ndt-users list created

[Aaron Brown <>]

Hi Bogdan,

Given the very low speed of the tests, NDT should be able to handle that. How long are the speedtest.net, and wget downloads? If they're more than 10 seconds, I could maybe see that accounting for the speed difference if the link has some loss. In all three cases, the download test is going to be the same thing, a loop of reads off a TCP socket, so I can't imagine there'd be too much variance in how that's implemented.

Cheers,

Aaron

On Jun 11, 2013, at 11:44 AM, wrote:

Hello

I've been testing a couple of ADSL 80/20Mbps lines using NDT and I'm
consistently getting very good averages/small variations for the uplink speed
(around 16Mbps), but poor estimates (30-40Mbps) for the downlink speed,
regardless of the server used. The performance appears to be consistent for
the same server, but varies from server to server (e.g. testing on serverA and
server, the average on serverA would be 32Mbps and the average on serverB
would be 37Mbps, with <3Mbps variation); RTT varies slightly between servers,
25-50ms. I've tested the lines using both wget and speedtest.net and been
getting consistently around 50-60Mbps, which is rather far from the NDT
estimates.

To check for consistency, I setup an NDT server on a remote network with Gb
connectivity and got around 20-40Mbps NDT estimate and around 45-60Mbps
download speed using wget.

Looking at the packet trace on the client, the transfer goes fine with 2-3
exceptions where it stalls rather badly due to packet loss/recovery.

I've used 3.6.3 and 3.6.5.1 for the NDT clients (using the more recent version
did not seem to make any difference)

Did I miss something in my tests/analysis? Any suggestions?

Best regards
Bogdan

ESnet/Internet2 Focused Technical Workshop
Network Issues for Life Sciences Research
July 17 - 18, 2013, Berkeley CA
http://events.internet2.edu/2013/ftw-life-sciences/