Date: Thu, 19 Mar 2009 16:35:31 +0100 From: Christian Meutes <christian@errxtx.net> To: Christopher Arnold <chris@arnold.se>, Peter Spekreijse <peter@spekreijse.net> Cc: Steve Bertrand <steve@ibctech.ca>, isp@freebsd.org Subject: Re: ISPs? Message-ID: <17AF069575D7E2B24912CFBA@tok> In-Reply-To: <alpine.BSF.2.00.0903191554160.5077@localhost> References: <153046.19925.qm@web63901.mail.re1.yahoo.com> <907077794.20090317173752@homelink.ru> <49C05E35.8070609@ibctech.ca> <001501c9a795$07058de0$1510a9a0$@com> <49C1C3D0.5060304@neely.cx> <CC6BF6C0-D134-4DE6-9D47-17E01AA71BBB@ekalb.net> <5F9EF08A583352985E262800@tok> <49C24561.5090301@spekreijse.net> <49C253FE.3010408@ibctech.ca> <49C2583D.30502@spekreijse.net> <alpine.BSF.2.00.0903191554160.5077@localhost>
next in thread | previous in thread | raw e-mail | index | archive | help
I am also very interested in pps with IMIX type of traffic you are having. How does FreeBSD with $preferred_BGP_daemon scales on IX with dozens or=20 hundred of peers and having two or more full feeds? How fast is the convergence in IGP/BGP? And how do you guys do=20 configuration changes without interrupting everything (reloading daemon)? --On Donnerstag, 19. M=C3=A4rz 2009 15:54 +0100 Christopher Arnold=20 <chris@arnold.se> wrote: > > > On Thu, 19 Mar 2009, Peter Spekreijse wrote: > >>> Peter Spekreijse wrote: >>>> We do use it for routing, using FreeBSD, booting from flash, running >>>> completely in RAM. We have created a solid state BGP/OSPF router with >>>> FreeBSD. Our border routers run Quagga (bgp and ospf) but we are in = the >>>> process of moving to OpenBGPD / OpenOSPFD. Our internal routers = already >>>> use OpenOSPFD. We are using Network Appliances from portwell as >>>> hardware (8 * 1 Gbit/sec ethernet). We're in process of testing other >>>> appliances. >>> > What thruput and PPS are you seeing on theese? > >> We've two full bgp tables in our border routers (275K+ prefixes per >> table). The quagga version we use get's real busy if one of our BGP >> peers disappears suddenly. It starts recalculating the routing table >> and neglects the BGP sessions. Sometimes the other BGP sessions >> time-out, then we lose every route. >> > Opps very bad... > > Have you tried going SMP so one CPU can dio the BGP thread and the other > forwarding and updating the route table? > > Or are you using polling(4)? In that case have you tried to give more % > of the CPU to the userland processes? > > /Chris > > -- > http://www.arnold.se/chris/ > _______________________________________________ > freebsd-isp@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-isp > To unsubscribe, send any mail to "freebsd-isp-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?17AF069575D7E2B24912CFBA>