Wasn't one of the big reasons behind the TeliaSonera vs Cogent spat a little while ago the large difference in traffic going one away across the peering agreement versus the other way and that Cogent wanted Telia to pay more for that difference?
I don't know what the ratio of data uploaded to YouTube versus downloaded is in comparison to Justin.tv or other big bandwidth heavy sites, but I would guess that Google still has to pay quite a hefty price for these peering agreements. ISPs after all don't just want more traffic across their network for the sake of it.
Each time the traditional media butchers a technical story I'm astonished anew. And then I get a second round of surprise when everyone accepts it so uncritically because it was in a non-blog context.
The down votes do not signify disagreement with the assertion. They signify that the direction you are choosing to take the discussion in with the comment is inappropriate for this situation.
1. I am in this forum since it's beggining. You think I don't know what is appropriate or what is going to be down voted?
2. Your percepetion that I want to pollute the discussion would be valid If only I have history of doing that.
i. define "inappropriate" and why/how this "situation" is affected. Since, when comments must allign with one scope?
ii. according to your "way" I could say the parent comment concerning "justin.tv bandwidth costs" has nothing to do with this thread, as it's essentially abstract and it contributes nothing.
3. Whoever is annoyed for a fact to be mentioned then it's not my problem.
1. I am in this forum since it's beggining. You think I
don't know what is appropriate or what is going to be down voted?
In fact, the discussion was about the original article, and bandwidth costs, not the possibility of Youtube/justin.tv having copyrighted content on them.
Were this a discussion of copyright, people would probably not feel that your assertions were off-base.
The comment from abstractbill reffered to Credit Suisse report which was about how the analysts have estimated Google's YouTube may lose $470M in 2009 and more in the future.
By reffering to that report, it means his comment reffered to justin.tv costs and not on how they do it. Meaning that it's not so expensive according to their experience.
So, If I am going to reply to that comment it means I am going to talk about money. If I am going to talk about money I have every right to comment that they are profiting from content owned by someone else, WHICH has nothing to do if its legitimate or not.
My comment was as much relevant as the parent comment I replied to.
As much for justin.tv, they should better follow YouTube strategy to look broke, because if people who own that content know they are actually profitable someone will knock their door.
Obviously it's very easy to play semantic logic games and come up with a free association that "proves" how on-topic you were. Here's the problem though:
People here think justin.tv's bandwidth costs are relevant to the article. They don't think you political agenda about piracy is relevant.
You are wrong. Youtube/Justin.tv don't "exploit copyright content." The law explicitly allows what they do.
The Online Copyright Infringement Liability Limitation Act creates a safe harbor for online service providers. It specifically says that those sort of sites are not liable for infringement as long as they take down content at the copyright holder's request (i.e., when the copyright holder issues a DMCA takedown notice). There are a lot of details which you can read about at http://en.wikipedia.org/wiki/Online_Copyright_Infringement_L....
Youtube/Justin have just as much right to host user uploaded content as copyright holders do to enforce their rights.
Can anyone find the original Credit-Suisse article? Are their estimates based on pure ISP costs, or do they incorporate the cost of running the infrastructure to serve that content?
Assuming he's right about Google's data transfer costs (I... know next to nothing about how the backbones operate), the amount of Google infrastructure dedicated to running Youtube can't be that cheap. That's a huge amount of data, that needs to be distributed to all their various local data centers, plus developers working full time on the project, etc.
I guess what I'm saying is, just because it's easy to want the bean counters to be wrong, doesn't mean they are.
One of the articles linked to from the main article says Credit Suisse estimated Google's bandwith at 30 million megabits/second. At google appengine prices, that would cost ~ $440 per second (I'm assuming non-SI units). Presumably google's real costs are much lower than that, but still...
I don't get this. Wouldn't most of YouTube's traffic be downstream from Google's servers to the hoards of individual internet viewers all over the world? How exactly does peering help?
If Google operates their own backbones, then Google's peering agreement with other tier 1 backbone providers helps. The peering agreement allows packets passed in/out from Google with other providers without extra bandwidth charge.
Until, of course, peering agreements are renegotiated. I think this is where other bandwidth providers will stick it to Google.
They only have leverage for the moment. It's like falling off a cliff. Everyone thinks . . . "so far . . . so good" on the way down. But it's not the period that you are falling that is important, it's how you stop when you reach the bottom.
Doubtful. Google's been around for years now. If the bandwidth providers were going to stick it to Google, they already would have.
Furthermore, Google and the bandwidth provider both stand to gain by cutting out any middlemen (i.e. tier 1 providers), so they both have an incentive to be reasonable.
It makes perfect sense that Google peers with downstream ISPs.
However, I think Mr. Turner's analysis is wrong. It is in the interest of the Tier 1s to continue to gouge Google. YouTube pushes out way more bytes out than it pulls in... even including video uploads, client-made HTTP GETs, and TCP ACKs. This asymmetry pokes holes in the presented argument (assuming that those in a peering agreement measure bandwidth in bytes per second and not packets per second, and that both parties strive for balanced bandwidth usage).
If Google were to route all traffic through a single Tier 1... the latter would be completely screwed. The ISP would constantly have to push all these YouTube bytes out to all the other Tier 1s, and it's all the other networks that get the leverage.
The only leverage for this clever Tier 1 is in exploiting their position as the exclusive route into Google. They may even be tempted to extort the other networks... pay lots of money or your user don't get access to Google. This would barely work. All the remaining Tier 1s would simply stop by the Googleplex to say "hi guys, here's a bunch of half-duplex links. We are going to push bytes to you but not take any from you. Ciao". Google, of course, would agree. This lone Tier 1 is effectively bypassed, and it still has to find a way to get its massive quantity of upstream traffic into everyone's networks.
A Tier 1 might offer Google a special deal if they feel they are underutilizing their uplinks, and therefore want to push the peering situation back into equilibrium. But, they would definitely put a cap on Google's bandwidth usage, so as to guarantee that things don't swing too far the other way.
It definitely makes sense for Google to peer with the downstream guys (the Tier 2s and 3s). This reduces costs for both parties. The ISP doesn't have to pay for upstream transit to route to Google, and Google doesn't have to pay for upstream transit to route into the ISP. It's a win-win situation. And, it also improves user experience by decreasing latency, a big priority for Google.
