[heinrichhartman]

That's not likely going to happen nore is it a sensible thing to ask for.

As an ISP how would you _guarantee_ the bandwidth for each endpoint? You would have to provision for the maximal capacity all over your network. Given that most people only utilize their channel 1% of the time, this is a huge waste!

What's sensible to ask ISPs to do is: (a) Communicate historically experienced bandwith at each region (b) Provided certified, standartized measurement facilities (software / hardware?!) that can be used to monitor of the link utilization/saturation levels. (d) Refund policy, when agreed service level targets (as measured in b) were not hit. (This should be legally mandated)

[Retric]

A sensible rule would be 98th percentile so 98% of the time every customer can get X bandwidth. This allows for ~10 hours of downtime or congestion a month. They can still over provision, they just can't outright lie.

[diamondlovesyou]

Except they don't control external hosts. Cox Comm doesn't control that person hosting a server on AOL dialup.

[lagadu]

No, but they can guarantee a certain speed within the borders of their network.

[mioelnir]

Back when communication was still phoneline-connection oriented this was done all the time. You had a contractual quality target of 0.0n% of calls that either aborted or did not connect at all over the yearly average. You then over-provisioned your phone system to hit that target. That could mean adding two additional lines you never need just so new year's eve does not screw you over.

There are mathematical models that describe the probability of a new call coming in at any given time. Add the system in terms of how many connections it can have active and how many it can queue, and you can calculate your required sizing for a given quality level.

As a sidenode, this is also why ISDN flatrates were doomed, because the always-connected nature of them broke the models the system was based on. And why new phone companies renting capacity from established ones could offer cheaper connections, they simply rented at a much higher allowed connection error rate.

Using similar, well, maybe even much easier math, you can calculate that your current system at your desired maximum utilization level allows for 432KiB/s downstream for every customer, but if the overall network is underutilized you can achieve up to the n MiB/s your connection is rated for.

Then you add for example hierarchical traffic shaping where queues are allowed to borrow unused bandwidth from other queues. But it is a huge investment, no doubt.

Also, guaranteed bandwidth is imho not that different from a service level target in Bit. You'll have to refund if you break the SLA, the same as if you break your promise.

[DasIch]

At the moment at least in Germany you have ISPs advertising bandwidths and even selling contracts with bandwidth that can never be reached not because they don't have the capacity but because it's not possible to reach them given the technology in the ground.

We're currently in place that is completely unacceptable towards customers. Now you don't have to go completely in the other direction but when you advertise and sell someone a contract involving a certain amount of bandwidth, you should have to provide that 99% of the time modulo schedule downtime. I think that would be perfectly achievable and fair to ISPs and it wouldn't require them to provision for maximal capacity either. I mean it's not like you should be able to sue them into the ground when they only manage 89.9999% or something.

[TillE]

Right, they give you a discount when your max speed is below a certain level. But they really should have the data on hand to provide you with an approximate maximum speed along with their availability check before you buy.

[nitrogen]

As an ISP how would you _guarantee_ the bandwidth for each endpoint?

You guarantee that you will obtain x mbps of peering/transit per customer, per network. Then you advertise x as your minimum speed, even if off-peak usage can spike to 30 times x.

[Swizec]

> As an ISP how would you _guarantee_ the bandwidth for each endpoint?

When I still lived in Slovenia, I had 20/20 FTTH. Fiber went directly into a router in my bedroom. My bandwidth was always exactly 20/20. No matter what.

Now I have Comcast. Speedtest says I get 80/6. On Friday and Saturday evenings Netflix and Facebook and many other things often experience issues. Now I can't confirm any of this. If you run speedtest, it's fine. If you ping something, there's no packet loss. But it just doesn't feel very fast and reliable under normal use.

[revanx_]

As someone living in Slovenia and still running on ADSL2+ I'm jelly. Regarding your 20/20, highly suspect, it's likely you were doing speedtest tests against your ISPs server which would make sense but you would never get that past their gateway.

[Swizec]

I wasn't doing speedtests though, I was running torrents 24/7. High school was fun like that. Torrents were the only way to get music, movies, and TV shows back then.

[TheSpiceIsLife]

> My bandwidth was always exactly 20/20. No matter what.

This can't possibly be true. The ISP providing the service can only guarantee a particular speed to the boarder of their network.

Once you connect to a service not hosted by your ISP they can't guarantee anything.

[Swizec]

Sure, they can't guarantee speed with a particular service, but they can guarantee width and reliability of my pipe. Believe me, back then I was downloading so much crap that my internet was stuffed full at all times. I'd notice any dip in service.

And when I was on ADSL, I did notice those dips. A lot of them. With FTTH, they went away.

[Dylan16807]

There's a difference between "they guaranteed it to be that 100.000000% of the time" and "in practice, it actually was that the entire time". The latter can definitely be true.

[sliverstorm]

You would have to provision for the maximal capacity all over your network. Given that most people only utilize their channel 1% of the time, this is a huge waste!

Agreed, everybody loves the idea of fixed allocation, but never seem to consider how much that costs. If you could buy overprovisioned 25mbps @ $50/mo, or fixed allocation 25mbps @ $5,000/mo, which do you think most people would go for...

[Retric]

Everyone is playing by the same rules so they would all end up advertising 2.5 mbps @ 50$ / month with speed boost if that's what they can provide. The important point is it does not actually take much bandwidth to do streaming video etc, but if you can't provide reasonable bandwidth when people want to use it then it's false advertising.

PS: Don't forget back haul is actually a relatively minor cost for most ISP's. Until it hits ~20% it's just not going to have a major impact on peoples bills.

[tajen]

I can watch a Netflix movie in 3G at 320kmh (TGV, in France). So you're right, video consumes fewer bandwidth than we imagine. But it's common that I can't view a proper Youtube video on ADSL. ISPs are really being deceptive when they provide a connection to "the Internet".

[nordify]

It's a well known fact that certain ISPs in France will not peer with YouTube and thus their connections are congested.

This is not a technical problem, but a business decision. The at only solution is to vote with your wallet and chose an ISP that sucks less.

[rhino369]

That information is even less useful than what ISPs provide now. Home users really have no reason to value the worst case scenario speed since it never happens.

If my ISP was going to tell me either the limit they cap me at and the speed the can guarantee me at 100% use, I'd pick the former.

[Retric]

If you buy a 30MPG car your not going to actually get 30MPG as it will depend on your driving habits. However, it's based on a meaningful test so you can do meaningful comparisons.

As it stands an ISP can advertise 100 mbps service and fail to show Netflix streams when a 10 mbps connection on another ISP can easily handle 2 of them in HD. Thus, consumers need something meaningful.

PS: Picture trying to compare gas MPG if car companies could report MPG while costing down the side of a mountain. That's a perfect recipe for Honda to optimize the wrong things just like ISP's do now.

[sliverstorm]

If utilization is generally 1% of service today, you'd be looking at 0.25mbps @ $50/mo, not 2.5

[Retric]

You are assuming back haul is 100% of their current costs AND they don't increase it any AND that's 1% of peak not 1% of average including 3AM. Further, this is only limiting them to saying what their current network is.

Honestly, this is like car company's advertising their top speed when dropped from an aircraft and then complaining when they need to list actual horsepower.

[nitrogen]

That is not what transit actually costs though.

[mseebach]

Transit isn't everything you're paying for. Your local street cabinet isn't kitted out like your local datacentre and thus prices won't be the same.

[nordify]

No, but everything that's not transit is a fixed cost and thus as long as the internal network is properly designed and implemented, only transit costs matter. The rest (i.e. the fixed costs) will be covered by the monthly subscription fee.

[mseebach]

It is very expensive fixed cost that the customers need to pay the amortisation of. Also, it's only fixed as long as it's never maintained or until it needs to be upgraded. You're not going to be satisfied with 25mbit a few years from now.

[nordify]

Maintenance is a percentage of the fixed costs. Upgrades replace old equipment as they have been paid off and decommissioned, so just another fixed cost that replaces the old fixed cost.

So fixed costs all around.

[rhino369]

It would take a massive upgrade to most ISP internal networks to allow each home to use 25 mbit at the same time. There are bottlenecks that would have to replaced.

[nordify]

True, but it would still be a fixed cost to upgrade.