Subject: Re: OT: Routers
From: Scott Johnson (sjohnson@akghetto.com)
Date: Wed Mar 05 2003 - 23:38:46 AKST
I know this is very shameless, but I have a bunch of 4 port nics for sale if
anyone out there is considering building their own router. 4 port nics come
in VERY handy for such a project.
Please email me off list if interested.
----- Original Message -----
From: "Christopher E. Brown" <cbrown@woods.net>
To: "James Dory" <jdory@gci.net>
Cc: <aklug@aklug.org>
Sent: Wednesday, 05 March, 2003 12.37
Subject: Re: OT: Routers
>
> On Wed, 5 Mar 2003, James Dory wrote:
>
> > We don't have any available old hardware here except maybe one or two..
> > scrounging some up would cost too so I have to keep that in mind.
>
> Even purchasing new x86 linux based routers tends to be alot cheaper
> than comparable Ciscos.
>
> Unless I am way off on my prices, I know where you can get a flash
> based book sized 3 10/100 port ethernet router (about 20% faster than
> a 2621) for about a grand less than CDW lists a 2621 without support
> contract for. That is an industrial type flash booting no moving
> parts LinuxRouter, with more interfaces, hardware watchdog and in a
> smaller chassis.
>
>
> If done right you should be able to do a very nice setup for well
> under 2k per location for all routing and switching gear.
>
>
> > Ok, I got those off the Cisco website I think, but could have been
> > misled easily by that website. I'm not familiar with pps rates anyway..
> > not sure how that relates to speed 'feel' if at all.
>
> They are the figures they list. I was just noting that I was giving
> real world worst case figures, not best case test workbench figures.
> A baseline 2600 will do 40,000pps best case(fastpath), in a very
> specific testing setup, a 486DX-133 with a PCI bus and a pair of EEPRO
> 10/100 will do 65,000/pps (fastpath) with the same workload and
> testbed configuration. (Interrupt mitigation, fast switching, etc
> enabled). I just wanted to compare apples to apples as it were.
> 28,000/pps was a worst case real world figure for the 486DX-133 box.
> One can normally divide the pps figure Cisco gives by 3 to get
> something sane. Cisco quotes pps figures based on the fastpath, and
> *most* of the time a router is running slowpath.
>
> So long as a router is running *well below* its slowpath pps figure it
> really has nothing to do with how it feels. A good rule of thumb,
> worst case *real-world* pps should be no more than 110% of router
> slowpath pps. (Preferred is no more than 90%).
>
>
> You need to figure overall router ability based on slowpath router
> pps, and a routers aggregate bandwidth first, then on available
> interfaces and interface speed. You can easily pack 8 100Mbit-FD
> links in a 7206, for an interface aggregate of 1600Mbit, but the main
> interconnect bus limits a 7206 to no more than 400Mbit aggregate in
> slowpath, and not much more than that in fastpath.
>
>
> You can stuff alot of interfaces in a router, you just have to be
> careful not to bounce off of pps or bus bandwidth limits.
>
> --
> I route, therefore you are.
>
>
>
> ---------
> To unsubscribe, send email to <aklug-request@aklug.org>
> with 'unsubscribe' in the message body.
>
---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.
This archive was generated by hypermail 2a23 : Wed Mar 05 2003 - 23:39:20 AKST