> 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-....
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-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-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.
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-....
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-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-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.