Thanks Adam,
in fact, I'm already with 2 upstreams, each one with 2 links, I do
load balancing with them, announcing a /24 from one link to another
according to the traffic, these 2 existing ISPs have /24 filters.
I'm adding a new ISP with 3 links, and this 3rd ISP wants to change a
unique /18 filter; we plan to use all 3 ISPs at the same time, not
for backup only.
BR
At 15:42 12/08/2010, Adam Armstrong wrote:
On 12/08/2010 13:21, RAZAFINDRATSIFA Rivo Tahina wrote:
Hi all,
I'm BGP multihomed with 2 ISPs and have a /18 subnet.
I declared the /18 block in RIPE database as 64 /24.
I'm adding a new ISP and he asked me to modify the 64 /24 in RIPE to a
unique /18.
How does this change affect the existing routing with 2 former ISPs?
How can I announce a /24 prefix from one ISP to another? He is talking
about "no export", what is the purpose of this attribute?
Your new ISP thinks you're being incredibly stupid by announcing 64
/24s (thereby taking up 64 times as much space in the routing table
as you should be, could you imagine if the global table was
64x300k?), It's a justified reaction, in some parts of the UK you'd
be hung upside down by your feet for splitting a /18 into 64 /24s.
If you haven't done this to do some form of fisherprice traffic
engineering, you need to stop announcing the /24s to the existing
ISPs and just announce the /18 supernet to them (otherwise all
traffic will come in via them, as they'll have more specifics).
It's rare to have a good reason to split netblocks up like that, but
if you do, you could try splitting it into less tiny blocks, like 2
/19s or 4 /20s.
no-export is a well-known BGP community which tells the router not
to announce that route outside of the AS (or confederation).
adam.
_______________________________________________
cisco-nsp mailing list cisco-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/