Slashdot Mirror

That statement and those like it are misleading. Netflix's problem is that they were paying the cheapest "Tier 1" ISP to "dump" traffic to other "Tier 1" ISPs who refused to upgrade their side of the connections due to this asymmetric data flow. "Tier 1" ISPs want to keep traffic balanced, and upgrade links based on this.

What people are asking for is like complaining that the highway system won't build enough roads to your warehouses and that there are traffic jams into the cities you want to ship to. What do the likes of Amazon do to overcome this burden? They build their warehouses closer to where their customers are and where there are not road blocks.

Taking this back to the ISP world, that means Netflix needs to pay more money to be located on "Tier 1" ISPs which have better connections to their customers, or they need to pay their end-customer ISPs to be co-located on their networks.

This is what people forget: the Internet isn't just some network. It is a network of networks, and each ISP has their own network, and they peer (connect) at various places to exchange traffic. The top few ISPs peer settlement-free (aka "Tier 1" ISPs), and then some ISPs peer for other reasons (mostly to defer their uplink costs to the "Tier 1" ISPs). Practically all of the cable companies peer settlement-free with each other, but really because they don't compete in a single market and all that traffic passed direct bypasses the "Tier 1" ISP uplink.

Back to the warehouse analogy - whose fault is it if Amazon can't get products to their customers? Whose fault is it if they build their warehouses where there are traffic problems, or when their customer demand outpaces the roads from the warehouses?

Most people can't explain how the Internet really works with financial settlements and peering. They just want to pay the cheapest fees and get it all.

Great educational link: Internet Peering.

That's not how it works, either. There's settlement-free peering (the "real" peering) and there's paid peering and there's transit. Peering is when two networks exchange data which is destined for the respective other network. Transit is when two networks exchange traffic that is destined for some other network. Transit is the real cost factor, because nobody carries transit traffic for free: You want me to carry your data to someone else? What's in it for me? Whether peering is settlement free or not is a matter of negotiation. Most networks publish peering policies in which they describe where and with whom they will peer and what the conditions are. For example: Google, Comcast, Verizon, and few others. If you want to dig even deeper, there's a database of peerings (use guest login). It is indeed often a matter of size, and the resulting negotiating power, who pays whom. There are however "peering sluts": CDNs will typically peer settlement-free with anyone above a relatively small minimum size, even though CDNs are true behemoths on the internet. That's because their business depends on reaching everybody, and settlement-free peering is still a lot cheaper than the transit for their huge traffic flows. Netflix is in this category, for the same reason.

> It was also low bandwidth. The modern Internet is a lot of expensive to produce and deliver content.

Bandwidth costs have dropped exponentially since then.
We are looking at a 2000x drop in pricing from 1998 to the end of this year.
In 1998 it was $1200.00/Mbps by 2015 it will be $0.63/Mbps

Oh, did we forget that there was little to serve back then and few who watched? If content prices changed linearly... expectations, eyeballs, wasteage and content *weight* probably escalated exponentially. I would like to know the numbers for total MB served on media content-heavy sites. Bandwidth potentials are not the whole picture. Volume adds to it because text, standards for bitrates and general consumption hunger where probably orders of magnitude against the 2000x improvements. now you have JS taking half of your homepage loads. A single image straight off a camera can clock in 5MB when even jpegs over 0.5MB are wasteful.

I recall when 60MB for one single 20-min anime episode in low quality RM files was max goal for a night (100+MB for avi, wmv). That is all one could download per diem / night due to modem-call-vs-talk-time exclusivity logistics. Binge watching was not the problem it is with instantly-gratifying FLASH videos today.

> It was also low bandwidth. The modern Internet is a lot of expensive to produce and deliver content.

Bandwidth costs have dropped exponentially since then.
We are looking at a 2000x drop in pricing from 1998 to the end of this year.
In 1998 it was $1200.00/Mbps by 2015 it will be $0.63/Mbps

wow... I dont know why I bother. You don't understand peering... at all... Learn more here: http://drpeering.net/ Here's a very dumbed down example of what I think is going on. No I don't have proof, but this would be typical of one of these disputes... and no, peering negotiations are never nice and friendly. They more resemble bar brawls. http://i.imgur.com/uLQvkIi.jpg

Unless you're downloading games regularly, watching high def videos online, or torrenting stuff, data caps should never be a problem.

The trouble is a lot of people are now doing most of the above. People who aren't don't care about caps, since they'll never get close to 100gb a month without those three.

The problem is that the hogs of today are examples of what the average customer could be doing 10 years from now.

10 years ago, 3 Mbps down/384 kbps up DSL was widespread. Streaming video was uncommon. There was no YouTube, and Facebook was exclusive to Harvard. Windows XP SP2 was not yet released. Perhaps a 10GB cap would have been reasonable, and data hogs would pay exorbitant prices for cable.

Now, "selfie" is a thing. My chattering devices are constantly looking to communicate with Google (Android), Apple (iOS and MacOS), and Microsoft (Windows Update). I probably go right past 50GB regularly, and I don't do Netflix. This has been enabled by the relentless falling prices of Internet transit. (Historical trends) Games, Netflix, and BitTorrent aren't the only things pushing bandwidth usage up.

So, it's alarming to see a Comcast exec, proposing in 5 years to limit the Internet to the Internet of today. He wants to stop progress. It's especially galling because he's already double-dipping. He's trying to triple-dip. He's charging consumers for the cable Internet connection, he's charging Netflix for access to those consumers, and now he wants to charge again to use that connection that 2 parties have already paid for.

For anyone who actually wants to know as much as possible about the situation before they defend their position to the death Dr. Peering has a somewhat unbiased writeup about the arguments from both sides of the Netflix/Comcast debate.

My take is that the overselling of all-you-can-eat broadband will need to come to an end as Comcast (any ISP that employs the practice) simply cannot sustain an uncongested and reliable network as long as they rely on it. An unpopular idea but one that needs to be considered.

For anyone who actually wants to know as much as possible about the situation before they defend their position to the death Dr. Peering has a somewhat unbiased writeup about the arguments from both sides of the Netflix/Comcast debate.

My take is that the overselling of all-you-can-eat broadband will need to come to an end as Comcast (any ISP that employs the practice) simply cannot sustain an uncongested and reliable network as long as they rely on it. An unpopular idea but one that needs to be considered.

Let's say half of those $60 are internal cost and profit, leaving $30 for the internet connection and cost of leasing copper to the customer. Let's further assume that half of those $30 lease the copper and the remaining $15 pay for the uplink, all transit. I'll refer you to this for an idea what you get for $15. (I calculate about 6Mbps dedicated, and that is with rather generous assumptions about the other costs of an ISP. 6Mbps nonstop amounts to roughly 2TB/month.)

You're not going to get more if you don't demand more. People who are advocating volume limits and/or throttling are either ISP shills or stupid like a brick. Americans in particular are getting fleeced by the ISPs.