3 ms·

For that matter, why can I get 200GB/mo at 50mbit/s plus a server for $20 from a VPS provider, and $.15/GB beyond that ($.10 if purchased in advance in the case

by jsz0 15y ago

For that matter, why can I get 200GB/mo at 50mbit/s plus a server for $20 from a VPS provider, and $.15/GB beyond that ($.10 if purchased in advance in the case of Linode)? Why do consumer ISPs need to charge so much more? The ISP has to maintain and upgrade thousands of miles of physical plant to deliver data to customer's homes in the geographic regions they cover. For a typical cable ISP for instance there's about 3-5 miles of copper between your modem and the the fiber node and typically dozens of miles between the fiber node and the provider's headend/datacenter where they uplink to the Internet -- which incidentally for smaller rural ISPs may itself be hundreds of miles away from competitive rates on data. Going back to the cost of maintaing/upgrading the infrastructure a small plant of only ~1000 miles typically requires at least a dozen full time employees simply to maintain it day to day and that's not including things like management, network engineers, customer service, technical support, equipment, contracted labour, customer installation, etc. On top of that you've got the cost of leasing access to the poles/underground conduits and the cost of powering the 3-5 miles of plant plus whatever operational costs you have to put gas into your fleet of vehicles, the cost of the vehicles themselves, insurance, etc. When you put this all together you're looking at some pretty huge overhead on the consumer side of Internet access plus whatever profit the company wants to make to justify all of the above. It's not a charity after all.

nitrogen 15y ago
That still doesn't justify the 250GB cap. Once the line is installed and maintained, using it more or less doesn't change the cost of installing and maintaining it. At that point we're talking about bulk bandwidth, which if you live in a metropolitan area, cannot cost significantly more than what Linode is charging.
- jsz0 15y ago
  In the case of cable Internet you're dealing with a shared medium. Your ISP may have a 10Gbit uplink in their headend but that does not extend to your home. In reality with 4 channel DOCSIS 3 bonding the speed available "on-the-wire" is about ~150Mbit/sec and it's shared between about 200 customers at minimum. (in reality it's usually a lot closer to 300-500 customers) So if 15 customers out of the 200 want to use 10Mbit/sec 24x7 there is no bandwidth left over for the other 190 customers. Just to be clear this part is not strictly an infrastructure challenge as the physical cable plant is not inherently limited to ~150Mbit/sec. It simply depends how many DOCSIS channels are being sent out and how many customers share them. They originate from a CMTS in the provider's headend. A fully loaded CMTS sets you back about $500-$750k. So cable ISPs figured out capping data can save them quite a lot of money. The alternative is a) they make less money b) they charge the customer more. We know A certainly isn't going to happen in a non-competitive market and B is something most customers would not like so we get c) capping. Telco or FTTH are a completely different beast though. Much harder to justify capping there. For telco it's usually because there is a limited uplink to the DSLAM. As of a few years ago it was not uncommon to have a few T1s feeding a DSLAM. FTTH is rarely capped because the fiber that goes to your home is capable of pushing 10Gbit/sec with the right optics. WISPS have the same issue as cable -- in fact WiMAX is very similar to DOCSIS and some WISPs actually do use DOCSIS over wireless at 10Ghz.
  - nitrogen 15y ago
    In the case of cable Internet you're dealing with a shared medium. Your ISP may have a 10Gbit uplink in their headend but that does not extend to your home. In reality with 4 channel DOCSIS 3 bonding the speed available "on-the-wire" is about ~150Mbit/sec and it's shared between about 200 customers at minimum. (in reality it's usually a lot closer to 300-500 customers) So if 15 customers out of the 200 want to use 10Mbit/sec 24x7 there is no bandwidth left over for the other 190 customers. I realize that the cable is shared, but remain unconvinced that a flat cap is in any way better than other solutions (such as dividing available bandwidth proportionally among active users, with higher-paying customers getting a larger slice). It shouldn't matter if someone's using their 10mbit/s 24/7, it should only matter if they're using it during peak hours. On the flip side, the "average" user who wants to stream Hulu, Youtube, and Netflix is likely to be doing their streaming at the same time as most other customers, causing more degradation to other customers' experience per hour of usage than the heavy downloaders. Capping the heavy users doesn't change the fact that the average users all want to stream video at the same time. Also, how can the cable company possibly offer me the 50+mbit/s connection for which I currently pay over $100/mo if the 4 256-QAM downstream channels (totaling ~150mbit/s as you said) are shared by so many customers? Those channels only use 24MHz of the total spectrum.
    - jsz0 15y ago
      I agree the flat cap isn't going to be the end of this. It just happens to be the easiest solution right now and the heavy users are low hanging fruit. As for the 50Mbit on a 150Mbit bonding group -- that's he magic of oversubscription. It's amazing how little bandwidth people actually use. The number of heavy users who are doing 50Mbit/sec non-stop are really the fringe cases. It's actually quite hard to figure out how to max out a 50 or 100mbit/sec for long periods of time. The traffic patterns are extremely bursty. For example a single uncapped modem's traffic. (in MB/sec) http://i.imgur.com/KL1YD.png http://i.imgur.com/KL1YD.png (120Mbit/sec is the bottleneck of my router)

by jsz0 15y ago

nitrogen 15y ago

That still doesn't justify the 250GB cap. Once the line is installed and maintained, using it more or less doesn't change the cost of installing and maintaining it. At that point we're talking about bulk bandwidth, which if you live in a metropolitan area, cannot cost significantly more than what Linode is charging.

jsz0 15y ago
In the case of cable Internet you're dealing with a shared medium. Your ISP may have a 10Gbit uplink in their headend but that does not extend to your home. In reality with 4 channel DOCSIS 3 bonding the speed available "on-the-wire" is about ~150Mbit/sec and it's shared between about 200 customers at minimum. (in reality it's usually a lot closer to 300-500 customers) So if 15 customers out of the 200 want to use 10Mbit/sec 24x7 there is no bandwidth left over for the other 190 customers. Just to be clear this part is not strictly an infrastructure challenge as the physical cable plant is not inherently limited to ~150Mbit/sec. It simply depends how many DOCSIS channels are being sent out and how many customers share them. They originate from a CMTS in the provider's headend. A fully loaded CMTS sets you back about $500-$750k. So cable ISPs figured out capping data can save them quite a lot of money. The alternative is a) they make less money b) they charge the customer more. We know A certainly isn't going to happen in a non-competitive market and B is something most customers would not like so we get c) capping. Telco or FTTH are a completely different beast though. Much harder to justify capping there. For telco it's usually because there is a limited uplink to the DSLAM. As of a few years ago it was not uncommon to have a few T1s feeding a DSLAM. FTTH is rarely capped because the fiber that goes to your home is capable of pushing 10Gbit/sec with the right optics. WISPS have the same issue as cable -- in fact WiMAX is very similar to DOCSIS and some WISPs actually do use DOCSIS over wireless at 10Ghz.
- nitrogen 15y ago
  In the case of cable Internet you're dealing with a shared medium. Your ISP may have a 10Gbit uplink in their headend but that does not extend to your home. In reality with 4 channel DOCSIS 3 bonding the speed available "on-the-wire" is about ~150Mbit/sec and it's shared between about 200 customers at minimum. (in reality it's usually a lot closer to 300-500 customers) So if 15 customers out of the 200 want to use 10Mbit/sec 24x7 there is no bandwidth left over for the other 190 customers. I realize that the cable is shared, but remain unconvinced that a flat cap is in any way better than other solutions (such as dividing available bandwidth proportionally among active users, with higher-paying customers getting a larger slice). It shouldn't matter if someone's using their 10mbit/s 24/7, it should only matter if they're using it during peak hours. On the flip side, the "average" user who wants to stream Hulu, Youtube, and Netflix is likely to be doing their streaming at the same time as most other customers, causing more degradation to other customers' experience per hour of usage than the heavy downloaders. Capping the heavy users doesn't change the fact that the average users all want to stream video at the same time. Also, how can the cable company possibly offer me the 50+mbit/s connection for which I currently pay over $100/mo if the 4 256-QAM downstream channels (totaling ~150mbit/s as you said) are shared by so many customers? Those channels only use 24MHz of the total spectrum.
  - jsz0 15y ago
    I agree the flat cap isn't going to be the end of this. It just happens to be the easiest solution right now and the heavy users are low hanging fruit. As for the 50Mbit on a 150Mbit bonding group -- that's he magic of oversubscription. It's amazing how little bandwidth people actually use. The number of heavy users who are doing 50Mbit/sec non-stop are really the fringe cases. It's actually quite hard to figure out how to max out a 50 or 100mbit/sec for long periods of time. The traffic patterns are extremely bursty. For example a single uncapped modem's traffic. (in MB/sec) http://i.imgur.com/KL1YD.png http://i.imgur.com/KL1YD.png (120Mbit/sec is the bottleneck of my router)