Hi all
Can I know how many ipv6 full bgp table routes now?
how many memory can run one ipv6 full bgp table?
how many peer for ipv6 in Router reflector you suggest?
Do you suggest to separate the ipv4 and ipv6 in router reflector?
Thank you for your info
Hi all
Can I know how many ipv6 full bgp table routes now?
Right now there are about 15k routes.
how many memory can run one ipv6 full bgp table?
This depends on the platform.
how many peer for ipv6 in Router reflector you suggest?
This depends on your architecture.
Do you suggest to separate the ipv4 and ipv6 in router reflector?
I recommend keeping your network as congruent between IPv4 and IPv6 as possible, with dual-stack.
- Jared
Can I know how many ipv6 full bgp table routes now?
Right now there are about 15k routes.
8k when you filter based on IRR.
Not based of IRR =D
Foundry CER2K
12111 BGP
Number of Neighbors Configured: 7, UP: 5
Number of Routes Installed: 22866, Uses 1966476 bytes
Number of Routes Advertising to All Neighbors: 53961 (41844 entries),
Uses 2008512 bytes
Number of Attribute Entries Installed: 22746, Uses 2047140 bytes
Number of Neighbors Configured: 6, UP: 5
Number of Routes Installed: 40326, Uses 3468036 bytes
Number of Routes Advertising to All Neighbors: 34987 (34987 entries),
Uses 1679376 bytes
Number of Attribute Entries Installed: 31290, Uses 2816100 bytes
Hi,
Why?
Regards, K.
For one thing, doing otherwise violates the principle of least astonishment.
Other reasons include simplified network diagrams, improved reliability, easier troubleshooting,
reduced complexity, and often lower costs of operations.
Owen
I asked a similar question a few years ago:
http://seclists.org/nanog/2007/Aug/653
Most of the answers came back along the lines of "keep your routing
boundaries congruent." Doing so makes documentation and troubleshooting
simpler -- having non-congruent boundaries is more complex and error
prone.
However, if a network you're running calls for non-congruency -- go for
it! Just be cognizant of the trade offs.
Amen to that. Not too long ago, I blew about 3 hours trying to debug an odd
networking issue on my laptop - finally tracked it down to the fact that my
IPv4 default route was pointing out the ethernet on the docking station, but
IPv6 was defaulting to the wireless card. Took a while because I knew *damned*
well that Fedora had long ago included my patch to allow specifying a
preference metric for multiple interfaces, and that I had set it to prefer the
ethernet when both were connected.
Turns out that the patch worked just fine for v4, but nobody ever carried it
forward for v6....
(I probably should cook up a patch for the v6 side.. 