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

missing products table

Hi Harry,

I've updated the table at
http://www.unidata.ucar.edu/staff/anne/sunnyLoss.html.  This shows the
middle 14 hours of about 18 hours of logging on each machine.  This time
period contained a multihour period of frequent disconnects followed by
a multihour time of no disconnects.

Regarding latency values, my initial approach was to grab a product from
some particular time and use its latency, sort of a snapshot approach.
These are the values listed under 'latency'.  But, now I think that
approach is too narrow.  In particular, these values don't appear to
correlate with the average and max latency pqbinstats statistics sunny
sent in for that time period.  Perhaps I need to calculate similar
statistics.

I was disappointed to find that data loss did not appear to correlate
with the disconnect problem.  I now think that notifyme, or perhaps the
way I'm running notifyme, is not accurate.  Again, there was a variation
between the total number of products reported by notifyme and that
reported by sunny via pqbinstats.

It's possible that these notifymes may be running further behind, hence
latency values that are consistently higher than those sent by
pqbinstats.  Perhaps the python script contributes to the problem.  The
next step would be to track the latencies across connections to see if
this one is anamolous or representative.  I wish I had more confidence
in these results.

Thoughts or comments?  I'll be meeting with Russ on Friday about these
issues (out of the office tomorrow).

Anne
-- 
***************************************************
Anne Wilson                     UCAR Unidata Program            
address@hidden                  P.O. Box 3000
                                  Boulder, CO  80307
----------------------------------------------------
Unidata WWW server       http://www.unidata.ucar.edu/
****************************************************