ndt-users list created

[Richard Carlson <>]

Claudio;

As the note below indicates the theoretical limit is an estimate of  
how a TCP connection would operate over an ideal network.  There is a  
well established formula that uses the packet size, rtt, and loss rate  
to predict this limit.  The NDT server extracts those values from the  
web100 data captured during the s2c test and computes the limit value.

In some cases, the theoretical limit will be less than the measured  
value (as you note below).  This may be caused by such factors as:

1) the NDT server uses the average RTT for this calculation.  RTT  
varies during the course of this test due to packet queueing effects.   
If the average RTT is significantly higher than minimum value, the  
theoretical limit will be lower than it should be.

2) The loss rate is inferred by looking at the information derived  
from the receiver.  If the calculated loss rate is off, the  
theoretical limit calculation will be off.

3) the formula is based on TCP's response to congestive loss.  When  
non-congestive loss events occur, such as over home broadband  
connections, then the theoretical limit will be lower than expected.

One of the task on the developers to-do list is to investigate ways to  
improve the accuracy of this calculation.

Rich

On Mar 29, 2009, at 9:19 AM, 

 wrote:

> Dear Sirs,
>
> I don't understand what is the theoretical limit. I found this  
> definition in your materials:
>
> Theoritical Limit - The maximum throughput that a connection could  
> achieve over the end-to-end path. This value is calculated using  
> web100 values for Round-Trip-Time, Packet size, and Packet loss. A  
> properly functioning TCP implementation will never exceed this  
> value. This value does not take the physical network interface  
> speeds into account. This value is meaningful when packet loss rates  
> are large, RTT's are big, or packet sizes are small.
>
> In particular i don't understand how it is possible that the speed  
> of my inbound test results faster ( 2.46 Mbps ) than my theoretical  
> limit ( 1.81 Mbps )
>
> Maybe i have some problem with english!
>
> Thank you so much
>
> Yours faithfully,
>
> Claudio Opizzi

Richard Carlson

1000 Oakbrook Dr
Ann Arbor, MI  48104

P: 734-352-7043
C: 630-251-4572