BIRD performance in a dynamic scenario

Yakimus // Netassist NOC yakimus at netassist.ua
Mon May 26 22:14:43 CEST 2025


Please clarify how do you describe the configuration for your dynamic peers? 

-- 
Mykola Yakymenko
NETassist

On Mon, May 26, 2025 at 06:40:33PM +0000, Maxime Peim (mpeim) wrote:
> Any idea if there is a way for dynamic neighbours?
> 
> De : Bird-users <bird-users-bounces at network.cz> de la part de Maxime Peim (mpeim) via Bird-users <bird-users at network.cz>
> Date : mercredi, 21 mai 2025 à 14:32
> À : Yakimus // Netassist NOC <yakimus at netassist.ua>
> Cc : bird-users at network.cz <bird-users at network.cz>
> Objet : Re: BIRD performance in a dynamic scenario
> Indeed, if you know the peer in advance, right ?
> What about dynamic neighbor ?
> 
> De : Yakimus // Netassist NOC <yakimus at netassist.ua>
> Date : mardi, 20 mai 2025 à 15:56
> À : Maxime Peim (mpeim) <mpeim at cisco.com>
> Cc : bird-users at network.cz <bird-users at network.cz>
> Objet : Re: BIRD performance in a dynamic scenario
> Hello!
> 
> You can add the comunity directly in the inbound filters of the peer
> 
> --
> Mykola Yakymenko
> NETassist
> 
> On Tue, May 20, 2025 at 07:43:28AM +0000, Maxime Peim (mpeim) via Bird-users wrote:
> > Hi everyone,
> >
> > First of all, thanks a lot for BIRD !
> >
> > I have some questions about the dynamicity of BIRD wiht BGP.
> > In a scenario where routes needs to be updated accondingly depending on some external informations, I find out that those informations would need to be written into the BIRD configuration file.
> >
> > As an example, if we need to add a bgp community based on the peer address (already known), we would need something like that :
> >
> > ```bird
> > function get_community_from_peer_address(ip addr) -> pair {
> >                 case addr {
> >                                <addr1> : (1,1) ;
> >                                <addr2> : (245, 12434) ;
> >                                …
> >                                else : (0, 0) ;
> >                 }
> > }
> > ```
> >
> > And we could update that function if needed in the future but the configuration will need to be parsed each time.
> > If the rate of update is high, it can become a bootleneck.
> >
> > So I would like to know if I am going the right direction doing that way ?
> > Is there a plan for a direct API to BIRD internals instead of passing a configuration file ?
> >
> > Thanks in advance for your answer,
> > Maxime Peim
> >


More information about the Bird-users mailing list