allocations/filtering, etc.

Jeremy Porter jerry at fc.net
Thu Mar 14 02:02:52 UTC 1996


I had promised a number of people a paper on economics of route filtering, and  I haven't quite
got it finished yet.  However with regard to current whining, I will post some of the numbers from
my current  work in progress.  I would like to encorage anyone that is working with
non-cisco routers, that has information on router limits with regards to memory utilization
per prefix, I would like to hear from you.  Right now my paper is basied with only
numbers I have from various members of Nanog, and my personal numbers in my routers.

-----
from "Economics of Route Filtering", Jeremy Porter, Feb 1996, unpublished.
-----
Current theory is that a large number of backbone routers will overload when 
more than about 64,000 prefixes are announced into the global backbone. There
are currently about 35,000 networks being announced. With providers filtering
routes with prefixes longer than 2^18 bits, with allocations remaining 
between 207.0.0.0 and 223.255.255.255, we have 17 networks of size 2^8, or
17048 networks of size 2^18.  This total is less than the 70,000 "doomsday" 
routing table. The 2^19 filter length also fulfills the recommendation some 
people have been making of 2048 prefixes per class A space.

This gives us with maximum density 4.4 million hosts on the Internet.  With 
50% utilization we are still talking about 2.2 million new hosts on the 
Internet.

Currently the reserved spaces have not been allocated except for one notable
exception which is a large block anyway.  Filtering on just the 2^19 limit 
gives us a total routing table size between 2.6 million prefixes and 5.3 
million prefixes.  Given current router technology one could estimate an 8 
fold growth in router table size to utilize all 5.3 million prefixes.  Thus 
future routers could need 512 megabytes of memory.  This is within the realm
of today's technology.

Now if we decided that we need our current routers to function, filtering on
2^14 out of 64.0.0.0 - 95.255.255.255 will give an additional 16384 prefixes
in a worst case allocation.  This is just slightly larger than current routers
will hold.  With moderate levels of aggregation in 207.0.0.0 - 223.255.255.255
and 64.0.0.0 - 95.255.255.255 we should be able to keep current routers under
the doomsday limit.

-- 
Jeremy Porter, Freeside Communications, Inc.      jerry at fc.net
PO BOX 80315 Austin, Tx 78708  |  1-800-968-8750  |  512-339-6094
http://www.fc.net



More information about the NANOG mailing list