[112803] in North American Network Operators' Group
Re: real hardware router VS linux router
daemon@ATHENA.MIT.EDU (Deric Kwok)
Wed Mar 18 15:52:10 2009
In-Reply-To: <7BF22750-2082-4169-A5B2-8E61E20FAE33@daork.net>
Date: Wed, 18 Mar 2009 15:51:59 -0400
From: Deric Kwok <deric.kwok2000@gmail.com>
To: Nathan Ward <nanog@daork.net>
Cc: nanog list <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
Anymore success to use multiple CPU to bind NIC to increase the performance
Thank you
On Sun, Feb 22, 2009 at 4:52 AM, Nathan Ward <nanog@daork.net> wrote:
> On 22/02/2009, at 8:27 AM, Leen Besselink wrote:
>
> If you had to choose, it's probably smarted to go with OpenBSD, it has a
>> lot better integration of packet filter, bgpd-daemon, ospf, vrrp-like,
>> etc.
>>
>
> If you have one eBGP session in your whole network, sure.
>
> However if you have more than one, BGP cannot do the "Prefer the path with
> the lowest IGP next-hop metric" thing, as OpenBGPd does not know metrics
> from OpenOSPFd. Someone commented that OpenBSD would be able to do this soon
> as metrics were added in to the routing code in -current, but I have not
> tried this personally and a quick couple of queries on Google didn't reveal
> anything other than internal OpenOSPFd stuff.
>
> I have however used OpenBGPd and OpenOSPFd with great success on routers we
> put at single-homed customer sites for a small business-only ISP I used to
> work at. We used BGP communities to put prefixes in to PF tables, and then
> shaped and accounted based on that. (Here in NZ we have a few thousand
> domestic prefixes, which transit to/from is often cheaper than transit
> off-shore).
>
> --
> Nathan Ward
>
>
>
