On Sun, Jul 27, 2014 at 10:39 PM, Leon Towns-von Stauber
<[email protected]> wrote:
>
> I've been waiting to wade into this conversation, avoiding the 
> back-and-forth. I'll try to get all my contribution, FWIW, into this email, 
> but to make my position clear, Derek has been saying most of the things I 
> would've.
>
> The main point that needs to be made, and which almost never is in these 
> discussions or in articles about Netflix or Net Neutrality, is that there is 
> nothing fundamentally new here (along the lines of what Paul said). Internet 
> peering agreements have a couple of decades of tradition behind them (as do 
> conflicts about peering). True peering, in which money doesn't change hands, 
> has always been based on an assumption of roughly equal traffic. If you want 
> to hook up to someone else's network and aren't willing or able to offload 
> about as much traffic from your peering partner as you intend to send, then 
> you're going to pay for it.
>
> It's true what one discussion participant said, to a certain extent: the fact 
> that it's Netflix sending all this traffic is none of Verizon's business. 
> Verizon's direct issue is with Level 3, and normally it would be Level 3 
> obligated to pay Verizon; the case for Verizon to pay is nonexistent, AFAICT, 
> based on traditional arrangements.

First of all, you are completely correct that peering wars have been
going on for several
decades now.   So in some sense, this isn't new.  In the old days,
everyone wanted to be considered a "tier one" network because that was
supposedly how you got settlement free (no money) peering.   Typically
this was more likely to happen if there were roughly equal flows in
both direction.    Back then most networks had a mix of customer types
connected to them so in the end the thing that mattered was how many
customers you had.  The more customers, the more clout, and the higher
chance that you could get someone to pay you to peer with them.

As data hosting centers took off and residential speeds increased,
different networks started to specialize.   I would suggest that we
now have server networks, eyeball networks, and (maybe) business
networks.  With this specialization, efficient network protocols were
going to unbalance flows with a large net flow from server to eyeball
networks.   Interestingly, the eyeball networks have designed their
internal networks under that assumption.   They almost always provide
asymmetrical bandwidth (more download then upload).   And this is GOOD
for the eyeball networks' business.
They can offer more of what the eyeballs want (downloads) for less
investment in equipment.

Of course, there is nothing requiring things to be done efficiently.
There is no reason that the proprietary clients (or javascript code)
that are used to communicate with the
server network hosts couldn't be written to spew large data flows back
towards the
server networks.  If a large company that many eyeballs want to access
decided it wanted to do so, it could theoretically dial the ratio of
in/out bound data to practically any
number it wanted.   It controls both ends of the network via running
code at both ends.  I suspect that most server networks are engineered
in such a way that they could actually handle the increased inflow
while the eyeball networks would find their last mile networks
saturated.   Not good for the eyeball networks.   Even worse, given
your view of how peering should be done, would this mean that Verizon
should start paying Netflix for the privilege of talking to them?
(Which some have reported has already been stated as a future
concern.)

Now, I don't suggest the above scenario as one which I think is a good
idea.   But I did want to point out that the idea that paying for
peering should be all about aggregate direction of netflows is
shortsighted.   That just happens to be the way that it was done in
the past, not some natural law.   The Internet has changed so what was
once truly a many to many set of networks flows has become more of a
few to many set.   I think the charging model needs to change as well
and not be based solely on the direction of the aggregated flows.
Unfortunately, I don't know what that new charging model should be.

Bill Bogstad
_______________________________________________
Discuss mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to