Fiber to the home/GPON is also a shared medium. One OLT (the fiber device in the headend) port generally goes out to a splitter in the field that then goes to 128,64,or 32 customers which has about ~2.5 gbps of downstream speed that is probably also shared with the video service. That is generally more capacity than DOCSIS but it isn't a fiber straight from the customer house back to the provider network.
Modern DOCSIS networks are much more capable than the linked diagram. The general guideline is that your top speed package shouldn't be any more than half of the node capacity (which ends up being similar to GPON that often does a 1gbps service on a shared ~2.5gbps network). The video portion of a HFC network also doesn't share the bandwidth that has been assigned to Data.
The diagram shows a node of 500-2000 homes - that is a HUGE number, I've worked on networks with nodes much larger than they should be and about 250 is the biggest node. ~100 customers per node is more common and in newer or upgraded HFC networks the goal is normally N+1 or N+0 (Node plus one active device or zero active devices - the linked diagram is N+10) those numbers get down to the 32-64 customers/node range.
I do work on DOCSIS networks for small providers but some of the tools the big guys use to monitor the RF quality is pretty amazing. They are able to read the RF Spectrum of every modem in the network and find common impairments in the cable plant and they can also assign individual modems different modulation profiles, so if your inside wiring and drop cable are capable of it your modem gets a higher modulation profile (and higher speed) than your neighbors that have crappy radio shack cable (the inside wiring has a huge impact on the signal quality). Traditionally you would have to set the service to work for the lowest common denominator but the emerging technologies are changing that. That doesn't necessary mean the big guys don't have capacity issues in other places (example: Netflix peering) but there is quite a bit of work and technology behind the last mile delivery.
You also can't ignore the business aspects - there is significant capital investment in upgrading any last mile network. I can take a group of 200 customers at 50mbps and upgrade them to 200mbps and the overall traffic pattern doesn't hardly change at all. That makes it hard for operators to justify to shareholders why spending hundreds of thousands or millions of dollars to go from 500mbps to 1gbsp is a good investment.
Perhaps not mentioned (that i see thus far) is how density drives the economics of telecom network deployment. Network will be deployed (and upgraded) in direct proportion to the availability of revenue (homes passed)/dollar of capital investment. Those who live in high density cities will have abundant (1Gbs) and relatively cheap bandwidth. Most people living in remote, rural areas will not receive abundant (1Gbs) bandwidth unless someone subsidizes it.
I live inside the perimeter in Atlanta, GA, where ATT provides a 1Gbs fiber connection for $70/month. I just tested it at 878 Mbs down, 650 Mbs up. And that performance has been consistent for over 2 yrs, with no outages except after hurricane or ice storm.
Before that, Comcast couldn't deliver 25 Mbs and went down at least once a month, for hours at a time. Comcast now offers a 1 Gbs service here, but I wouldn't trust its reliability.
Chattanooga, TN has three plus private ISP's. The local power company (EPB) fought AT&T and Comcast tooth and nail to obtain its telecom services license from the City of Chattanooga/Hamilton County. EPB offers 1 Gbs for $68/mo, 10Gbs for $300/month - residential service - and according to family members, the service is excellent (TV and phone offerings also available). EPB offers 1Gbs service to 98% of the city. Neither At&T nor Comcast are close in coverage, but both claim to offer 1 Gbs service.
The key to EPB's success - they already owned the rights of way required for fiber infrastructure (electric grid), they had the capital to commit to the long term investment required for telecom payback (7-10 years), they deployed a fiber-based network from the outset, they hired experienced telecom people to run that business, and they were/remain a respected corporate citizen dedicated only to serving their local area.
Those rights of way are a key consideration often overlooked when outsiders evaluate telecom economics. LOTS of legal and regulatory (local, state, federal) hurdles involved in obtaining them if one doesn't already have them.
It's because software services are built according to the infrastructure of their typical customer. If symmetrical connections were the standard in america, how would that change the software landscape today?
Modern DOCSIS networks are much more capable than the linked diagram. The general guideline is that your top speed package shouldn't be any more than half of the node capacity (which ends up being similar to GPON that often does a 1gbps service on a shared ~2.5gbps network). The video portion of a HFC network also doesn't share the bandwidth that has been assigned to Data.
The diagram shows a node of 500-2000 homes - that is a HUGE number, I've worked on networks with nodes much larger than they should be and about 250 is the biggest node. ~100 customers per node is more common and in newer or upgraded HFC networks the goal is normally N+1 or N+0 (Node plus one active device or zero active devices - the linked diagram is N+10) those numbers get down to the 32-64 customers/node range.
I do work on DOCSIS networks for small providers but some of the tools the big guys use to monitor the RF quality is pretty amazing. They are able to read the RF Spectrum of every modem in the network and find common impairments in the cable plant and they can also assign individual modems different modulation profiles, so if your inside wiring and drop cable are capable of it your modem gets a higher modulation profile (and higher speed) than your neighbors that have crappy radio shack cable (the inside wiring has a huge impact on the signal quality). Traditionally you would have to set the service to work for the lowest common denominator but the emerging technologies are changing that. That doesn't necessary mean the big guys don't have capacity issues in other places (example: Netflix peering) but there is quite a bit of work and technology behind the last mile delivery.
You also can't ignore the business aspects - there is significant capital investment in upgrading any last mile network. I can take a group of 200 customers at 50mbps and upgrade them to 200mbps and the overall traffic pattern doesn't hardly change at all. That makes it hard for operators to justify to shareholders why spending hundreds of thousands or millions of dollars to go from 500mbps to 1gbsp is a good investment.