[c-nsp] Growing BGP tables

[Oliver Dewdney]

Is there any more news on this problem? As far as I can see there are two
problems

1) The size of the BGP tables. These will only get bigger, but I believe
that there are steps that can be taken: limit prefix length and have a
default route, add more dram, and get Cisco to use better memory structures
to store the bgp tables ( with the risk of introducing bugs )

2) The size of the FIB. This is currently a mirror of the RIB. The issue is
that when the routing is done in hardware this FIB is loaded into the TCAM
of a limited size. I think that the FIB should be an aggregated view of the
RIB so hopefully have far fewer entries. As an example: Take a Edge router
connected to two internet peers and one internal router with one local
network, with a default route going over one of the peers. If you exclude
local subnet entries, you would have one 0/0 entry for the default route,
one x.x.x.x/n entry for the local network and then only entries for subnets
that you choose to use the other peer to transit. Looking at current
statistics that would be about 20,000 routes for a 50/50 split. A lot less
that the current 200,000 entries that are currently populated in the FIB.

Oli Dewdney


This e-mail is only intended for the person(s) to whom it is addressed and
may contain confidential information. LB Icon does not accept responsibility
for any loss or damage caused by this email or any attachments. Unless
clearly stated to the contrary, any opinions or comments are personal to the
writer and are not made on behalf of LB Icon. If you have received this
e-mail in error, please notify us immediately at info at lbicon.co.uk and then
delete this message from your system. Please do not copy it or use it for
any purposes, or disclose its contents to any other person. Thank you for
your co-operation.

LB Icon is the business name of Aspect Internet Holdings Limited, Aspect
Technologies Limited, LB Icon Limited and Escador Limited.