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

20040614: LDM related Question (cont.)



>From: Tarun Banka <address@hidden>
>Organization: CSU
>Keywords: 200406140800.i5E807tK001480 LDM

Tarun,

>Thanks for the info.

No worries.

>I saw some latency traces using the link provided in the 
>following mail. I could notice that lots of variation in latency for 
>sunset.atmos.colostate.edu on any day. I could see latency in the range of 
>8000 seconds or so, could you please tell what's the reason for this high 
>latency.

The latency problem was related to something at CSU/the link from UCAR
to CSU.  Compare the latencies for the same feed(s) to other machines
downsteam of thelma.ucar.edu.  Here are a couple of comparisons for the
NNEXRAD (NEXRAD Level III products) feed:

sunset.atmos.colostate.edu
http://my.unidata.ucar.edu/cgi-bin/rtstats/iddstats_nc?NNEXRAD+sunset.atmos.colostate.edu

emo.unidata.ucar.edu
http://my.unidata.ucar.edu/cgi-bin/rtstats/iddstats_nc?NNEXRAD+emo.unidata.ucar.edu

rossby.met.sjsu.edu
http://my.unidata.ucar.edu/cgi-bin/rtstats/iddstats_nc?NNEXRAD+rossby.met.sjsu.edu

Another topic:

One thing that the LDM needs in order to be able to deliver data
reliably is accurate clocks.  A example of strange latencies results
when the receiving LDM's clock is not being set.  Here is one example:

rainbow.al.noaa.gov
http://my.unidata.ucar.edu/cgi-bin/rtstats/iddstats_nc?NNEXRAD+rainbow.al.noaa.gov

The stair step pattern in the latencies shows that the clock on rainbow
is continuously getting worse.  (I will be contacting the LDM admin for
rainbow to get this corrected).

>For some real-time applications this much latency might not be 
>acceptable,

Understood.

>As far as I understand latency here means delay in transferring 
>data from the Thelma to the Sunset machine.

The problem was at CSU, not with the relay being provided by
thelma.ucar.edu.  If the problem was due to network outage or router
glitches at CSU, any method of getting the data over the Internet would
be just as bad.

>Regards,
>Tarun

Cheers,

Tom
--
NOTE: All email exchanges with Unidata User Support are recorded in the
Unidata inquiry tracking system and then made publically available
through the web.  If you do not want to have your interactions made
available in this way, you must let us know in each email you send to us.