[132130] in North American Network Operators' Group
Re: Low end, cool CPE.
daemon@ATHENA.MIT.EDU (Joel Jaeggli)
Tue Nov 16 13:29:58 2010
Date: Tue, 16 Nov 2010 19:36:10 +0800
From: Joel Jaeggli <joelja@bogus.com>
To: Eugen Leitl <eugen@leitl.org>
In-Reply-To: <20101112153049.GA28998@leitl.org>
Cc: NANOG list <nanog@nanog.org>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
On 11/12/10 11:30 PM, Eugen Leitl wrote:
> On Fri, Nov 12, 2010 at 10:10:30AM -0500, Jason Lewis wrote:
>> Everytime I'm in the market for a device like you describe, it comes
>> down to the limitations of consumer devices. You can't get all those
>> things in a low cost solution. I end up rolling my own. My latest
>> system is this http://www.supermicro.com/products/system/1U/5015/SYS-5015A-PHF.cfm
<snip>
>> , with Endian http://endian.com/en/community/download/ and an
>> additional dual port nic. With all the parts (HD,NIC) it's under
>> $400.
>>
>> It's an atom board, so you could put whatever you wanted on it. I
>> have a 50mbps net connection and it doesn't have any issues.
>
> Works well on GBit/s as well. I haven't measured the throughput
> yet, though. Should be ~500 MBit/s, assuming a single Atom core
> is about equivalent to a Pentium 3 at the same frequency.
An atom should easily be able to forward some high fraction of a gig
between two pci-e 1x connected interfaces certainly in the soho context
such a box can do ipsec at farily reasonable rates as well.
Regarding equivalence to a PIII an atom is a scalar rather than super
scalar device. it is slower clock for clock than a pIII but there are
also multicore variants and of course they run faster at loewr poper
consumption rates than the equivalent PIII derived embedded processor
such as the intel a800
>
