Robert Davies <robert@statsresearch.co.nz> said:
> This answers some questions raised by some follow-ups to
> my previous
> postings (public and private), see below for earlier
> posting.
>
>
> Does the software upgrade to the USR 9105 make any
> difference? The version
> is now 1556_011604-2.12L.01.a0_23a and it makes no
> appreciable difference.
>
>
> What are the line-speeds reported by the modems?
>
>
> Nokia M1122
> Downstream Upstream
> Attainable bit-rate 2752 576
> Actual bitrate 2336 608
>
>
> USR 9105
> Downstream Upstream
> Attainable rate (Kbps) 3808 928
> Rate (Kbps) 3520 832
>
>
> The difference between these figures and the observed ones
> below may be due
> to the time it takes the packets to wend their way through
> Windows XP's new
> fire-wall or, maybe, you can never get the rates the
> router claims. I
> don't think the differences in this case are large enough
> to get excited
> about.
>
>
> What about time between retrains? The log over 15 hours on
> the USR 9105
> showed one retrain following what I interpret as a short
> ADSL outage. So
> I don't think retrains are an issue.

I was busy when this first came up and then forgot about it
but I think I was the original person who requested the
info.

Craig was right, I was talking about the line speeds
reported by the modem. As Craig has pointed out, to my
knowledge there is no reason I know of any decent modem
should not be able to achieve the line speed it's connected
at. Usage tests on FTP servers and such are likely to be
somewhat unreliable since there are other factors which can
affect the speed. A sustained transfer over a long period
(say half an hour or so) during off peak times (3 am maybe)
repeated several times should give a consistent value but
since most modems report line speeds, it probably isn't
necessary to go to so much trouble.

I now have a M1122. The connected line speed varies from
2600-3000kbps down and about 800 up. I don't think there has
even been a retrain. It generally connects at the lower end
of the spectrum but is variable (at least in the 10x or so I
tested it). I do have the latest firmware.

A friend also tested his DSE ADSL router here and it
connected at around 3080kbps. Whether it always gets this
and whether it can maintain this speed I don't know, it's
not been tested.

For future reports, it might also be interesting to see
attentuation and SNR margins (and any other statistics that
are relevant to line conditions).

As an aside, the reason your actual measured bitrate is
lower then the connected bitrate is because of overhead.
I've just found this out (even though I've observed it for
about 2 years) but it's actually quite widely reported (do a
Google). There is a ~10% ATM overhead and ~3% TCP/IP
overhead. Based on what I've read, ATM uses 53 byte cells,
of which only 48 bytes are data. This gives 10.417%
overhead. For AAL5, there is also a 8 byte SAR header for
every IP packet and so padding may be necessary depending on
packet size.

There are a few ISPs (some Aussie business connections I
think and ADSL in Malaysia) which limit your line speed by
limiting the maximum connection rate. This of course means
you do not actually achieve that rate in transfers due to
the overhead. This also means that you'll need to have a
line speed of about 2400kbps to be able to achieve the
planned 2mbps plans.

These sites may help:
http://sd.wareonearth.com/~phil/net/overhead/
http://www.techfest.com/networking/atm/atmaal.htm
http://en.wikipedia.org/wiki/Asynchronous Transfer Mode
http://www.ietf.org/rfc/rfc2364.txt

-- 
This message is part of the NZ ADSL mailing list. 
see http://unixathome.org/adsl/ for archives, FAQ, 
and various documents. 
To unsubscribe: send mail to majordomo@lists.unixathome.org 
with "unsubscribe adsl" in the body of the message