Slashdot Mirror
YouTube's Bandwidth Bill May be Zero
MrShaggy writes "Credit Suisse made headlines this summer when it estimated that YouTube was costing Google a half a billion dollars in 2009 as it streamed 75 billion videos. But a new report from Arbor Networks suggests that even though Google is approaching 10 percent of the net's traffic, it's got so much fiber optic cable it is simply trading traffic, with no payment involved, with the net's largest ISPs. 'I think Google's transit costs are close to zero,' said Craig Labovitz, the chief scientist for Arbor Networks and a longtime internet researcher. Arbor Networks, which sells network monitoring equipment used by about 70 percent of the net's ISPs, likely knows more about the net's ebbs and flows than anyone outside of the National Security Agency."
I really don't see why Google would be paying much. It seems the guy who wrote that article now discovered how peering works.
Routing graph for YouTube AS
Routing graph for Google AS
YouTube alone has direct peering contracts with AT&T, RETN, TINET and via Google AS with Net Access, NTT Communications, Telia, Level3, SIG, Sprint, Global Crossing, MFN, Cogent, Port80, Internet2 and AOL.
Depending on the terms, it means Google can also act as a peering or transit point between these companies and or even have an IXP's at their locations, so theres incentive for ISP's to sign up beneficial transit agreement, especially considering Google has data centers around the world. Google has more power than Tier 1 ISP's alone. The article's note about "serving customers YouTube faster" is a moot point - Google's infrastructure and routing contracts alone act as a great incentive for ISP's to make a peering agreement with Google.
Google already ran the fiber for other purposes. So that cost was already planned for, well before they acquired YouTube. So, yes, it cost them nothing extra.
Owning and maintaining all that fiber is costing Google money. Even if they are not paying anything to other providers for handling YouTube traffic it is using bandwidth on their own fiber that they could otherwise sell or use for something else.
Warning: this article may contain humor, sarcasm, parody, and perhaps even irony. Read at your own risk.
The Wired article is from last fall. Arbor's blog post this week by Labovitz has better information. The most interesting data is a chart showing how 60 percent of Google's traffic takes advantage of direct peering, up from 40 percent a year earlier. Given the volume of traffic, we're talking about, there's some meaningful economics in that change.
RichM
Data Center Knowledge
.05Mbps = 50kbps. Sound familiar? He means dial-up! I'll be here all week for simple math help.
The question is, though: To whom would they be selling those gobs of bandwidth? The nature of bandwidth, overall, remains geographically fixed; you can't sell (much) of your bandwidth capacity in the united states to a company in Japan; they still need the pipes going, overall, from Point A to Customer B.
At the volumes in which they are dealing with, they don't really have a lot of customers who can conceivably use that much bandwidth. So it's definitely in their best interests to trade with them preferentially.
If the options are A) Trade to defer costs, or B) Try to sell to others and discover nobody else wants to buy a tenth of our capacity, they'll usually find that A) is a smarter business decision.
"To pass through the jungle; silence, courtesy, ferocity, as the occasion demands." -- Kamau, "Proper Passage"
More accurately, this is like saying "I don't own a car, so my petrol costs are zero", and everyone in the comments going "But that doesn't include your bus tickets or the time you spend walking!", and completely missing the point.
Almost all companies lease their offices. They could buy them and save rent. It would possibly be cheaper. They don;t though. They don't want all that capital tied up in property. They can use it for business expansion instead.
So Google owns a bunch of fibre. This has a capital cost. That's money that could have been invested somewhere else, so it's not free. They could have leased the fibre from a third party. Presumably they worked out that it would be cheaper not to do this. They could probably have saved money by leasing bandwidth from a third party. The third party would then be able to amortise the costs over several customers if there's surplus bandwidth. Having capital tied up like this isn't "free".
This is what I am hearing:
One person says Google's bill is zero, because they run the infrastructure themselves.
Another person says Google's bill is not zero because they have to maintain the network.
It's all about perspectives: Do you count internal cost or not in the discussion. Obviously it cost "something" for the infrstructure. Is it a fixed cost internally which can be minimized and absorbed or is it an external bill which can increase significanly as the business expands.
I think the point of the article is to debunk inaccurate speculations from traders who have no technical and real commercial knowledge who may be trying to trash Google's stock for short gain. Not necessarily figure out how many Washingtons Google has to shell out.
Then again, where would be the fun of slashdot if we can't go back and forth on the chicken-and-the-edd argument...
So, even their power supplies are in perpetual beta!
A post a day keeps productivity at bay.
The problem with this argument is that these guys' physical maintenance bills are significantly higher than most everybody else's. We may quibble whether this counts as an upstream bandwidth cost; it's upstream from the customer, but not from the ISP. But even in the second case, strictly speaking, peering is basically buying some of somebody else's bandwidth and paying not with money, but with some of your own bandwidth. But you still have the costs incurrent in delivering that "payment."
Are you adequate?