The customers are paying for the service. They bear the cost.
The idea that ISPs bear the cost of Internet traffic which benefits tech companies is at the centre of their agreement against net neutrality, and it's a fallacious argument.
The ISP is free to decide the price and the quality of the service offered. That's how they recoup their costs, by charging their customers the correct price for the service.
Maybe I was not clear enough, I was not defending ISPs. What I meant was that they will bear the extra cost if every subscriber maxes out their bandwidth usage so the cost (and prices therefore) will go up.
What customers pay is an averaged out bandwidth usage from my understanding.
The customer has already paid for a dedicated bandwidth. If the ISP has used an outdated model that doesn't account for current usage, that's their problem. That's how business risk and models work.
They can, and should, respond by modifying their plans accordingly. Once the consumer's contract is up, renew them with a more modern usage model. Until then, you eat the cost for not doing your only job.
The theory that "we didn't build in a margin of error and therefore someone else is responsible" is dangerous to society.
ISPs SHOULD see losses. Their investors aren't entitled to guaranteed profits.
But if everyone uses their connection in inefficient ways (such as letting their video streaming run for hours unwatched) everyone will simply end up paying more at some point.
Only in the United States, because of the lack of ISP competition.
In Europe, as usage goes up, prices stays flat, because ISPs there know that the cost of a marginal bit of data is almost nothing, and they actually have to compete with each other on customer service and price.
The idea that everyone should simply pay the same regardless of use is wrong.
Cell phone plans have data tiers and pay-per-use. There's no reason cable companies can't do this either, except it prevents them from fudging costs.
I probably use as much bandwidth as anybody else, and I'd prefer to be accurately charged for it rather than have my provider pull shady net neutrality things or try to become an ad network or buy a media company to "bundle" their services or sell my data to third parties to juice their income.
I don't necessarily disagree but go back to any of the discussions on here about data caps and read all the screams about stifling innovation, etc., etc.
I see both sides. Honestly, I think there was a better argument for it when the heavy users were a small percentage of users downloading a bunch of "Linux distributions" on bittorrent. Today, it's more likely to be a family that uses Netflix a lot.
Metering would change usage for better or worse and the fact is that a lot of domestic phone plans don't have caps, or at least meaningful ones any longer, although they may throttle data at some point.
I have often wonder why Netflix doesn't uses its Fast.com to offer CDN services.
They could also have offer their Encoding System as Services. I upload a video file, sets a few requirements and I get Netflix Quality encoding in return without me fiddling anything.
The answer to both of those is "because Netflix is a video delivery service, not an infrastructure service". It would be a whole new skillset for the company to offer such a service. Running those things, and selling them, are two very different things.
I guess so, It is just a nice idea. It was one day I notice how Fast the response time were in Fast.com compared to others. I am not even sure if Google Edge Node in the ISP even works with the GCP CDN, because their performance don't seems to be leading the pack.
The Market for CDN and Online Video Encoding that Netflix could get are properly in the sub annual $1B range, may be too small for Netflix to even consider.
I just thought the more traffic passing through those FreeBSD Server the better.
It's only incompatible when ISPs actually act on their natural preference for consumption to go towards in-house edge caches. It's not that much different from the ancient art of configuring http servers and caching proxies to get along well (back even caching proxies were still a thing).
So Akamai, CloudFlare et al. will locate servers in a large ISPs data centre. TIL, I just assumed once a resource had been cached then they'd effectively have that collocation (logically); I'm surprised if avoiding the cost of the first hit really saves anything for static content.
Obviously anyone streaming video, it's different.
I wonder if BBC iPlayer (major streaming service in UK) collocate with ISPs.
Will those servers have a local (to the ISP) IP, or do the Akamai IPs get physically routed to the AS?
It is a CDN, but what's new is that the openconnect appliances are distributed not just in datacenters, but directly to ISPs. Traditional CDNs aren't in your ISP.
Do Akamai has Servers right in ISP's DC / Network? I know they are within 1 hoops in lots of places, but never heard they are actually inside ISP's network, which is what Netflix OpenAppliance and Google Edge Node has done.
There's basically 2 costs to running a network, buying routers/switches/laying cable, and electricity.
The only thing that matters is peak usage because then you have to buy faster network equipment. If I buy 100MBit internet from my ISP, they basically need to spend, once, whatever 100Mbit of a port looks like, 1/10th of a gigabit port, 1/100th of a 10G port etc. They only have to do that once, yet I pay them every month, they're fine.
You’re only seeing part of the picture there. Hardware is approximately a one time cost, if you ignore upgrades and maintenance, and electricity is a consideration, but the other part is bandwidth to get data to that port.
ISPs also need a way to get your packets off their network and across to other people’s networks, which are possibly on the other side of the globe. A lot of this is done via peering agreements with other providers, where both parties will agree to connect their networks at a common data centre without charging for data going across that link, but there will still be some data that needs to go across transit providers - these are the companies providing things like transatlantic fibre, or interconnects between different parts of the US. They charge for a pre-agreed amount of bandwidth, and when you burst above that it gets expensive.
These agreements are the core of the net neutrality debate. Traditionally peering agreements would see both parties having a roughly equal amount of traffic going back and forth, however in a world with services like Netflix what you see is a huge amount of bandwidth being used on side of the link. This is why ISPs think Netflix and co should have to pick up the cost of those links.
I don’t agree with that stance since the reason ISPs have customers is often the ability to access those services. However, it’s not quite as clear cut as people make out.
But the same rule that basic cost equals hardware + setup + energy + maintenance also applies to the upstream provider, and their upstream provider, all the way along the selected route including the biggest interchange touched. The economic arrangements are different, but the cost base is just that.
Sure, and they purchase very expensive equipment to cover the peak bandwidth demand then spread out those capital costs in the form of a monthly bill.
If you want more bandwidth, it's "free" as long as there is room available, otherwise the whole chain needs to upgrade so as not to degrade the service of other customers. On a competitive market, the margins available and the prices should tend to go as tight as possible and the economics will lead to pay for usage.
They bear the cost to build the infrastructure, but their cost to deliver data is negligible. As long as their infrastructure can handle the maximum capacity spikes, they're basically printing money.
