loco at isi.net
Mon Jul 29 23:36:22 UTC 1996
On Mon, 29 Jul 1996, Peter Lothberg wrote:
Roy <garlic at garlic.com>
|} > They use the ping times to figure out which server would be closest.
|} > All the servers are not located in the same place. The idea is that
|} > european users may receive better service from a european server.
Peter Lothberg <roll at stupi.se>
|} The network topology and geography does not match very well, I would guess
|} that the network center of Europe is likely to be somwhere on the US
|} Only Stockholm have multiple international E3 links....
There is still a rather high bit of latency to cross the Atlantic. Not to
mention that within Europe, not everyone has an STM-1 backbone. I'm under
the impression (anyone feel free to correct me :), that the majority of
the infrastructure is based on multiple E1s with some E3 connections and
perhaps a few SDH links. One could probably take this picture and copy it
a few times to represent several other countries and/or whole continents;
remove pieces and get a picture of even more countries.
Distributing web servers to remote corners of the world can only be
characterized as a good thing. Not to mention the added value that is
gained in localized content *and* advertising. If the end-user is
shuffled to a 'local' server rather than the 'master' server, the end-user
is theoretically getting a higher bandwidth connection, perhaps content in
his/her local language and perhaps directed marketing.
If traffic is to be moved away from exchange points, a good way of doing
it is to move the content. Developing a mechanism for end-users to access
that content transparently around the world is another discussion
More information about the NANOG