Hacker News Reader

The real question is whether Verizon is saying "Use the other bridges" or "If you pay us we can open another bridge for you". What I've read mostly points to the latter and Netflix is either taking a stand against this practice or Verizon wants to much money.

> Realistically it's absurd that Netflix isn't use a whole variety of ISPs to manage the load themselves. Other big data providers do it. Why can't they?

IIRC, Netflix offered to provide hardware to all the major ISPs that would cache Netflix content at the ISP data centers. Not a single one took them up on the offer.

That was the Open Connect Hardware. Free big storage boxes for all ISP's who wanted them.

They also offered to change their entire distribution model to P2P as-to keep the traffic burden inside the consumer isp's network (which effectively costs them nothing, it's ingress traffic that is the major costs).

Level3 has even offered to split the costs of upgrading the Big5 ISP's peering connections.

All options were refused.

> They also offered to change their entire distribution model to P2P as-to keep the traffic burden inside the consumer isp's network (which effectively costs them nothing, it's ingress traffic that is the major costs).

Why wouldn't they do this anyway? How does the ISP have any say in whether or not Netflix distributes their content via P2P? Or am I misunderstanding?

P2P would make the service quality depend on how many people were online, etc... I think they'd rather sort it out another way.

But regardless, even with that offer, Comcast scoffed at it, citing Netflix would still be a "large burden" to their network. (which would be even more-so false than it currently is).

But they weren't going to pay for rack space and power. Also, IMO offering free rackspace violates net neutrality since a start up couldn't get a similar deal.

Netflix could do something differently, the question is really whose responsibility is it? The fact that the data gets to the edge of Verizon's network uncontested means the congestion is on Verizon's network. The fact that the traffic was requested by their members, and then delivered uncongested to them tells me it's on Verizon to fix the issue.

In an ideal world Netflix would pay a single ISP for access to "the internet" and their traffic would be handled properly across peered networks. Unfortunately it's not an ideal world, and so Netflix has made a bunch of philosophical compromises to increase service for its customers (including setting up other Fast Lanes). But that doesn't mean we should expect Netflix to continue to do this.

And Verizon says that they have plenty of open peering points and it's Netflix's responsibility to deliver it in a responsible manner.

>In an ideal world Netflix would pay a single ISP for access to "the internet" and their traffic would be handled properly across peered networks.

Netflix can do that now, but its not what they are doing. It's paying an ISP to send it over only their peering points by choosing the lowest bidder. It saves Netflix on bandwidth costs. A normal ISP will send the data by any means to get it there. Netflix's ISP is just letting their saturated points sit there. That's the ISP's fault.

>The fact that the traffic was requested by their members, and then delivered uncongested to them tells me it's on Verizon to fix the issue.

You could easily make the argument that Netflix is selling their service and is the one sending the data. It should be on their ISP.

> Verizon is saying, use a different bridge they are all open.

You're assuming Verizon isn't over-subscribing all of their other peering connections too.

> Both are valid arguments. However, one thing to consider is that Netflix could switch ISPS tomorrow and start dumping their load on a different bridge.

Your analogy is broken because peering connections aren't like bridges. Expanding a bridge costs millions of dollars and can't reasonably be moved from where you build it. 10Gbps switch ports do not cost millions of dollars and can trivially be moved to a different peer if loads change.

According to Level3 - a 10Gbps link bears only a one-time cost of $10K-$15K USD.[1](see comments)

Chump change for major ISP's.

[1] http://blog.level3.com/global-connectivity/chicken-game-play...

>You're assuming Verizon isn't over-subscribing all of their other peering connections too.

The PR release linked as the topic of this thread claims it's only 44% on the other connections.

>Your analogy is broken because peering connections aren't like bridges. Expanding a bridge costs millions of dollars and can't reasonably be moved from where you build it. 10Gbps switch ports do not cost millions of dollars and can trivially be moved to a different peer if loads change.

There isn't just one peering ports there are many many peering points.

It's also cheap for Netflix to use another ISP who won't cram it down saturated peering points.

Finally, if it's so cheap why won't LEVEL3 or Cogent pay 100% of the cost?

> The PR release linked as the topic of this thread claims it's only 44% on the other connections.

Percentages don't tell you anything. The fact that some 10Mbps link is at 23% utilization is of no utility. Neither is the capacity of links not connected to the relevant customers. The real question is, if Netflix used some other transit provider and their traffic went over some other Verizon links, wouldn't those links still get saturated?

> There isn't just one peering ports there are many many peering points.

...and?

> It's also cheap for Netflix to use another ISP who won't cram it down saturated peering points.

If that were true then there would be no dispute.