The US is very well connected. Besides the two main access points, MAE East
in
Washington DC, and MAE West in San Francisco, there literally hundreds of
NAP,
national access points, in major cities. Many regional providers have
maintained
their peer presence at these NAPs.
Also, you have no control over the number of hops to your site. These huge
Cisco
routers will determine the best routes and dynamically adjust them. Your
forward
and reverse routes may also not be symmetric because of that.
Which backbone would give you the best connected site on the Net? None! In
the
past three years, in my former existence, we bumped around UUnet, Digest,
Global
Center, Frontier, ... There is no difference in connectivity. It is more of
the
support problems. And, avoid ones who claim they are well-connected through
one
of the MAE points. They are slow as molass in the winter, even if there are
fewer hops.
My answer is simple. Don't worry if you are in the US.
Outside the US is a different story!
Bao
-----Original Message-----
From: Robert Hoffman [mailto:">rob@frankenlinux.com]
Sent: Monday, May 15, 2000 9:46 PM
To: ">ale@ale.org; Jeff Hubbs
Subject: Re: [ale] OFFTOPIC: Structure of the Internet
Damn Jeff, you're always asking the easy questions aren't you? :-)
While I probably know about as much as you do on the subject, two things
cross my mind:
You should always get fewer and faster hops if you are co-locating with one
of the backbone providers like UUNET. Many of the other providers are simply
re-sellers for the big boy's services.
I know that there is a bottleneck when going overseas. For instance, all
traffic to Europe is routed through the boys at the NSA in Washington. It
seems to me that global response would be faster if you had a co-lo on each
continent you are serving.
Of course, real data on worldwide bandwidth saturation would be nice.
-Rob
---------- Original Message ----------------------------------
From: Jeff Hubbs ">Jhubbs@niit.com>
Date: Mon, 15 May 2000 14:57:38 -0400
>Can anyone point me to information regarding the global structure of the
>Internet in a telecomm/routing sense (this is distinct from the recent
>article(s) about the structure of the Web, which has more to do with
linking
>structures than anything else)? The question I'm trying to ask is
>essentially this: what physical places (ISPs/telcos, I'm presuming) that
do
>co-location for Web servers are going to tend to be the fewest router hops
>away from the most end users? Is there a way to evaluate the degree to
>which an ISP/co-lo is in the telecomm "core" of the Internet?
>
>The larger issue I'm assessing here is to what extent is it necessary to
>deploy servers globally to serve a modest Web function that has a global
>reach, assuming that the function in question isn't some kind of
>massive-bandwidth thing. If I'm a business in Chattanooga and I'm a
>subscriber to a Web service, how bandwidth-intensive a service does it have
>to be before it matters whether the Web service's servers are in Atlanta,
>London, or Tokyo?
>
>- Jeff
>--
>To unsubscribe: mail ">majordomo@ale.org with "unsubscribe ale" in message
body.
>
--
To unsubscribe: mail ">majordomo@ale.org with "unsubscribe ale" in message
body.
--
To unsubscribe: mail ">majordomo@ale.org with "unsubscribe ale" in message body.