Cable providers spend more each year just in upgrades to their copper network than the cost of upgrading the same number of customer to all fiber. It's not an issue of money. They have a good thing going and unless there is a reason to go fiber, they don't want to rock the boat. "Don't fix what ain't broken". The only reason they are not going fiber is because of the lack of competition. Google comes in with 1gb for $70 and they freak out. I'm sure Google could easily offer 1gb for $100 if they couldn't get all of the help from the city.
I'm not sure exactly what you meant be "contended", but an ISP should never oversubscribe their last mile to the point of congestion. Modern last mile networks, even if contented, should have no congestion as long as they're not horribly oversubscribed.
This is probably not the entire reason, but ISPs were already complaining about the 4mb HD streams, yet alone Super HD at 8mb. Netflix wanted to wait for A and B to get widely used before going to open up Super HD to all because it does reduce bandwidth costs for Netflix, and Netflix set a standard that they can say "these other ISPs are doing just fine with our OpenConnect services", in the case that ISPs were to complain en masse. I'm sure if Netflix didn't have OpenConnect pushed to most ISPs and they offered Super HD, almost every ISP across the USA would become one voice as their networks blew-up over night.
CDNs paid ISPs because the demand for cheaper bandwidth was in high demand by service providers, so 3rd party CDN service cropped up to provide these services. They worked as a middle-man to quickly negotiate lower costs for bandwidth.
At this point, Netflix is not in a mad rush to lower costs at the expense of giving in to ISP demands and are willing to play chicken with ISPs. With normal CDN services, a single service provider did not have the negotiating power and CDN providers knew this. It is different with Netflix. They are huge and they have the means. If ISPs want to play hard-ball, Netflix is willing to use transit, but the ISPs don't want to use transit.
ISPs are complaining that Netflix traffic costs too much, so Netflix offers them a way to reduce their bills by peering to an IX CDN or a local CDN. The ISPs don't want none of that, they're used to getting paid for CDN services. They are just used to being in a position of power, but this time Netflix is on near equal footing and these giants are butting heads. Netflix is willing to meet ISPs half way and treat them as equals, but ISPs don't want to be equal, they want to resell access to their customers and act as gatekeepers to double-dip.
The main point is customers have already paid their fair share of bandwidth. With current prices, ISPs should be able to give all users dedicated bandwidth just based on transit costs alone, ignoring the huge reduction in costs for ISPs when peering is involved. For $5,000/month, you can get 100gb/s of bandwidth at your local IX to peer with Netflix and YouTube, which represent about 50% of all bandwidth usage. So ISPs can get 1/2 of their peak bandwidth usage at a rate of $0.05/mbit, then they turn around and charge you $80 for a 50mb connection and bitch when you try to use 4mb/s for a few hours. How is that fair?
Obviously this does not represent all ISPs, especially small rural ISPs, but it does represent most ISPs that are within a few hundred miles of a major IX.
Netflix still has to use transit providers for customers who's ISPs do not peer with Netflix at an IX or use an internal Netflix CDN. I also question what you mean by "across the Internet". I've gotten Netflix data from across the USA in the past. Midwest to Washington over the exact same route I get when trying to connect to any other Washington service for stuff like Mumble or Minecraft servers. I now get my Netflix data over Level 3 to a Cogent datacenter in Chicago. That is a regional CDN, but it still uses transit to get to me.
Your circuit wasn't that "saturated". They use TDMA with a fair scheduler. If you were getting less than (Total bandwidth)/(number of customers on port), there was something miss-configured. Newer GPON units have such a good TDMA bandwidth scheduler, that they can emulate T3 over a fully loaded link. That requires some tight timings. With a properly configured GPON system, you should not get latency, even when saturated, and you should have a lower bound of around 79mb/s of raw bandwidth, minus a bit for inefficiencies.
GPON supports SLAs, so you can put a minimum guarantee on both latency and bandwidth, while allowing users to consume excess idle bandwidth. Or do what my ISP does and only sell what you have. They give you a reserved slice of bandwidth, which means GPON will make sure you always get your bandwidth, no matter what. The question is if the upstream routes can handle it.
I fully agree, but they could also just design their network to not have congestion at the "local" level. I think this was estimated to cost about 1% more to the entire cost of a new fiber network.
We're talking about backbone links, so very high speeds. QoS is a bad thing here. It comes down to this. Would you rather have a 100gb link with QoS or a 400gb link with no QoS. They're the same price.
Business class may have higher priority when going over the same internal link. If that link is congested, the residential users will feel it, but as long as the total number of business bandwidth doesn't go past link capacity, they won't feel it.
Last and middle mile can cheaply be designed with a 1:1 provisioned to available bandwidth, all the way up to 1gb/1gb. The only issue is the trunk, but with a large population, peak usage from day-to-day is almost identical. The entire Internet is designed around the network being idle 95% of the time. You target peak usage. It's a good thing that peak usage is quite stable and easily predictable.
In my many years of owning cars, nearly every one of their speedometers has been +-(3-5mph) off. Older cars worse than newer ones. Some of them were accurate at lower speeds around 25, but then even up to 10mph off at the high end. It seemed common enough that I assumed there has to be some room for error.
Not where I live in the USA. Business and Residential get the same service and the same options all with dedicated bandwidth also with dedicated fiber. Now if you want enterprise grade, you can get that also. There are some draw-backs to the non-enterprise grade fiber connections. You don't get that one-on-one interaction for any and every change to your network. I've had a few times where a mass upgrade caused a hick-up, but their network has been over-all great.
I recently had an issue with my fiber port that they were trying to diagnose and they had to switch me to another port. They said this was all done back at the CO, so it only took the guy on the phone about 5 minutes to get up from his desk and change my port. I spent more time waiting for their system to register that my ONT changed locations. If I had an enterprise account, I probably would have had little to no down-time.
Sabatier reactors take CO2 and hydrogen as input. So to do this, you first need to split water to create hydrogen, then you need to run it through yet another inefficient process to combine it into methane? Sounds like a horrible way to "store" energy. Once all is said and done, one would hope to at least break even after all of that waste.
Probably get a better ROI with huge banks of lead-acid batteries.
Current nuclear power designs are based on the old designs which were designs to produce dangerous stuff meant to be used by the military. If they wanted a safe nuclear power option, they would have went with some of the other options. Safety and health was never a primary concern. Newer nuclear power designs have less nuclear waste than coal power plants.
Letting links degrade isn't illegal. Throttling is. If links getting degraded was illegal, then getting congestion would be illegal. A link is degraded when it can no longer handle its peak load, yet you see this happen all the time. Verizon is just a bit more ballsy about doing it with high profile routes in an attempt to get more money out of other companies.
Throttling can be done passive aggressively and there is no law against that, that I know of. Just let links degrade, then give business users higher priority.
For personal reasons, he generally does not actively browse the web from his computer; rather, he uses wget and reads the fetched pages from his e-mail mailbox, claiming to limit direct access via browsers to a few sites such as his own or those related to his work with GNU and the FSF
Yet research has shown that paying people who do creative jobs, more money, makes them worse performers. Once you get past "being content", performance goes down for jobs that involve any amount of creative thinking. In this way, the free market does not work.
A great reason to use VLANs to enforce guaranteed bandwidth to the phones and separate them from the public Internet, at least within the ISP. The backbone will need to have special routes to maintain separation, but VOIP data is much much less than other traffic.
The original argument was that even with a UPS at home, if the middle-mile loses power, you're SOL. My counter argument was that fiber does not have this problem, you only need to power the both ends, not everything in-between.
Someone has to program the automation and that automation will have limits, and someone will have to program a better version. Anyway, someone needs to be able to describe a problem for a computer to automatically program it, and at that point, that is the new programming.
I bet the sales person would kick your ass at selling stuff. I'm great at programming, but I don't do around judging people based on how well they can program, just that they're good at something and they probably do that something better than me.
Hats off to you, sales person. The only sales people I've worked with have absolutely no understanding of programming. Need a feature added? Sure, our programmers can magically add that and have it ready by tomorrow.
Programming is the application of knowledge and encompasses all sciences. In order to program, you must be able to look at a problem, understand it, and break it down into its atomic pieces, the same type of thinking that must be done for all problems in any system.
Slashdot Mirror
Cable providers spend more each year just in upgrades to their copper network than the cost of upgrading the same number of customer to all fiber. It's not an issue of money. They have a good thing going and unless there is a reason to go fiber, they don't want to rock the boat. "Don't fix what ain't broken". The only reason they are not going fiber is because of the lack of competition. Google comes in with 1gb for $70 and they freak out. I'm sure Google could easily offer 1gb for $100 if they couldn't get all of the help from the city.
I'm not sure exactly what you meant be "contended", but an ISP should never oversubscribe their last mile to the point of congestion. Modern last mile networks, even if contented, should have no congestion as long as they're not horribly oversubscribed.
This is probably not the entire reason, but ISPs were already complaining about the 4mb HD streams, yet alone Super HD at 8mb. Netflix wanted to wait for A and B to get widely used before going to open up Super HD to all because it does reduce bandwidth costs for Netflix, and Netflix set a standard that they can say "these other ISPs are doing just fine with our OpenConnect services", in the case that ISPs were to complain en masse. I'm sure if Netflix didn't have OpenConnect pushed to most ISPs and they offered Super HD, almost every ISP across the USA would become one voice as their networks blew-up over night.
CDNs paid ISPs because the demand for cheaper bandwidth was in high demand by service providers, so 3rd party CDN service cropped up to provide these services. They worked as a middle-man to quickly negotiate lower costs for bandwidth.
At this point, Netflix is not in a mad rush to lower costs at the expense of giving in to ISP demands and are willing to play chicken with ISPs. With normal CDN services, a single service provider did not have the negotiating power and CDN providers knew this. It is different with Netflix. They are huge and they have the means. If ISPs want to play hard-ball, Netflix is willing to use transit, but the ISPs don't want to use transit.
ISPs are complaining that Netflix traffic costs too much, so Netflix offers them a way to reduce their bills by peering to an IX CDN or a local CDN. The ISPs don't want none of that, they're used to getting paid for CDN services. They are just used to being in a position of power, but this time Netflix is on near equal footing and these giants are butting heads. Netflix is willing to meet ISPs half way and treat them as equals, but ISPs don't want to be equal, they want to resell access to their customers and act as gatekeepers to double-dip.
The main point is customers have already paid their fair share of bandwidth. With current prices, ISPs should be able to give all users dedicated bandwidth just based on transit costs alone, ignoring the huge reduction in costs for ISPs when peering is involved. For $5,000/month, you can get 100gb/s of bandwidth at your local IX to peer with Netflix and YouTube, which represent about 50% of all bandwidth usage. So ISPs can get 1/2 of their peak bandwidth usage at a rate of $0.05/mbit, then they turn around and charge you $80 for a 50mb connection and bitch when you try to use 4mb/s for a few hours. How is that fair?
Obviously this does not represent all ISPs, especially small rural ISPs, but it does represent most ISPs that are within a few hundred miles of a major IX.
Netflix still has to use transit providers for customers who's ISPs do not peer with Netflix at an IX or use an internal Netflix CDN. I also question what you mean by "across the Internet". I've gotten Netflix data from across the USA in the past. Midwest to Washington over the exact same route I get when trying to connect to any other Washington service for stuff like Mumble or Minecraft servers. I now get my Netflix data over Level 3 to a Cogent datacenter in Chicago. That is a regional CDN, but it still uses transit to get to me.
Your circuit wasn't that "saturated". They use TDMA with a fair scheduler. If you were getting less than (Total bandwidth)/(number of customers on port), there was something miss-configured. Newer GPON units have such a good TDMA bandwidth scheduler, that they can emulate T3 over a fully loaded link. That requires some tight timings. With a properly configured GPON system, you should not get latency, even when saturated, and you should have a lower bound of around 79mb/s of raw bandwidth, minus a bit for inefficiencies.
GPON supports SLAs, so you can put a minimum guarantee on both latency and bandwidth, while allowing users to consume excess idle bandwidth. Or do what my ISP does and only sell what you have. They give you a reserved slice of bandwidth, which means GPON will make sure you always get your bandwidth, no matter what. The question is if the upstream routes can handle it.
Anyone that deletes evidence should be considered admitting that they did commit whatever they're being charged with, and with malicious intent.
I fully agree, but they could also just design their network to not have congestion at the "local" level. I think this was estimated to cost about 1% more to the entire cost of a new fiber network.
We're talking about backbone links, so very high speeds. QoS is a bad thing here. It comes down to this. Would you rather have a 100gb link with QoS or a 400gb link with no QoS. They're the same price.
Business class may have higher priority when going over the same internal link. If that link is congested, the residential users will feel it, but as long as the total number of business bandwidth doesn't go past link capacity, they won't feel it.
Last and middle mile can cheaply be designed with a 1:1 provisioned to available bandwidth, all the way up to 1gb/1gb. The only issue is the trunk, but with a large population, peak usage from day-to-day is almost identical. The entire Internet is designed around the network being idle 95% of the time. You target peak usage. It's a good thing that peak usage is quite stable and easily predictable.
In my many years of owning cars, nearly every one of their speedometers has been +-(3-5mph) off. Older cars worse than newer ones. Some of them were accurate at lower speeds around 25, but then even up to 10mph off at the high end. It seemed common enough that I assumed there has to be some room for error.
Not where I live in the USA. Business and Residential get the same service and the same options all with dedicated bandwidth also with dedicated fiber. Now if you want enterprise grade, you can get that also. There are some draw-backs to the non-enterprise grade fiber connections. You don't get that one-on-one interaction for any and every change to your network. I've had a few times where a mass upgrade caused a hick-up, but their network has been over-all great.
I recently had an issue with my fiber port that they were trying to diagnose and they had to switch me to another port. They said this was all done back at the CO, so it only took the guy on the phone about 5 minutes to get up from his desk and change my port. I spent more time waiting for their system to register that my ONT changed locations. If I had an enterprise account, I probably would have had little to no down-time.
Sabatier reactors take CO2 and hydrogen as input. So to do this, you first need to split water to create hydrogen, then you need to run it through yet another inefficient process to combine it into methane? Sounds like a horrible way to "store" energy. Once all is said and done, one would hope to at least break even after all of that waste.
Probably get a better ROI with huge banks of lead-acid batteries.
Current nuclear power designs are based on the old designs which were designs to produce dangerous stuff meant to be used by the military. If they wanted a safe nuclear power option, they would have went with some of the other options. Safety and health was never a primary concern. Newer nuclear power designs have less nuclear waste than coal power plants.
Letting links degrade isn't illegal. Throttling is. If links getting degraded was illegal, then getting congestion would be illegal. A link is degraded when it can no longer handle its peak load, yet you see this happen all the time. Verizon is just a bit more ballsy about doing it with high profile routes in an attempt to get more money out of other companies.
Throttling can be done passive aggressively and there is no law against that, that I know of. Just let links degrade, then give business users higher priority.
For personal reasons, he generally does not actively browse the web from his computer; rather, he uses wget and reads the fetched pages from his e-mail mailbox, claiming to limit direct access via browsers to a few sites such as his own or those related to his work with GNU and the FSF
He's an "Amish" programmer?
Yet research has shown that paying people who do creative jobs, more money, makes them worse performers. Once you get past "being content", performance goes down for jobs that involve any amount of creative thinking. In this way, the free market does not work.
A great reason to use VLANs to enforce guaranteed bandwidth to the phones and separate them from the public Internet, at least within the ISP. The backbone will need to have special routes to maintain separation, but VOIP data is much much less than other traffic.
The original argument was that even with a UPS at home, if the middle-mile loses power, you're SOL. My counter argument was that fiber does not have this problem, you only need to power the both ends, not everything in-between.
Someone has to program the automation and that automation will have limits, and someone will have to program a better version. Anyway, someone needs to be able to describe a problem for a computer to automatically program it, and at that point, that is the new programming.
I bet the sales person would kick your ass at selling stuff. I'm great at programming, but I don't do around judging people based on how well they can program, just that they're good at something and they probably do that something better than me.
Hats off to you, sales person. The only sales people I've worked with have absolutely no understanding of programming. Need a feature added? Sure, our programmers can magically add that and have it ready by tomorrow.
Programming is the application of knowledge and encompasses all sciences. In order to program, you must be able to look at a problem, understand it, and break it down into its atomic pieces, the same type of thinking that must be done for all problems in any system.