[ale] Router DNS Tables - off topic

SanMillan, Todd tis3 at cdc.gov
Thu Sep 28 08:08:03 EDT 2000


I am not a router expert, but... As several people have pointed out, DNS is
not the issue.  Size of routing tables is.  I don't know as much about
backbone routers as I would like, but  one of the strategies of the IPv6
address space is to have a hierarchical address scheme.  The current IPv4
addresses are globally unique, but scattered all over, i.e.. you have no
idea if 123.001.100.000 is anywhere near (in network space) 123.002.100.000.
The IPv6 addresses are arranged like US Zip Codes or direct dialing codes
(Country Code- Area Code - Exchange - Number), so the core routers only need
to hold information on how to get to say North America, then the next level
only needs info on how to get to the Southeast, then the next level knows
how to get to Atlanta.  My level examples are probably wrong in scale, but
you get the idea.  As it stands now, the core routers have to hold an entry
for each and every advertised route, with no aggregation designed in.  In
IPv6, you can't request a network number block direct from the highest
authority any more.  Only the regional registries will get blocks and they
will be huge contiguous blocks that allow aggregation.  They in turn will
issue aggregate blocks to large providers, who issue to smaller providers
and end points.  It will burn a lot of addresses to provide for aggregate
routes, but the space is huge, at 128 bits (course, that was the impression
when IPv4 (32 bit addresses) was designed, so we'll see how huge it is in 20
years).

-----Original Message-----
From: Lathe [mailto:magius at wittsend.com]
To: ale at ale.org
Sent: Wednesday, September 27, 2000 8:55 AM
To: ale at ale.org
Subject: [ale] Router DNS Tables - off topic


I know this is somewhat off topic for a Linux group, but I am curious
about the dynamics of routers.

A friend and I got into a discussion on the allocations of IPv4
addressing as opposed to the IPv6 addresses and something odd came up.
It was explained to me that an existing problem that the internet has is
that the a lot of the major routers that are in use are running out of
memory in their DNS tables.  Once these table are full, the routers then
do a complete memory flush and therein lies the problem.  Once that
happens, the routers will no longer be able to forward datagrams to
address blocks.

I'm not as familiar with routers as I'd like to be.  But this doesn't
quite sound right to me.  I was curious if anyone could possibly explain
the dynamics of the memory tables to me.  Or, if this system IS correct,
could someone tell me so.

Thanks in advance,

Scott Warfield


--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message
body.
--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.





More information about the Ale mailing list