Google Fiber: No Charge For Peering, No Fast Lanes 238
An anonymous reader writes "Addressing the recent controversy over Netflix paying ISPs directly for better data transfer speeds, Google's Director of Network Engineering explains how their Fiber server handles peering. He says, 'Bringing fiber all the way to your home is only one piece of the puzzle. We also partner with content providers (like YouTube, Netflix, and Akamai) to make the rest of your video's journey shorter and faster. (This doesn't involve any deals to prioritize their video 'packets' over others or otherwise discriminate among Internet traffic — we don't do that.) Like other Internet providers, Google Fiber provides the 'last-mile' Internet connection to your home. ... So that your video doesn't get caught up in this possible congestion, we invite content providers to hook up their networks directly to ours. This is called 'peering,' and it gives you a more direct connection to the content that you want. ... We don't make money from peering or colocation; since people usually only stream one video at a time, video traffic doesn't bog down or change the way we manage our network in any meaningful way — so why not help enable it?'"
Re:terminology (Score:5, Informative)
The phrase 'Fiber Server' does not appear in the article.
Blame the submitter.
Servers (Score:2, Informative)
Nope. They explicitly permit non-commercial servers. From the Fiber use policy: https://support.google.com/fiber/answer/2659981?hl=en&topic=2440874&ctx=topic:
"However, personal, non-commercial use of servers that complies with this AUP is acceptable, including using virtual private networks (VPN) to access services in your home and using hardware or applications that include server capabilities for uses like multi-player gaming, video-conferencing, and home security."
Re:We don't make money from peering or colocation (Score:4, Informative)
...and they have no-compete clauses with each other
No they don't. That would be an illegal cartel, and they know it. No, they have "gentleman's agreements" with each other not to compete. Which amounts to the same thing, but the only proof is the indirect evidence that they never actually compete, and so it's not particularly actionable in court.
And don't look now, but Ma Bell is very nearly completely reconstituted. The only piece missing is Pac Bell. Of course the FCC and FTC will remain determinedly oblivious to that fact.
Re:One person a bottleneck doesn't create... (Score:4, Informative)
Re:terminology (Score:5, Informative)
The price of a T1 hasn't really changed all that much. Due to LEGAL requirements for the SLA associated with a T1, its unlikely to change for the foreseeable future.
Now getting far more than a T1's worth of bandwidth for far less is easy, but thats not a T1 nor does it come with the SLA that will have the provider working at 3am on a Sunday morning to get it back on line as required BY LAW.
Just because you get 1.5mbit of data doesn't mean you're getting an actual T1.
LECs are still the only ones who can offer a T1 for the most part.
If you knew what the terminology you are using actually meant you wouldn't have made such statements.
Uverse is 45 Mbps over new(ish) wire (Score:4, Informative)
Uverse maxes as 45 Mbps and requires a minimum of UTP drop. It does not work over "the existing infrastructure" (untwisted pair) unless that infrastructure has recently been upgraded.
Re:One person a bottleneck doesn't create... (Score:4, Informative)
For providers and data demanded by Comcast customers. I also complain about that, but I'm not in a position to get quoted for it. Does that make it less valid a complaint? Keep in mind, every bit Netflix ever sent me was at my request. I am responsible for that traffic. Why is Comcast trying to charge someone else for my choice?
That is a very valid comment, from a consumer's point of view. The answer is not simple, but let me try and simplify it.
Consumer broadband connections are always oversubscribed. This means that for every 100Mbps customer, an ISP will have an X amount of actual bandwidth available. For example, an oversubscribtion rate of 1:25 means that for every 25 100Mbps subscribers, only one 100Mbps link will be provisioned. The reason for this is that building a network is ridiculously expensive, and this is the only way to make it affordable for consumers. Another reason is that very few consumers will actually use the bandwidth. For example, I have a 25Mbps link from Charter. If I look at my stats, I barely used 512kbps on average over a month. However, when I do actually download something, I see that my link is more than what I pay for; I usually get over 30Mbps.
Further down the road, that oversubscription becomes a bit blurred. Most large ISPs have big networks, with multiple entry and exit points. In short, there are two ways in which traffic flows from one network to another: via direct peering, or via a paid transit provider. Direct peering is most of the time handled at an Internet Exchange point. Members will all connect to each other, and peer whenever they come to an agreement. Transit is when I pay a third party to transport my packets to someone else. So, let's say I am Comcast, and I need to transport packets to and from AT&T. If I do not have a peering with AT&T, I will find someone who does. Let's say Level 3 Communications does have a peering agreement with AT&T. I can then pay Level 3 Communications to transport my traffic to AT&T. The path will then become: Comcast Level 3 AT&T.
In this example, interconnect 1 is paid, and interconnect 2 may be paid transit or free peering (from Comcast's perspective, that is irrelevant).
There will be many of these entry and exit points. Today, most of these interconnects are at 10 Gigabit Ethernet speeds. This does mean, that in a lot of cases. the aggregate bandwidth between two networks on the internet is 10Gb. Let's turn back to our example. In this case, it is actually reversed. Comcast has a peering agreement with Level 3, and Netflix pays Level 3 for transit traffic. So what ends up happening is:
Comcast Level 3 Netflix
While "peering" sounds free, it is in fact not. Peering requires network ports and backhaul capacity. If Comcast has 1 10G peering port with Level 3, they only need to haul back 10G to the rest of the network. If Comcast upgrades that to 10G, then not only do they need an additional 10G port, but also the capacity to transport that traffic further downstream. This is regardless of the rest of Comcast's network. They may have (and probably will) have hundreds of other 10G peering points elsewhere, with tons of capacity. That is useless here, since all of a sudden all traffic is centralized to 1 entry point.
Now, from Comcast's point of view, I totally understand their reluctance to invest significantly only to haul back the traffic of Netflix, which then makes a profit off of it. Is it the right thing to do? That's debatable. But to bluntly say that all of a sudden Comcast is the root of all evil, that's a bit too far.
Re:terminology (Score:4, Informative)
Who runs their own radio network with multiple towers instead of cell phones?
I cannot think of anyone. Even police and emergency services have switched to cell phones, albeit on a dedicated network. A network where 1.5Mbps per tower is woefully insufficient, of course.
Modern cell towers use SyncE.