Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Leolo]

Hi everyone,

Please, I need help to understand how exactly is uplink speed measured.

I'm using manual uplink speed instead of automatic. My ADSL2+ modem synchronizes at 8192/1024 kbps, but that's not the usable bandwidth. The losses caused by overhead are very clearly explained here:

http://pflog.net/dsl_overhead/

I've also manually set the value of "Connection Type" to xDSL, instead of auto. (I prefer manual settings)

The HELP web page of my router says the following:

Choosing xDSL or Other Frame Relay Network causes the measured uplink speed to be reported slightly lower than before on such connections, but gives much better results.

Does the value of "Connection Type" also affect the uplink speed when it's manually set?

Or is the value of "Connection Type" ignored when uplink speed is set manually?

Should I put 1024 kbps in manual uplink speed?

Or should I put the usable bandwidth in uplink speed (approximately 870 kbps) ??

Thank you very much,
Kind regards.

[EddieZ]

Just use 1024 when you have that upload bandwidth. I also have an ADSL (XS4ALL) 8/1 subscription. On manual I give it 1024 and the auto makes it 920 or so. But manual is the best setting in my experience.

[Leolo]

Thanks for the reply, EddieZ, but that's not enough for me. I'd really like to have a meaningful explanation.

I've made a few tests with different values, checking the actual speed with a website that gives me very repeatable results. It's located in Spain, the country where I live.

http://www.testsdevelocidad.es/medir-jazztel.html

These are the results (The first value is the number I entered in my router. The second value is the upload speed measured by the website. They are in kbps. I tested three consecutive times for each value, to make sure the results were repeatable)

1024 -> 864
1000 -> 864
980 -> 865
960 -> 865
950 -> 858
940 -> 853
930 ->  843
920 ->  831

The download speed was always 6960 kbps (with a small error of +-1 kbps)

The optimal value is probably somewhere between 950 and 960 kbps (I need to do a few more tests to really nail it down).

But to me there's one clear conclusion: if you have an ADSL modem, you should not put the upstream synchronization values given by your ISP. It will only confuse your router, because that will make it think that you have a connection faster than it really is.

It's clear that the router doesn't take into account the overhead caused by the encapsulation of Ethernet packets into ATM cells.

But, the question now is: does the router take into account the overhead caused by TCP/IP headers?

Kind regards.

[lotacus]

I'm by far no means an engineer so my guess is moot but, one would say that tcp headers are part of the encapsulation when the PPOE modem establishes an ATM connection.

I suppose it may be possible through some sort of equation to see how it reacts if you choose a source other than xDSL and put your results from your first test against the results from the second test.

The only way you are truely going to know is to reverse engineer the firmware and see what equation (if any) is implimented per connection type.

[EddieZ]

It's clear that the router doesn't take into account the overhead caused by the encapsulation of Ethernet packets into ATM cells.

I guess the setting Framedelay Y/N will do that.

But, the question now is: does the router take into account the overhead caused by TCP/IP headers?

Kind regards.

The automatic bandwidth setting will give you a lower setting to ensure that the TCP/IP headers are taken  into account. That's my idea about the mechanism.

I think you should do testing with a more controled environment if you want to be scientific and exact up to the last kb. You are not taking into account the WAN latency and ISP variables. Those can vary too.