[arin-ppml] debunking the myth that Moore's law helps
Dave Feuer wrote:
> ---------- Original Message ----------------------------------
> From: "Michel Py" <michel at arneill-py.sacramento.ca.us>
> Date: Thu, 17 Dec 2009 19:45:34 -0800
>> Joel, the WRT610n does NAT at 300Mb/s? I mean measured in the
>> world, not only manufacturer's marketing?
> In "real word" tests it does about 50% to 60% of that. The older
> 600 was a bit faster. But still not bad for something you can
> pick up for $179 or less at WalMart.
> Sent via the WebMail system at connetrix.com
Here is something that I wrote after we took the opportunity to play
with one, for an hour about a year ago:
> One of our campus departments recently purchased a Linksys WRT610N,
> which is a consumer-grade SOHO (Small Office/Home Office)
> NAT/Firewall/Wireless AP, which cost somewhere in the neighborhood of
> Since the device has a 1/100/1000 UTP uplink, I was asked how to
> determine what the throughput limit of the device was. They didn't
> care much about the wireless part, they bought it mostly for the GigE
> uplink. (And we didn't test the wireless part).
> The WRT610N arrived on Friday, and after several minutes of
> preparation, we used the Windows desktop that happens to be near the
> uhmanoa measurement machine to do a simple iperf test.
> They departmental guys had already installed DD-WRT ( Linux-derived
> open source OS for little routers --
> http://www.dd-wrt.com/dd-wrtv3/index.php ) on it before I got to it,
> but not tested, and although the DD-WRT page says it works with
> WRT610N, we did not get the outside interface to pass traffic.
> So we re-flashed back to the Linksys firmware, and then we were able
> to test.
> As a control, iperf was run between the two test hosts over a 3 meter
> Systimax-D cable, which is supposed to meet or exceed category 6.
> Doing "-r" TCP tests, it was shown that the test hosts reliably
> obtained in excess of 900 Mbps in either direction over a half a dozen
> tests, using just the control wire.
> The WRT610N is a 4-port 10/100/1000 UTP switch, and a NAT/Firewall
> with a fifth port as an uplink. When we tested with one host on a
> switchport, and the other on the uplink port, the max throuhgput was
> 130 Mbps in either direction. When a third machine was added to
> another switch port, the combined throughput was 130 Mbps. The
> bottleneck appears to be at the NAT/Firewall function.
> When 4 hosts were connected to the switchports, leaving the uplink
> un-populated, the built-in switch performed admirably, simultaneous
> iperfs yielded > 900 Mbps throughput.
-------------- next part --------------
An HTML attachment was scrubbed...