NDT-DEV email list created

[Sebastian Malecki <>]

Hi!

Thanks for your reply.
I'll look throuht it again and update wiki using your suggestion.
Build graph is good idea. I try to do it. It should be much better than just table as it is now and should answer your question.

Regards,
Sebastian Malecki

W dniu 02.07.2014 20:33, Aaron Brown pisze:

Hey Sebastian,

CurRTO isn’t listed. Is that the same value as “tcpi_rto”?

The MinRTT/MaxRTT/AvgRTT could be recreated by taking snapshots regularly (which gets done when the snaplog is enabled), and then iterating over it, and calculating the stats from the tcpi_rtt variable. You could do something to get the MaxRTO used in the middlebox test.

My understanding is that ‘tcpi_rcv_space’ the advertised receive window which would mean that ‘MaxRwinRcvd’ and ‘CurRwinRcvd’ could be calculated similar to the RTT/RTO stats above.

You list a bunch of variables in “UI”. What does that mean? e.g. is CurRwinRcvd just listed somewhere in the UI as ‘CurRwinRcvd’?

Also, do any of the values calculated get used in other values that get calculated? e.g. if ‘cwndtime’ got used as part of another heuristic, we need to know that since the lack of SndLimTime values will affect those calculations, and anything that’s calculated from those. What would be good would be to build some dependency graph of all the variables that NDT computes back to the web100 variables so that we can see what all would be affected.

Cheers,
Aaron

On Jul 2, 2014, at 7:10 AM, Sebastian Malecki  wrote:

Hi!

Continuing the previous thread, I investigated, which function could work
without web100/10g. It's not much, but in future could be use to create some
light NDT.

As i wrote on wiki page
(https://code.google.com/p/ndt/wiki/TCP_INFOvsWeb100Web10g#Function_work_witho
ut_Web100/10g) it,s:
  - C2S throughput test
  - S2C throughput test
  - Detect S2C & C2S throughput test: Packet queuing
  - Simple firewall test (after little change in calculate testTime)
  - Connection type (in client)
Rest is depend from Web100/10g specific variables. 

Regards,
Sebastian Malecki