Imagine releasing a CPU manual that explains all of the transistors and quantum physics math behind their layouts. FPGAs and stuff you can document, but advanced electronics requires some real math and physics backgrounds.
The Internet works on standards, not ad-hoc network layer DNS resolving based on the application layer logic. try building a sky-scraper when using lots of different standards, see how well it works.
With a base cost of about $30 for last mile infrastructure and trunk bandwidth costing less than $1/mbit, an ISP could sell a reserved+dedicated 50mb/50mb for $100/month and make a $20 profit. By next year, the trunk bandwidth cost will be about halved on average, so $75/month and turn a $20 profit, then the next year $62.5/month and turn a $20 profit. So about 2 years from now, a 200/200 dedicated connection with full non-blocking reserved bandwidth will cost about $80 and your ISP could sell for $100 and turn a $20 profit. Based on averages.
I rural farm area, the cost of FTTH is about $5k per house passed, assuming mostly above ground. At $40/m. It will take about 11 years to pay it off at that rate, but quite a few people closer to the $70 range, which puts it at 6 year mark, but that ignores lots of the on-going costs. Yes, need some government money. The good news is once it's installed, you're good for the next century.
As for a trunk you don't need to purchase a new trunk every time you need to add bandwidth, you just purchase a new line card and add another 500gb. Current tech can push about 288tb/s over a 72 fiber trunk and that will give you a range of 700km with no signal regenerators. Just make sure you lay enough fiber the first time. Fiber is cheap, the man-hours to lay the fiber is not.
A single low cost Active Ethernet chassis can handle 500 users all using 1gb/1gb at the same time with a full non-blocking backplane and 500gb of uplink. This single chassis is housed in a datecenter because of the 80km range of the fiber and port. Once this fiber is laid, it's good for about 50 years.
Cable on the other hand requires a node that's about as expensive as the fiber chassis, except this node must be housed in the field, needs to have power lines trenched to it, have cooling, heating, handle high and low humidity, have UPS batteries. This single node can only handle up to a few hundred users and must share the sub 1gb of bandwidth. The actual COAX requires filters and amplifiers fitted every so much distance, it is very sensitive to noise and has transient signal issues quite often. The COAX needs to be upgraded, replaced, and new filters and amps need to be setup and maintained.
You can't just let anyone and everyone dig, the citizens will fire the mayor and pass a law that only allows 1-2 companies to ever dig, just like we have now. If you want it to be a free market, then get ready for someone to pull out a gun to defend their land if someone tries to dig it up.
Camfrog lists bandwidth usage about 38KB/stream, which means about 270 streams on a 100mb pipe. I don't see regular people getting together and video chatting with 269 other people that often to make use of that 100mb pipe that you claim a "regular" person could with a 100mb connection.
It makes sense to have no such thing as "tiers"; simply have all users in the 100 megabits tier, and those who have higher bits per second peak usage on their link for longer periods of time, get the call, and have a choice to either pay more -- leave, or request that their downstream be capped to 1 megabit.
A user stuck downloading slowly for a long is a base load that makes poor use of trunk bandwidth, while a user with a fast fat pipe does very short bursts of data that are transient and make good usage of the trunk. So not only does efficiency go up, but usage time goes down.
Something like 10% of people executed are later proven innocent when purposefully hidden evidence surfaces. Those district attorneys really really like to win.
Yes, lets murder innocent people in a cruel way./sarc
The larger ISPs don't buy bandwidth, they buy ports. When a port gets past 50% utilization of the 95th percentile, they add a new port. Just because you decided to start remotely backing up your 10TB storage system and it will take you a month to do your initial sync, doesn't mean crap on your ISP's 500gb trunk. you're just a single data point in an ocean of averages.
95th percentile of your 500gb link at 50%? Add another 100gb port. Bigger ISPs are now doing 400gb-500gb increments. All over the same fiber of course, so their dark-fiber lease doesn't go up at all. They just place an order for a new line card, 2 days later they get it in the mail, they spend 12minutes installing it, and there you go, another 500gb.
For internal it's this easy, for peering it's a bit more complicated since you need to wait for provisioning the ports. But the main point is that the more industry practice is to have plenty of free bandwidth to have room for burst. Even if the burst is over a 30 day period because some random user decided to sync 10TB, it's still burst.
Most enterprise connections are priced on 95th percentile, which means only the peak sets the price, non peak has no bearing on costs. Per MB is an 80/20 rule that assume all users transfer during peak hours, but a smart user can time their bulk transfers during off-peak, which completely breaks the assumption made by per MB.
Two big reasons why ISPs don't like lots of upload is most of their last-mile infrastructure is old copper-based shared connections where over-subscription can cause congestion if too many high data users are active at once. The other thing is they want to discourage business users from using a cheap connection.
Once you get to more modern last-mile no-congestion fiber designs like Sonic.Net, you cease to care about a few heavy users, unless your plan is to price gouge and fleece all you can.
What an ISP really doesn't want is to hand out 1gb connections where the assumption is most users don't make use of their bandwidth, then they get a few super heavy users and are effectively running a datacenter that can make use of the upstream all the time. You may suddenly go from 10 heavy BitTorrent seeders to 100 business users.
Trying to properly handle a 1gb/1gb connection takes some decent hardware. Your average $200 gamer router might be able to handle a total of 1gb of a few dataflows, but get 10,000+ connections, your router is going to crash-and-burn. Even BitTorrent doesn't let you upload 1gb/s all the time. There is only so much demand and 1gb is A LOT of bandwidth. You can transfer 3GB is less than 30 seconds. Even if you were the only seeder, 120 leechers can download all 3GB in less than one hour from you.
1gb is currently so high above what people are used to, it's akin to giving someone 100TB of storage and asking them to find legitimate ways to fill it up. You can put an entire lifetime of BluRay movies in there and still not fill it up. It takes work. Even your most hardcore computer users will have a hard time using up 1gb without turning it into a chore and throwing serious money at it.
Libraries can have unsupervised children and the library is responsible. It's like saying Schools block porn. Wow, imagine that. anyway, most libraries that I know will allow blocks to be removed at request by an adult for only their account. Even my local uni allows the general public, that can prove they live in the city, to have access to computer labs where they explicitly state "if you're viewing something offensive like porn, be discrete and cease viewing immediately or move if someone is offended"
The first thing the lab assistants tell you is "if you see someone in a corner with their monitor turned, they're probably watching something offensive, don't sit next to them if you're easily offended".
It becomes the biggest sucker game: who gives the unlimited quota to people with the most upload bandwidth: they get to seed for all the other networks.
I was talking to a small ISP owner and he told me he doesn't mind people who upload a lot because the connections the ISP pays for are all symmetrical and the overwhelming bulk of users download, not upload. This means upload is free because the download sets the cost. He said even with all of his heavy torrent seeders, his upload usage is much lower than download, so it doesn't cost him a penny.
Just look at Sonic.Net. They sell 1gb/1gb to residential for $70. They just started a business option of 1gb/100mb for $40/month. Wouldn't residential users rather be paying for that? Well, businesses tend to operate on off-peak hours. This means that 1gb of bandwidth is FREE for Sonic.Net. I guess you could say that residential users are subsidizing the business users in a way, but at the same time, they are not. It's not about how much you consume, it's about WHEN you consume. Residential users tend to consume during peak hours.
You fail to account that the 1mb/1mb line costs as much as the 1gb/1gb line. What they will do is give you a 1gb line rate limited to 1mb. Then you also fail to account that your average 50GB/month user costs the same amount of bandwidth as your typical 800GB/month user.
I could easily shape my traffic in a way that I could transfer 4TB in one month and cost my ISP less money than someone transfer 10GB. I just do all of my transfers during off-peak hours.
Time and time again, research has shown data caps, short of extreme ones, make absolutely no difference on costs or congestion. 5% of the users make up 90% of the total data transferred in a 24 hour period, but only a small portion of peak traffic during the 2 hours of the day that actually matter.
Many of the larger ISPs don't even pay for mbits, they pay for ports. Got a 10gb port? You pay a flat rate no matter its usage. 100% usage, 0% usage, same price. But you need more ports if you go over. Yeah, well a 100gb port is only 2x the price of a 10gb port. Now they have 400gb ports and soon 1tb ports.
My point was that you could sell 25/25 for $70/m with no over-subscription and 100% dedicated bandwidth for 100% of all users. That price is almost competitive with the 30/4 connections Charter sells for $60 around here, and I'm sure transferring 10.5TB in one month that a 30/4 connection can push would get you a nasty-gram. A 25/25 connection could transfer about 16TB in one month.
That's a statistical anomaly that is less than the rounding errors of the many other users using their connection to browse FaceBook and look at cat pics. In a normal large population, a single person, no matter how badly they abuse their connection, will not make a difference on the trunk.
TWC is working on a plan that you can cut $5 off of your $100/month bill if you accept a 5GB cap instead of a 250GB cap. Obviously they must think that 245GB is worth about $5, yet they want to charge $10 per 10GB for overages. They're just making up numbers. Just think about text messaging. They charge more per byte to text message than to rent time+data from the Hubble telescope or Mars Rover.
Yes, if electricity was free to generate, how much would it cost to deliver it? The logical answer is not to charge based on consuming the self-replenishing resource, but charge based on the cost to distribute it. In other words, you don't charge based on usage, you charge base on load.
Bandwidth in infinite in that it can be created without restriction, but there is only ever so much available at a given moment. As long as bandwidth being consumed is less than the rate at which it is replenished, it does not cost anything, only the infrastructure to deliver the bandwidth costs.
At slightly less than $1/mbit for dedicated bandwidth on the backbone, you could sell a 25/25 connection for $70/month($25/m for bandwidth costs and $45 for everything else), and make money even if EVERY SINGLE USER used 100% of the bandwidth 24/7 all month long.
A typical, medium-sized/30 floor apartment building can have 300 units in it. Are you going to run an OC-192 line to a single building?
No, you're going to run 300 strands of fiber for about $300/kilometer, for an overall average one time fee of about $3,000 per unit. Then each unit will get it's own 1gb fiber port which will cost about $60 and supply them with a 1gb ONT for another $150, which is all part of the $3,000. Then you pay up-stream bandwidth an electrical costs from there on out. The fiber you laid will last about 40-100 years and the 1gb connection will take about 3 years to earn itself back.
I don't understand the whole OC-48 stuff. Upstream providers don't even advertise in those units anymore, they just sell 1gb/10gb/100gb ethernet ports. Want 1gb of dedicated bandwidth from a Tier 1 ISP? $6,000 per month, that's 1,000mb.
Price for Darkfiber, about $115/mile/month. Need two fibers, say an IX is 200 miles away, that's about $50k/month. New DWDM multiplexing photonic integrated circuits are able to get 1tb/s-8tb/s, depending on the company. While I can't find pricing, they do advertise crazy numbers like "80% cheaper per mbit than previous models". Anyway, those are sunk costs, not reoccurring. So assume 1tb/s. 100gb runs about $5k/month at an IX, so another $50k. So for $100k/month, you got yourself 1tb/s of bandwidth that you can get for free from YouTube and Netflix since peering is free with them at any IX. That's about $0.10/mbit.
The industry average real-world pricing being sold per mbit from a Tier 1, is less than $1 when buying in bulk(10gb-100gb+).
Slashdot Mirror
Imagine releasing a CPU manual that explains all of the transistors and quantum physics math behind their layouts. FPGAs and stuff you can document, but advanced electronics requires some real math and physics backgrounds.
Since you can get rid of stop signs and stop lights, I would say that they are better. Well, in non-residential zones anyway.
The Internet works on standards, not ad-hoc network layer DNS resolving based on the application layer logic. try building a sky-scraper when using lots of different standards, see how well it works.
With a base cost of about $30 for last mile infrastructure and trunk bandwidth costing less than $1/mbit, an ISP could sell a reserved+dedicated 50mb/50mb for $100/month and make a $20 profit. By next year, the trunk bandwidth cost will be about halved on average, so $75/month and turn a $20 profit, then the next year $62.5/month and turn a $20 profit. So about 2 years from now, a 200/200 dedicated connection with full non-blocking reserved bandwidth will cost about $80 and your ISP could sell for $100 and turn a $20 profit. Based on averages.
I rural farm area, the cost of FTTH is about $5k per house passed, assuming mostly above ground. At $40/m. It will take about 11 years to pay it off at that rate, but quite a few people closer to the $70 range, which puts it at 6 year mark, but that ignores lots of the on-going costs. Yes, need some government money. The good news is once it's installed, you're good for the next century.
As for a trunk you don't need to purchase a new trunk every time you need to add bandwidth, you just purchase a new line card and add another 500gb. Current tech can push about 288tb/s over a 72 fiber trunk and that will give you a range of 700km with no signal regenerators. Just make sure you lay enough fiber the first time. Fiber is cheap, the man-hours to lay the fiber is not.
A single low cost Active Ethernet chassis can handle 500 users all using 1gb/1gb at the same time with a full non-blocking backplane and 500gb of uplink. This single chassis is housed in a datecenter because of the 80km range of the fiber and port. Once this fiber is laid, it's good for about 50 years.
Cable on the other hand requires a node that's about as expensive as the fiber chassis, except this node must be housed in the field, needs to have power lines trenched to it, have cooling, heating, handle high and low humidity, have UPS batteries. This single node can only handle up to a few hundred users and must share the sub 1gb of bandwidth. The actual COAX requires filters and amplifiers fitted every so much distance, it is very sensitive to noise and has transient signal issues quite often. The COAX needs to be upgraded, replaced, and new filters and amps need to be setup and maintained.
So, fiber is cheaper, more reliable, and faster.
You can't just let anyone and everyone dig, the citizens will fire the mayor and pass a law that only allows 1-2 companies to ever dig, just like we have now. If you want it to be a free market, then get ready for someone to pull out a gun to defend their land if someone tries to dig it up.
Camfrog lists bandwidth usage about 38KB/stream, which means about 270 streams on a 100mb pipe. I don't see regular people getting together and video chatting with 269 other people that often to make use of that 100mb pipe that you claim a "regular" person could with a 100mb connection.
It makes sense to have no such thing as "tiers"; simply have all users in the 100 megabits tier, and those who have higher bits per second peak usage on their link for longer periods of time, get the call, and have a choice to either pay more -- leave, or request that their downstream be capped to 1 megabit.
A user stuck downloading slowly for a long is a base load that makes poor use of trunk bandwidth, while a user with a fast fat pipe does very short bursts of data that are transient and make good usage of the trunk. So not only does efficiency go up, but usage time goes down.
Something like 10% of people executed are later proven innocent when purposefully hidden evidence surfaces. Those district attorneys really really like to win.
/sarc
Yes, lets murder innocent people in a cruel way.
The larger ISPs don't buy bandwidth, they buy ports. When a port gets past 50% utilization of the 95th percentile, they add a new port. Just because you decided to start remotely backing up your 10TB storage system and it will take you a month to do your initial sync, doesn't mean crap on your ISP's 500gb trunk. you're just a single data point in an ocean of averages.
95th percentile of your 500gb link at 50%? Add another 100gb port. Bigger ISPs are now doing 400gb-500gb increments. All over the same fiber of course, so their dark-fiber lease doesn't go up at all. They just place an order for a new line card, 2 days later they get it in the mail, they spend 12minutes installing it, and there you go, another 500gb.
For internal it's this easy, for peering it's a bit more complicated since you need to wait for provisioning the ports. But the main point is that the more industry practice is to have plenty of free bandwidth to have room for burst. Even if the burst is over a 30 day period because some random user decided to sync 10TB, it's still burst.
I got 50mb of dedicated bandwidth that I don't share for $100/month. I just don't get an SLA.
Network equipment doesn't "wear out", it ages. Best make full use of it while you can.
Most enterprise connections are priced on 95th percentile, which means only the peak sets the price, non peak has no bearing on costs. Per MB is an 80/20 rule that assume all users transfer during peak hours, but a smart user can time their bulk transfers during off-peak, which completely breaks the assumption made by per MB.
Two big reasons why ISPs don't like lots of upload is most of their last-mile infrastructure is old copper-based shared connections where over-subscription can cause congestion if too many high data users are active at once. The other thing is they want to discourage business users from using a cheap connection.
Once you get to more modern last-mile no-congestion fiber designs like Sonic.Net, you cease to care about a few heavy users, unless your plan is to price gouge and fleece all you can.
What an ISP really doesn't want is to hand out 1gb connections where the assumption is most users don't make use of their bandwidth, then they get a few super heavy users and are effectively running a datacenter that can make use of the upstream all the time. You may suddenly go from 10 heavy BitTorrent seeders to 100 business users.
Trying to properly handle a 1gb/1gb connection takes some decent hardware. Your average $200 gamer router might be able to handle a total of 1gb of a few dataflows, but get 10,000+ connections, your router is going to crash-and-burn. Even BitTorrent doesn't let you upload 1gb/s all the time. There is only so much demand and 1gb is A LOT of bandwidth. You can transfer 3GB is less than 30 seconds. Even if you were the only seeder, 120 leechers can download all 3GB in less than one hour from you.
1gb is currently so high above what people are used to, it's akin to giving someone 100TB of storage and asking them to find legitimate ways to fill it up. You can put an entire lifetime of BluRay movies in there and still not fill it up. It takes work. Even your most hardcore computer users will have a hard time using up 1gb without turning it into a chore and throwing serious money at it.
Seinfeld had an episode about this. It involved a missing button, blurry vision, and a bit of bust.
The viewport is not your head, but the screen. It is 3D once you recognize that.
Libraries can have unsupervised children and the library is responsible. It's like saying Schools block porn. Wow, imagine that. anyway, most libraries that I know will allow blocks to be removed at request by an adult for only their account. Even my local uni allows the general public, that can prove they live in the city, to have access to computer labs where they explicitly state "if you're viewing something offensive like porn, be discrete and cease viewing immediately or move if someone is offended"
The first thing the lab assistants tell you is "if you see someone in a corner with their monitor turned, they're probably watching something offensive, don't sit next to them if you're easily offended".
It becomes the biggest sucker game: who gives the unlimited quota to people with the most upload bandwidth: they get to seed for all the other networks.
I was talking to a small ISP owner and he told me he doesn't mind people who upload a lot because the connections the ISP pays for are all symmetrical and the overwhelming bulk of users download, not upload. This means upload is free because the download sets the cost. He said even with all of his heavy torrent seeders, his upload usage is much lower than download, so it doesn't cost him a penny.
Just look at Sonic.Net. They sell 1gb/1gb to residential for $70. They just started a business option of 1gb/100mb for $40/month. Wouldn't residential users rather be paying for that? Well, businesses tend to operate on off-peak hours. This means that 1gb of bandwidth is FREE for Sonic.Net. I guess you could say that residential users are subsidizing the business users in a way, but at the same time, they are not. It's not about how much you consume, it's about WHEN you consume. Residential users tend to consume during peak hours.
You fail to account that the 1mb/1mb line costs as much as the 1gb/1gb line. What they will do is give you a 1gb line rate limited to 1mb. Then you also fail to account that your average 50GB/month user costs the same amount of bandwidth as your typical 800GB/month user.
I could easily shape my traffic in a way that I could transfer 4TB in one month and cost my ISP less money than someone transfer 10GB. I just do all of my transfers during off-peak hours.
Time and time again, research has shown data caps, short of extreme ones, make absolutely no difference on costs or congestion. 5% of the users make up 90% of the total data transferred in a 24 hour period, but only a small portion of peak traffic during the 2 hours of the day that actually matter.
Many of the larger ISPs don't even pay for mbits, they pay for ports. Got a 10gb port? You pay a flat rate no matter its usage. 100% usage, 0% usage, same price. But you need more ports if you go over. Yeah, well a 100gb port is only 2x the price of a 10gb port. Now they have 400gb ports and soon 1tb ports.
My point was that you could sell 25/25 for $70/m with no over-subscription and 100% dedicated bandwidth for 100% of all users. That price is almost competitive with the 30/4 connections Charter sells for $60 around here, and I'm sure transferring 10.5TB in one month that a 30/4 connection can push would get you a nasty-gram. A 25/25 connection could transfer about 16TB in one month.
That's a statistical anomaly that is less than the rounding errors of the many other users using their connection to browse FaceBook and look at cat pics. In a normal large population, a single person, no matter how badly they abuse their connection, will not make a difference on the trunk.
TWC is working on a plan that you can cut $5 off of your $100/month bill if you accept a 5GB cap instead of a 250GB cap. Obviously they must think that 245GB is worth about $5, yet they want to charge $10 per 10GB for overages. They're just making up numbers. Just think about text messaging. They charge more per byte to text message than to rent time+data from the Hubble telescope or Mars Rover.
Yes, if electricity was free to generate, how much would it cost to deliver it? The logical answer is not to charge based on consuming the self-replenishing resource, but charge based on the cost to distribute it. In other words, you don't charge based on usage, you charge base on load.
Bandwidth in infinite in that it can be created without restriction, but there is only ever so much available at a given moment. As long as bandwidth being consumed is less than the rate at which it is replenished, it does not cost anything, only the infrastructure to deliver the bandwidth costs.
At slightly less than $1/mbit for dedicated bandwidth on the backbone, you could sell a 25/25 connection for $70/month($25/m for bandwidth costs and $45 for everything else), and make money even if EVERY SINGLE USER used 100% of the bandwidth 24/7 all month long.
A typical, medium-sized/30 floor apartment building can have 300 units in it. Are you going to run an OC-192 line to a single building?
No, you're going to run 300 strands of fiber for about $300/kilometer, for an overall average one time fee of about $3,000 per unit. Then each unit will get it's own 1gb fiber port which will cost about $60 and supply them with a 1gb ONT for another $150, which is all part of the $3,000. Then you pay up-stream bandwidth an electrical costs from there on out. The fiber you laid will last about 40-100 years and the 1gb connection will take about 3 years to earn itself back.
I don't understand the whole OC-48 stuff. Upstream providers don't even advertise in those units anymore, they just sell 1gb/10gb/100gb ethernet ports. Want 1gb of dedicated bandwidth from a Tier 1 ISP? $6,000 per month, that's 1,000mb.
Price for Darkfiber, about $115/mile/month. Need two fibers, say an IX is 200 miles away, that's about $50k/month. New DWDM multiplexing photonic integrated circuits are able to get 1tb/s-8tb/s, depending on the company. While I can't find pricing, they do advertise crazy numbers like "80% cheaper per mbit than previous models". Anyway, those are sunk costs, not reoccurring. So assume 1tb/s. 100gb runs about $5k/month at an IX, so another $50k. So for $100k/month, you got yourself 1tb/s of bandwidth that you can get for free from YouTube and Netflix since peering is free with them at any IX. That's about $0.10/mbit.
The industry average real-world pricing being sold per mbit from a Tier 1, is less than $1 when buying in bulk(10gb-100gb+).