[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[CONDUIT #ADU-250071]: CONDUIT latency graph

This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.

  • Subject: [CONDUIT #ADU-250071]: CONDUIT latency graph
  • Date: Mon, 07 Aug 2006 11:12:54 -0600
> Greetings,
> 
> Do you keep an archive of CONDUIT latency graphs?  We are currently
> having problems in receiving the NCEP GFS grid files (most of the
> files are incomplete and variable size), but the NCEP NAM files we
> are also pulling off the feed are coming in fine just prior to the
> GFS group.   The latencies for the GFS files are currently exceeding
> 12000, and this seems exceptionally high compared to other hosts we
> have scanned from around the West.  This problem appears to have
> begun sometime during 29-30 July, and thus we were wondering if you
> might still have a latency graph for this period.  Otherwise, I
> suppose we can try to examine latency numbers from the log file.  If
> you have any thoughts or suggestions regarding our problem, those are
> welcome too.  Many thanks.
> 
> -Tim Brown

 Tim,

We don't pre-generate any gif images, but I can tweek the stats page
to do a longer time series plot. We keep about 15 days of the stats available.

I have posted plots for ldmhost and playa at:
http://www.unidata.ucar.edu/staff/chiz/gifs/dri_ldmhost_conduit.gif
http://www.unidata.ucar.edu/staff/chiz/gifs/dri_playa_conduit.gif

Looks like reception was never great since ldmhost had a history of large 
latencies, 
but they did get worse on july 29.

Your feed request looks like you are just asking for NAM 212 and GFS products.
Also, you should use the name idd.unidata.ucar.edu rather than thelma.ucar.edu
although the thelma alias will go to the idd cluster, and this should be 
unrelated
to your latency issues- but the idd.unidata.ucar.edu name will be portable.

It looks like the data flow initially comes through with no latency, but 
shortly thereafter
the latency just climbs off to whatever the maximum duration back in the 
upstream queue is.
The pattern appears to be replicated with playa which you just started up this 
past weekend, 
with a newer LDM distribution, and CONDUIT being the only request on that 
machine. Thgat
should eliminate all question about your LDM boxes themselves.

I would suspect that your campus IT department is throttling your network so 
that your
bandwidth is limited. Can you find out if that is the case, and as a 
suggestion, your campus
should pass through all traffic to port 388 uninhibited.

Steve Chiswell
Unidata User Support
> 
> 


Ticket Details
===================
Ticket ID: ADU-250071
Department: Support CONDUIT
Priority: Normal
Status: Open