Date: Wed, 03 Mar 2004 06:09:29 +0100 From: Arnaud Pignard <apignard@frontier.fr> To: freebsd-net@freebsd.org Subject: Re: My planned work on networking stack Message-ID: <6.0.3.0.2.20040303054929.06bc9988@213.161.193.184> In-Reply-To: <p06002014bc6a36fcdf29@[10.0.1.3]> References: <4043B6BA.B847F081@freebsd.org> <200403011507.52238.wes@softweyr.com> <20040302031625.GA4061@scylla.towardex.com> <20040302042957.GH3841@saboteur.dek.spc.org> <20040302082625.GE22985@cell.sick.ru> <p0600200ebc6a27773c31@[10.0.1.3]> <20040302125935.GA25835@cell.sick.ru> <p06002014bc6a36fcdf29@[10.0.1.3]>
next in thread | previous in thread | raw e-mail | index | archive | help
At 14:24 02/03/2004, Brad Knowles wrote: >At 3:59 PM +0300 2004/03/02, Gleb Smirnoff wrote: > >> Haven't you understand? I'm the "person who has real-world experience >> in running zebra in ISP environments with multiple upstreams and taking >> full views". > > Do you have multiple connectivity to two separate metro area=20 > exchanges, with multiple upstreams at each? Most large cities are lucky= =20 > to have a single major metro area exchange, and the author of bgpd for=20 > OpenBSD works at an ISP located in Hamburg which is lucky enough to have= =20 > two major NAPs, and he has multiple connectivity to both. He was the one= =20 > ragging on zebra/quagga. Among other things, he said he had real=20 > problems keeping sessions up with zebra/quagga when neighbors were= flapping. I know some small/medium ISP in France how are 100% zebra (or quagga) with= =20 at least 2 full net table with at least 130000 pfx When said small they have at least an average of 10-20 Mbps. As far as i=20 know one is 100% zebra more than 100 Mbps and seems stable. On our side, we have a Zebra with receving 2 full table & +130 peers as a=20 backup router without any trouble now (we have 3 cisco & 1 zebra) 0.92 & 0.93 was unstable. BGPD crash many times on peer routing table or=20 full net table. since 0.93b uptime wasn't broken Our zebra router already handle more than 60 Mbps without problem and so=20 few cpu use that's my cisco router was jalious ;) Currently i make so use for peering with ~40 Mbps since some months and i'm= =20 very happy with it. the design of zebra won't be interrested for have fast & evolutive solution= =20 regarding juniper or cisco 7x00 (except 7100 how is soo slow ;) Regarding 1ghz pc vs cisco 7x00 ... bpg & routing use less cpu on pc...=20 filter and such thing are much faster (7206vxr for example when i done some= =20 test) However where zebra bgp daemon is so crap and so slow is when flapping or=20 when clearing big session. But it's a design problem no ? I would like find very usefull to have bgpd integrate into base system. But= =20 maybe i'm not very objective since i work all the day with AS/Routing. Regarding interfaces that's PC router can deliver, All major carrier deliver FastEthernet / GigaEthernet or 10 Gig... OCx or Ex or Tx are mostly dead in Europe for deliver any ISP. Except=20 Global Crossing how was the last in France, *all* major carrier deliver=20 FastEthernet or GIG. All gix are in E/FE/GE (lynx / ams-ix / fr : parix / freeix / sfinx etc...) Regards, --=20 Arnaud Pignard (apignard@frontier.fr) Frontier Online - Op=E9rateur Internet
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?6.0.3.0.2.20040303054929.06bc9988>