Fiber To The Home program that cuts through current regulations. The one-time costs are undeniable but they can be easily and predictably amortized over a long period of time. The theory that networking technology can't keep up is technically baseless.
  - rayiner 9y ago
    > 1. Juniper and Cisco aren't nearly as important as they used to be and their importance is growing less every day. If I were running a multi-billion ISP, I'd use Juniper and Cisco gear too. > 2. Network technology almost certainly improves by 25%+ per year in terms of price-to-performance. Generational upgrades will make it uneven but that's not an unrealistic number to hit on average. It took 12 years to go from 10 gigabit ethernet to 100 gigabit (about 21% per year over the decade), but I'd be very surprised if the price was the same (i.e. that 100 gig switches cost the same now as 10 gig switches cost back then). Carriers and data centers moved directly from 1G to 10G. But it seems like with 100G, there is significant pricing back pressure creating the need for intermediate speeds: http://www.mellanox.com/blog/2016/03/25-is-the-new-10-50-is-the-new-40-100-is-the-new-amazing http://www.mellanox.com/blog/2016/03/25-is-the-new-10-50-is-.... But let's say a 100G router today costs the same as a 10G router cost 12 years ago. 21% per year is just about what you need to keep up with projected demand growth over the next five years: https://www.cisco.com/c/dam/m/en_us/solutions/service-provider/vni-forecast-highlights/pdf/United_States_2021_Forecast_Highlights.pdf https://www.cisco.com/c/dam/m/en_us/solutions/service-provid.... Even if 100% of your cost was scalable at the same level, you're talking about keeping up with demand for the same money, not less money. Of course, 100% of your costs aren't scalable. The last mile DOCSIS part of your network is not scaling 20% per year. > 3. The "last mile" is the hard part but sticking with cable (DOCSIS) technology is not a requirement. Verizon tried that. It's barely breaking even on FiOS even though the average FiOS customer brings in more than $100/month in revenue. > There's no reason we couldn't have a massive federal Fiber To The Home program that cuts through current regulations. I can think of way better uses for hundreds of billions of federal dollars than subsidizing Netflix's business model: http://www.afr.com/technology/web/nbn/australias-highspeed-internet-trails-thailand-kenya-20170511-gw34ln http://www.afr.com/technology/web/nbn/australias-highspeed-i.... For the forseeable future, fiber is a luxury product, and to the extent we want government-funded broadband to address "Internet access as a human right" issues, we should do it using cost-effective 5G wireless technology. > The one-time costs are undeniable but they can be easily and predictably amortized over a long period of time. The theory that networking technology can't keep up is technically baseless. It's not that networking can't keep up. It's that upgrading the existing cable network to keep up with demand growth is expensive, and replacing everything with fiber is even more expensive.
harryh 9y ago
You're talking about changes in absolute costs. I'm talking about relative costs. It doesn't matter how good networking equipment gets. It will always cost providers more money to service users saturating their connections than more typical consumers that don't saturate but care about burst speed. For every poor kid in Chicago that happens to be using a highly above average amount of total bandwidth per month there will be 50 poor kids in Chicago with more typical usage patterns that your plan would hurt.
- staunch 9y ago
  > It will always cost providers more money to service users saturating their connections ... I don't think it's an accident that so many people believe this myth. It sounds logical enough but it's completely false. Fact #1 You have to replace technology because it fails over time. Fact #2 The cost of networking technology is decreasing every year. Fact #3. The performance of networking technology increasing every year. Fact #4 The costs of running fiber to premises is a (mostly) one-time expense. There's no reason that users couldn't have more bandwidth for less money every year. Technology is magical like this.
  - harryh 9y ago
    None of those facts contradict my point as they, again, speak towards absolute rather than relative costs. Providers have a finite amount of bandwidth in their trunk connections. More bursty users can use a single trunk than saturating users. Therefore saturating users are more expensive. It's just like CPU. If you're bursting up to capacity but on average only using, say, 10% of available CPU then Amazon can pack more of your workloads on a single physical machine compared to a different user that is using 100% of CPU all the time. The fact the CPUs are always getting cheaper and faster doesn't negate this point.
    - staunch 9y ago
      Amazon has dropped the price of compute power many times for precisely the same reasons I'm describing. https://www.cloudyn.com/blog/analyzing-aws-ec2-price-drops-over-the-past-5-years/ https://www.cloudyn.com/blog/analyzing-aws-ec2-price-drops-o... This is exactly what would be happening if Comcast and AT&T had competition or smart regulation.

rayiner 9y ago

Internet bandwidth usage is growing 25% per year: https://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/vni-hyperconnectivity-wp.html https://www.cisco.com/c/en/us/solutions/collateral/service-p.... The amount of performance you get out of $1 of Juniper or Cisco gear is not going up 25% per year. And it's not even just about upgrading to the latest switch or router. You've got a cable network where each node used to support up to thousands of users, and each user might be several layers of amps away from the node. To accommodate newer generations of DOCSIS, you've got to split nodes and eliminate the intermediate layers of amps: http://www.lightreading.com/cable/ccap-next-gen-nets/cable-nodes-becoming-a-choke-point/d/d-id/728754 http://www.lightreading.com/cable/ccap-next-gen-nets/cable-n.... Replacing 1 node with 10 isn't cheap, nor are the (often unionized) technicians that have to drive around neighborhoods burying fiber and installing nodes.

staunch 9y ago
1. Juniper and Cisco aren't nearly as important as they used to be and their importance is growing less every day. Networking gear is headed towards commodity pricing. 2. Network technology almost certainly improves by 25%+ per year in terms of price-to-performance. Generational upgrades will make it uneven but that's not an unrealistic number to hit on average. 3. The "last mile" is the hard part but sticking with cable (DOCSIS) technology is not a requirement. There's no reason we couldn't have a massive federal Fiber To The Home program that cuts through current regulations. The one-time costs are undeniable but they can be easily and predictably amortized over a long period of time. The theory that networking technology can't keep up is technically baseless.
- rayiner 9y ago
  > 1. Juniper and Cisco aren't nearly as important as they used to be and their importance is growing less every day. If I were running a multi-billion ISP, I'd use Juniper and Cisco gear too. > 2. Network technology almost certainly improves by 25%+ per year in terms of price-to-performance. Generational upgrades will make it uneven but that's not an unrealistic number to hit on average. It took 12 years to go from 10 gigabit ethernet to 100 gigabit (about 21% per year over the decade), but I'd be very surprised if the price was the same (i.e. that 100 gig switches cost the same now as 10 gig switches cost back then). Carriers and data centers moved directly from 1G to 10G. But it seems like with 100G, there is significant pricing back pressure creating the need for intermediate speeds: http://www.mellanox.com/blog/2016/03/25-is-the-new-10-50-is-the-new-40-100-is-the-new-amazing http://www.mellanox.com/blog/2016/03/25-is-the-new-10-50-is-.... But let's say a 100G router today costs the same as a 10G router cost 12 years ago. 21% per year is just about what you need to keep up with projected demand growth over the next five years: https://www.cisco.com/c/dam/m/en_us/solutions/service-provider/vni-forecast-highlights/pdf/United_States_2021_Forecast_Highlights.pdf https://www.cisco.com/c/dam/m/en_us/solutions/service-provid.... Even if 100% of your cost was scalable at the same level, you're talking about keeping up with demand for the same money, not less money. Of course, 100% of your costs aren't scalable. The last mile DOCSIS part of your network is not scaling 20% per year. > 3. The "last mile" is the hard part but sticking with cable (DOCSIS) technology is not a requirement. Verizon tried that. It's barely breaking even on FiOS even though the average FiOS customer brings in more than $100/month in revenue. > There's no reason we couldn't have a massive federal Fiber To The Home program that cuts through current regulations. I can think of way better uses for hundreds of billions of federal dollars than subsidizing Netflix's business model: http://www.afr.com/technology/web/nbn/australias-highspeed-internet-trails-thailand-kenya-20170511-gw34ln http://www.afr.com/technology/web/nbn/australias-highspeed-i.... For the forseeable future, fiber is a luxury product, and to the extent we want government-funded broadband to address "Internet access as a human right" issues, we should do it using cost-effective 5G wireless technology. > The one-time costs are undeniable but they can be easily and predictably amortized over a long period of time. The theory that networking technology can't keep up is technically baseless. It's not that networking can't keep up. It's that upgrading the existing cable network to keep up with demand growth is expensive, and replacing everything with fiber is even more expensive.

harryh 9y ago

You're talking about changes in absolute costs. I'm talking about relative costs. It doesn't matter how good networking equipment gets. It will always cost providers more money to service users saturating their connections than more typical consumers that don't saturate but care about burst speed. For every poor kid in Chicago that happens to be using a highly above average amount of total bandwidth per month there will be 50 poor kids in Chicago with more typical usage patterns that your plan would hurt.

staunch 9y ago
> It will always cost providers more money to service users saturating their connections ... I don't think it's an accident that so many people believe this myth. It sounds logical enough but it's completely false. Fact #1 You have to replace technology because it fails over time. Fact #2 The cost of networking technology is decreasing every year. Fact #3. The performance of networking technology increasing every year. Fact #4 The costs of running fiber to premises is a (mostly) one-time expense. There's no reason that users couldn't have more bandwidth for less money every year. Technology is magical like this.
- harryh 9y ago
  None of those facts contradict my point as they, again, speak towards absolute rather than relative costs. Providers have a finite amount of bandwidth in their trunk connections. More bursty users can use a single trunk than saturating users. Therefore saturating users are more expensive. It's just like CPU. If you're bursting up to capacity but on average only using, say, 10% of available CPU then Amazon can pack more of your workloads on a single physical machine compared to a different user that is using 100% of CPU all the time. The fact the CPUs are always getting cheaper and faster doesn't negate this point.
  - staunch 9y ago
    Amazon has dropped the price of compute power many times for precisely the same reasons I'm describing. https://www.cloudyn.com/blog/analyzing-aws-ec2-price-drops-over-the-past-5-years/ https://www.cloudyn.com/blog/analyzing-aws-ec2-price-drops-o... This is exactly what would be happening if Comcast and AT&T had competition or smart regulation.