This is partly a result of MSDP's requirement for MBGP to poison routes
received from a peer in a neighbor AS. Rather than have to check whether
the NLRI is unicast or multicast, Juniper appears to have chosen to send the
routes regardless. See draft-ietf-msdp-spec-10.txt for further info.
regards,
chris
> -----Original Message-----
> From: andrew kevin [mailto:andrewkevin2000@yahoo.com]
> Sent: Monday, February 04, 2002 8:17 AM
> To: Julian Eccli; juniper-nsp@puck.nether.net
> Cc: juniper@groupstudy.com
> Subject: RE: [j-nsp] BGP route
>
>
> Hi,Julian,
> Do you imply that the implementation of Junos is to
> advertise the ebgp routes learing from the source?
> Since the whole internet routes is 130K today,it must
> impact not only on the routers, but also the bandwidth
> for those wasting routing info.
> Why not Junos stop sending it back automatically?
> In real world, it has to manually configure the policy
> you mentioned to do the performance tuning.
> Is there some special consideration of Junos?
> Why Cisco do it default?
> thank,
> Andrew
>
>
> --- Julian Eccli <je@juniper.net> wrote:
> > Answers inline...
> >
> > > -----Original Message-----
> > > From: andrew kevin
> > [mailto:andrewkevin2000@yahoo.com]
> > > Sent: Sunday, February 03, 2002 8:13 PM
> > > To: juniper-nsp@puck.nether.net
> > > Cc: juniper@groupstudy.com
> > > Subject: [j-nsp] BGP route
> > >
> > >
> > > Hi,
> > > I found it's strange that BGP of Junos can't
> > follow
> > > the split-horizon rule(BGP is a DV routing
> > protocol),
> >
> > split-horizon is not used in BGP and BGP is
> > considered a path vector
> > protocol.
> >
> > > after receiving EBGP neighbors route , it will
> > send
> > > back those route which learns from the original
> > one to
> > > the source, and the original will reject these
> > routes
> > > according to the AS-PATH attribute.
> > > It will waste the router's resource to send those
> > > routes.
> > > It is different with Cisco's implementation.
> > > Cisco will not send back to the original.It seems
> > even
> > > reasonable.
> >
> > RFC1771 states it's optional to advertise or not
> > advertise back from the
> > neighbor the route is received from in section
> > 9.1.3. In a modern day
> > environment it would not matter, and should have no
> > impact on router or
> > link performance since we only send back what is
> > sent to us.
> >
> > > Or must we configure a policy to preven Junos from
> > > sending updates back?
> > > Or any better method?
> >
> > You need to set up an export policy on the peer to
> > prevent readvertising
> > back to your neighbor.
> >
> >
> > -Julian
> >
>
>
> __________________________________________________
> Do You Yahoo!?
> Great stuff seeking new owners in Yahoo! Auctions!
> http://auctions.yahoo.com
>
This archive was generated by hypermail 2b29 : Mon Aug 05 2002 - 10:42:39 EDT