CIDR allocation question
vaf at Valinor.Stanford.EDU
Fri Apr 8 22:44:42 UTC 1994
Given a site that has a need for a certain number of
addresses, but probably will never need additional networks to fill
out a power of two allocation. Should we allocate the larger block to
the site anyway to reduce the size of the routing table, or should we
allocate two CIDR blocks?
For example, if a site will only realisticly need 6 class C's,
should we allocate a block of 8, or a block of four and a block of two.
The answer to this question depends on several factors. First, and probably
most importantly: will all of the networks in question be aggregated into a
larger block at the provider level or are there specific policy constraints
that mean this site needs to be explicitly advertised beyond your network?
If splitting them into two blocks means that both blocks are explicitly known
to the global Internet, then the rest of the world would prefer that you give
them a block of 8. If the rest of the world will only see a bigger aggregate
that includes the block of 8 or the blocks of 2 + 4, then we probably don't
care very much what you carry internally...
Next, will you be aggregating the advertisement at the site level to propagate
within your network? If you are not doing that or if there isn't any reason to
do so (i.e. you have sufficiently few routes within your network that you can
affort to carry all of the individual class-C's internally), then it probably
doesn't make much difference to your internal routing whether the site is one
block of two.
In general, we allocate blocks in power-of-two chunks, though there have been
some cases where our guesses about future growth were incorrect, resulting in
some sites with more than one block.
More information about the NANOG