Re: Study: Most Internet traffic bypasses tier-one networks

[David Farber <dave () farber net>]

Begin forwarded message:

From: "Mike O'Dell" <mo () ccr org>
Date: October 13, 2009 10:41:27 PM EDT
To: dave () farber net
Subject: Re: [IP] Re: Study: Most Internet traffic bypasses tier-one  
networks

David Farber wrote:
> Begin forwarded message:
> From: dewayne () warpspeed com (Dewayne Hendricks)
> Date: October 13, 2009 1:53:17 PM EDT
> To: Dewayne-Net Technology List <xyzzy () warpspeed com>
> Subject: [Dewayne-Net] Study: Most Internet traffic bypasses tier- 
> one networks
> Study: Most Internet traffic bypasses tier-one networks
> Telephony Online
> By Ed Gubbins
> The majority of Internet traffic now goes through direct peers and  
> does not flow through incumbent tier-one telecom networks, according  
> to a recent report from Arbor Networks, which sells network  
> management and security products.
> Tier-one incumbents were once the chief providers of connectivity  
> between content companies like Google and local or regional  
> broadband providers like Comcast. But over time, Google and other  
> content providers have built out their own infrastructure,  
> connecting more directly to end users and bypassing those tier-one  
> intermediaries.
> <http://telephonyonline.com/global/news/arbor-traffic-study-081309/>
> RSS Feed: <http://www.warpspeed.com/wordpress>
> -------------------------------------------
> Archives: https://www.listbox.com/member/archive/247/=now
> RSS Feed: https://www.listbox.com/member/archive/rss/247/
> Powered by Listbox: http://www.listbox.com

that is a completely wacky statement. i don't believe *anyone*
knows much of anything about "most Internet traffic", other than
it's carried in IP datagrams. the network operators have enough trouble
making sense of their own networks. i find it hard to believe that
some random 15th party has sweeping insights about the global
Big-I Internet. i don't even believe it about North America.

besides, that would seem to suggest that Google or Microsoft
or AOL is the only destination of importance, which is hardly the case.
a cosmic boat-load of bytes gets delivered by Akamai's servers, to
pick one alternative at random. (yes, i know they have some of
their own fabric, too.)

however, if the real thesis is that more organizations are building more
pipes to more places, and that is shifting traffic distributions,
that's hardly headline news. it's also not a zero-sum game!

the traffic flows on "tier-one" networks remains impressive and growing,
based on what i hear from senior network engineers at several
"tier-one" networks.

the fact that some organizations have grown to the size where it becomes
economical to build a "tier-one" network of their own should surprise  
nobody who's been paying attention. the fact the scale of their  
network and traffic satisfies the requirements for zero-dollar traffic  
exchange with other large networks doesn't hurt, either. it's *always*  
better to have a direct, one-hop path between where you are and where  
the packets want to go, and sometimes the economics and technology  
both work out to make it reasonable to do.

besides, whoever said the number of "tier-one" networks was established
by either immutable historical precedent or intrinsic physical constants
of the universe? tomorrow, somebody could decide pony-up enough  
dollars to build yet another "tier-one" network, either nationally or  
trans-nationally. After all, "It's only money!" To make it easy,  
Forbes publishes a list of people, at least 20% of whom could probably
do it with loose change from the sofas. (grin)

if they have a business plan which supports playing
the game at that scale, more power to them. it's hard argue that more
and better pipes is a bad idea, and if they are high-traffic paths,
it's even better.

as for hand-wringing because existing "tier-one" networks might "lose  
business" when someone like Google or AOL or Microsoft builds
their own "tier-one-scale" network, i'm not losing sleep worrying
about them. if they are doing their job on the sales side, things
should take care of themselves pretty quickly. and it might give
someone a respite of a quarter or two in the growth of network
operation expense - not the worst thing that could happen, either.

just as long as networks don't sell service for less than it takes
to deliver it, life will be OK. if, however, we have a repeat
of 1998-2000, where prices were positively psychotic - literally
unrelated to the actual cost of building and running the networks -  
things can get ugly again pretty quickly as well.

        -mo











-------------------------------------------
Archives: https://www.listbox.com/member/archive/247/=now
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com