[Difx-users] VDIF Benchmarking {External}

[Phillips, Chris (S&A, Marsfield)]

Hi Paul,

I would *highly* recommend you initially test vlbi_fake is producing the VDIF data you expect. Try running it independently with some sort of VDIF (or generic) UDP receiver to test. I have some of my own code you can use if you don’t have anything yourself.

A few other suggestions/comments:


  *   Avoid 5000 byte VDIF frames. That is some hang up from Mark5B 10K frames I think and has no purpose. Something closer to 8k (e.g. 8000 pr 8192 + header) makes more sense. Try “9000”.
  *   Vlbi_fake chooses a frame size which makes sense based on the data rate AND networking parameters, the 5000 probably include VDIF header and maybe network headers (I forget). You do need to setup. DIFX to match the frame size vlbi_fake uses, so some experimentation may ne needed
  *   Are you sure “Jumbo frames” are enabled (ie 9000 byte MTU enabled everywhere)
  *   Make sure the “threads” usage matches what you expect

Cheers
Chris



From: Difx-users <difx-users-bounces at listmgr.nrao.edu> on behalf of Paul Harrison via Difx-users <difx-users at listmgr.nrao.edu>
Date: Tuesday, 30 November 2021 at 10:50
To: difx-users at listmgr.nrao.edu <difx-users at listmgr.nrao.edu>
Subject: [Difx-users] VDIF Benchmarking {External}
Hi,

I am trying to do some benchmarking of DiFX 2.6.2 with the aim of sizing a cluster for e-MERLIN - I am attempting to run some VDIF tests on a cluster that I have access to with some 1Gb/s interfaces - so I am using as my starting point a 1Gb/s EVN NME - eMERLIN usually supplies VDIF a single 64MHz IF band 2 bit sampled with 2 threads. I am trying to send live VDIF with vlbi_fake


vlbi_fake -H ${ip[$i]} -year 2020 -dayno 305 -time 11:07:00 -novtp -p 12345 -bandwidth 64 -framesize 5000  -udp 5000 -vdif -d 300 -b 2 -nthread 2

but mpifxcorr just sits there - well it says things like

2021-11-29 18:11:46,181 DiFXAlert INFO    MPI[-1] compute-0-12.local libmark6sg   libmark6sg started

when it detects that the stream has started - I think that I have got the frame sizes correct as when they were not correct mpifxcorr was saying so (I had started with the “default” 5000 and then noticed that vlbi_fake was announcing that it was doing a framesize of 4096)

I enclose the contol files, and I would be grateful if anyone has any tips for getting this to run

Cheers
        Paul.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listmgr.nrao.edu/pipermail/difx-users/attachments/20211130/d2bb0997/attachment.html>