The article isn't wrong, but it's a new situation when content providers start paying to peer explicitly. Netflix and other large players already pay for CDN structures to get content close to the last mile. Now they are going to be paying for the last mile in some sense also.
Fine for Netflix to do, but if prices rise and the thresholds where this applies lower, new services will get shut down due to cost.
Hopefully some ISPs will not charge and market "fast access to whatever you want" and not "fast access to whatever we chose based on what they paid").
The entire point of the Netflix/Comcast paid peering arrangement was that Netflix didn't want to pay a "CDN structures to get content close to the last mile."
If they had paid a CDN for content delivery to Comcast's network (as thousands of other companies do), then none of this would have ever developed.
Netflix was just trying to find the cheapest possible way of getting the content that they deliver, to the eyeballs that were willing to pay for it.
The only way that "ISPs will not charge" is if we manage to get broadband competition in enough markets in the United States so that Quality of Service becomes a differentiation.
Yes Netflix offered that, and that's what the Comcast/Netflix paid-peering arrangement accomplishes, albeit in a slightly different manner than if Netflix OpenConnect Appliance had been dropped into a Comcast data center cage.
In order to connect to Comcast's network you need port capacity on Comcast's routers. Netflix pushes approx. 3 Terabits [1] onto the Comcast network, so they need Comcast to supply (configure, support, monitor, upgrade, power, etc...) approx. 300 10-GigE interfaces, as well as any incremental Line Cards, Routers, Cabinets that might be required to add that port capacity and bandwidth capability to their network. (3 terabits is enough bandwidth to get your attention)
In addition, now that there is an extra 3 terabits of incoming data onto their network, Comcast also has to scale their outgoing network (to the home users) by another 3 terabits as well, so there are incremental costs there. It could be argued (and I would do so) - that those 3 terabits have already been paid for by the cable-subscribers, and so Comcast should be prepared to provide sufficient capacity to deliver it - but regardless, it's an additional cost that Comcast needs to account for, regardless of whether they are installing appliances in their network, or peering with content providers.
Comcast provides these peering locations in at least 17 discrete locations in the United States [2] - Many of those locations are just general CoLos - I've heard of about 25% on that list - so many of them aren't "Dedicated Comcast Data Centers" - they are CoLos in which multiple customers can come together and interconnect with each other or carriers such as Level3, Cogent, etc....
So, in those facilities, from Comcast's perspective, providing Port Capacity is a matter of connecting a fiber patch cable to one of their 10 GigE interfaces, and exchanging some BGP routes with a content provider.
It's true that there are several routing relationships Comcast can establish with content providers, and some of those relationships might involve placing those Content Providers in AS7922 (Comcast's Autonomous System) - but, from a cost perspective, there are a lot of costs that are duplicated, regardless of whether Comcast establishes a peering relationship with Netflix, versus having them route from within Comcast networks.
You still need to install 300+ 10 GigE ports and all the associated costs that go along with that, and you still have to provide downstream capacity of 3 Terabits to home subscribers. Peering with a partner in a 3rd party CoLo instead of having to provide them cabinets/electrical/etc for the appliances avoids those cabinet/appliance costs.
Something to keep in mind, is that in addition to providing the three terabits streaming capacity on their network, Comcast would also have to provide Capacity to keep those Appliances "supplied" - each of the appliances would need about five Terabytes of content loaded each night [3, pg 13]. There are also some operational advantages to managing a Content Provider at "arms-length" distance with BGP, versus having to do custom routing on your own network to support them. Comcast already knows how to peer and interoperate with many carriers/partners using BGP, so no new procedures/technology required to add Netflix to that list.
In the end - Netflix and Comcast each negotiated in their own best interests. Netflix's access to the Comcast network was provided in roughly the same way that Content Providers have been getting access to networks for the last 10 years, and, until there is real competition in the broadband market, you'll continue to see content providers having to pay to get access to customer eyeballs, whether that's through a transit provider, a CDN, or a paid peering relationship like we see here.
Now - if end-users could chose from several competitive Broadband providers, then maybe it would have been Comcast paying Netflix in orde...
5 comments
[ 0.23 ms ] story [ 33.4 ms ] threadFine for Netflix to do, but if prices rise and the thresholds where this applies lower, new services will get shut down due to cost.
Hopefully some ISPs will not charge and market "fast access to whatever you want" and not "fast access to whatever we chose based on what they paid").
In order to connect to Comcast's network you need port capacity on Comcast's routers. Netflix pushes approx. 3 Terabits [1] onto the Comcast network, so they need Comcast to supply (configure, support, monitor, upgrade, power, etc...) approx. 300 10-GigE interfaces, as well as any incremental Line Cards, Routers, Cabinets that might be required to add that port capacity and bandwidth capability to their network. (3 terabits is enough bandwidth to get your attention)
In addition, now that there is an extra 3 terabits of incoming data onto their network, Comcast also has to scale their outgoing network (to the home users) by another 3 terabits as well, so there are incremental costs there. It could be argued (and I would do so) - that those 3 terabits have already been paid for by the cable-subscribers, and so Comcast should be prepared to provide sufficient capacity to deliver it - but regardless, it's an additional cost that Comcast needs to account for, regardless of whether they are installing appliances in their network, or peering with content providers.
Comcast provides these peering locations in at least 17 discrete locations in the United States [2] - Many of those locations are just general CoLos - I've heard of about 25% on that list - so many of them aren't "Dedicated Comcast Data Centers" - they are CoLos in which multiple customers can come together and interconnect with each other or carriers such as Level3, Cogent, etc....
So, in those facilities, from Comcast's perspective, providing Port Capacity is a matter of connecting a fiber patch cable to one of their 10 GigE interfaces, and exchanging some BGP routes with a content provider.
It's true that there are several routing relationships Comcast can establish with content providers, and some of those relationships might involve placing those Content Providers in AS7922 (Comcast's Autonomous System) - but, from a cost perspective, there are a lot of costs that are duplicated, regardless of whether Comcast establishes a peering relationship with Netflix, versus having them route from within Comcast networks.
You still need to install 300+ 10 GigE ports and all the associated costs that go along with that, and you still have to provide downstream capacity of 3 Terabits to home subscribers. Peering with a partner in a 3rd party CoLo instead of having to provide them cabinets/electrical/etc for the appliances avoids those cabinet/appliance costs.
Something to keep in mind, is that in addition to providing the three terabits streaming capacity on their network, Comcast would also have to provide Capacity to keep those Appliances "supplied" - each of the appliances would need about five Terabytes of content loaded each night [3, pg 13]. There are also some operational advantages to managing a Content Provider at "arms-length" distance with BGP, versus having to do custom routing on your own network to support them. Comcast already knows how to peer and interoperate with many carriers/partners using BGP, so no new procedures/technology required to add Netflix to that list.
In the end - Netflix and Comcast each negotiated in their own best interests. Netflix's access to the Comcast network was provided in roughly the same way that Content Providers have been getting access to networks for the last 10 years, and, until there is real competition in the broadband market, you'll continue to see content providers having to pay to get access to customer eyeballs, whether that's through a transit provider, a CDN, or a paid peering relationship like we see here.
Now - if end-users could chose from several competitive Broadband providers, then maybe it would have been Comcast paying Netflix in orde...