Netflix Launches Its Own Content Delivery Network

1sockchuck writes "Netflix has launched its own content delivery network to manage data delivery for its streaming video service. ISPs can choose to host caching appliances in their data centers, or peer with Netflix at Internet Exchanges. 'Netflix will provide either form of access at no cost to the ISP,' it said. As part of Open Connect, Netflix is sharing its hardware appliance design and the open source software components of the server. Does this mean Armageddon for the CDNs currently serving Akamai? Not really, according to analysts, citing the leverage Netflix had in dealing with providers."

Comments

What about Comcast?

[Anonymous Coward]

So Comcast is going to roll over for this? Comcast is an integrated Cable TV/ISP, which wants to favor their own delivery mechanism and content relationships. Only Common Carrier status for Internet delivery will break that stranglehold - lots of luck for achieving that in the USA!

- Leonard

Re:What about Comcast?

[TubeSteak]

If Netflix can get its hardware inside Comcast's network, does that mean Comcast won't count it against their data caps?

Re:What about Comcast?

[Bengie]

The actual problem is Comcast's congestion is in the last mile. No amount of ISP caching will reduce last-mile congestion.

Re:What about Comcast?

[dgatwood]

And the vast majority of those companies were given money by the government because for many years, every company they asked to serve those areas responded, "Hell, no". The free market works well when there are enough customers per square mile to make competition feasible. In more rural areas, true competition can't possibly work, which means the only viable alternative is a government-run last-mile infrastructure that leases access out to multiple competing ISPs.

Re:What about Comcast?

[dgatwood]

I think we can safely say what would have happened. Almost nothing. In the U.S., somewhere around a quarter to one fifth of the people are spread out across approximately 97% of the land mass, and the other 70-75% are concentrated in the remaining 3% of our land mass. Even with the best wireless technology we have at our disposal, covering much of the U.S. is infeasible because of geography (mountains, etc.). And the cellular phone industry has doubtless poured more money into wireless research than the total spending on ISPs nationwide.

Let's take my hometown in Tennessee as an example. Many of my friends did not have cable service because even with subsidies, it was not profitable. It costs about $30,000 to run a mile of coax or fiber, and in many places, that would serve only one or two households. Even if they stood to make $30 profit per month (unlikely), it would take over 40 years to break even, without factoring in such pesky things as interest.

So what about wireless? Realistically, most wireless Internet services cap out at somewhere on the order of ten miles or so. Assuming you placed a cellular tower ($150,000 or so) in a location where it could serve the areas between towns, there are many areas where a tower with a ten mile radius would serve only a low four-digit number of households). Many of these areas do not even have basic cellular phone service today from any carrier.

Given that probably 90% of those four-digits worth of people live in a town (and thus would likely already be served by a wired Internet provider because it would be profitable to do so, subsidies or not), you're talking about almost half a grand per household served. And that's just for the initial tower construction costs. On top of that, you have to add the cost of a trunk line out to the middle of nowhere (at $30,000 per mile times 10+ miles to the nearest town), plus hundreds of dollars in customer premises equipment costs for each household (that many of those households could not realistically afford). Granted, $2,000 per customer is a far cry from $15,000, but it is still something that no sane person would invest in.

Even if you could miraculously crank up the radius up to 30 miles, it would barely be profitable, and short of insanely tall towers, that's about where the curvature of the earth itself will bite you in the you-know-what.

And the bigger problem is that all this proposed infrastructure is unlikely to pay for itself before the technology becomes obsolete.

In short, there is just no feasible way to solve the last-mile problem except for either A. the government forcing private enterprise to build out the wired infrastructure in exchange for the right to serve other, more profitable areas or B. the government building the infrastructure itself. Your viable choices are basically the system we have now or socialism. Take your pick.

Re:What about Comcast?

[Bengie]

If you look at the complaints of Comcast customers, it's that "the internet is slow and I'm getting packetloss" during peak and they don't use P2P or anything hoggish. But you will notice that these people also mention they live in high density areas like large apartment complexes. You will also notice other people claim to be using Comcast in the same city and don't have any issues at all.

During the whole L3 vs Comcast issue, L3 requested 270Gb of additional peering bandwidth. Remember, L3 is a tier1 back-bone. This means Comcast and send data to L3 and L3 will route that data to anywhere in the world. Being that Comcast and L3 have a peering agreement, it effectively means Comcast gets its internet for FREE. When L3 comes to you and offers you 270Gb of free bandwidth and you turn it down, that means you don't have an issue at your trunk. That's enough bandwidth for almost 70k 4Mb data streams

If Comcast doesn't have have congestion at their trunk, then the only other place is their last mile or their middle mile(or whatever it's called). When you own your own network or lease fiber, upgrading the "middle mile" is almost free. The logical conclusion is that the last mile is the bottle-neck.

Re:What about Comcast?

[Jason Levine]

I know that when I was there in the mid 00s there were places in downtown nashville without cable OR DSL because they were both too fucking greedy to spend a dime on infrastructure while the places that did have it were oversubscibed horribly.

Of course, whenever the issue of those under-served towns getting municipal broadband comes up, the big ISPs like Comcast suddenly become very interested in developing there. Not enough to actually develop, mind you, but enough to lobby to squash the project on the grounds that it would be unfair competition. You know, should they ever decide to build there.

Re:OMG! A caching server!

[DarwinSurvivor]

I think the real story here (which the summary completely missed) is that Netflix is not just setting up caching mirrors, they are trying to get the competitors to host the caching servers! Most ISP's in the US (which is about the only place netflix works anyways) are the same companies that have been trying to destroy netflix to save their cable-TV interests.

awesome!

[asshole felcher]

Now they just need some content to deliver.

FreeBSD

[Jon Stone]

All these posts and no-one has mentioned it runs on FreeBSD?

Netflix's New Peering Appliance Uses FreeBSD