Hacker News Reader

Comment by jerf | Hacker News Reader

But some traffic is favored over others, by sheer network topology. Internet traffic costs money or peering agreements (which is ultimately just money that happens to cancel out). Traffic from the network you are currently connected to is much, much cheaper for the ISP.

I offer the opposite argument, which is that charging you for data sourced within the same network the same as external traffic is what would be the bizarre, anti-consumer behavior. If you want force anything on the ISPs, force them to allow others to home their services directly on their network for some reasonable fee, and then treat it also as uncharged traffic.

(I'm not sure it carries the day, so debate away. It should be put out there, though.)

View on HN

Traffic from the network you are currently connected to is much, much cheaper for the ISP.

That's not always true. Eyeball networks are often spectrum-constrained and congested on the last mile. When a DOCSIS network becomes congested, the most common relief is a node-split, and those are very expensive.

For what it's worth, often times if you rent a server, you get no charges for communicating with other servers in the same data center.

This is slightly different since the company you rent the server from is paying an ISP for internet traffic. If you don't use that bandwidth, the server company isn't charged and they pass the savings to you.

With this issue, you're still using the network bandwidth you paid for, it's still going on the internet, but its end point is a server owned by the ISP. You are not accessing a special subnet IP - you're using internet DNS and internet infrustructure to pass this "free" data. I would even bet that if there was damage to the network and there was no direct route to the server, your packets may end up routing outside of the ISPs inrastructure.

how much money does internet traffic cost?

how much cheaper is local traffic from the ISP versus traffic coming from a server 1500km away?

how much money does it cost to send internet traffic to varying distances.

there's no problem with the argument, but "much, much cheaper" doesn't mean anything in an industry where every byte can be measured and tracked.

we're not asking how much does Shaw feel like billing us today, we're asking how much does 1 Mb of data cost them to send over distance.

http://drpeering.net/white-papers/Internet-Transit-Pricing-H...

So roughly a penny a gigabyte as of 2011.

So Comcast's 250gb limit costs them $2.50?

I doubt it. This is just the cost of transmitting data from say Netflix to Comcast.

Don't forget costs such as network infrastructure, staff, etc etc.

To take an anology from the pharmaceuticals industy, the second GB costs one cent, the first one is a fair bit more expensive.

I would gladly pay per gig at that rate, and then some.

If you want to pay for the infrastructure and fixed costs too, go right ahead.

Anyone remember posting to USENET with 'rn' and you'd get a stern lecture that your post would "cost the network hundreds or thousands of dollars" are you sure you want to do this (Y/N).

Remembering, of course, that it's typically 95/5 billing. So only near-peak traffic matters, but it matters a lot.

There is a lot more complication and loophole-ability in forcing ISPs to allow others to home their services than there is in forcing net neutrality in its traditional definition.

That's not how the Internet business model works. Receiving data is effectively free. Transit providers charge the sender for the bitrate (high for the web host, low for the web user) and a fixed fee for the connection, which you will have to pay regardless. Peering is paid for per connection and often by the web hosts who want to avoid transit and get better latency to the ISPs customers. In terms of traffic, there is effectively no difference in cost between sending traffic internally or receiving it from an external source. The reason for data caps is that ultimately networks are capacity constrained, but mostly at the last mile, not at the peering and transit links.

This is totally false. Inter-AS traffic is often priced using bandwidth in peered connections. In a bandwidth discrepency situation one isp will usually have to pay a lot more for the inter-AS traffic.

Right, and that's the sender, which is what I said. Note that the amount of traffic is effectively irrelevant since the connection is supposed to be dedicated. The only reason why the amount of traffic matters is that you need to decide which side pays what.

No. Generally the agreement works like this: "we'll set up a peering agreement between us. As long as the traffic differential is within 3%, peering is free. Beyond that point, the exceeding party will pay an excess penalty of x/GB."

Reply on news.ycombinator.com