TO:   All GCN Socket Sites
RE:   Improvement in BATSE-FINAL packet distribution times


Some of you may have noticed that the arrival times for the BATSE-Final
Notices (type=22) were arriving at your sites about 110-125 sec after
the arrival of the BATSE-Original packets (type=1).  The time delay
should be about 35-40 sec after the type 1's.  When originally set up,
the program scanned the BATSE telemetry stream for 32 sec after the end
of the Original processing to integrate the lightcurve to get more counts
to make a better location calculation.  Which it does, but a problem developed
over time with all the additions of news sites receiving any of the various
forms of e-mail Notices (e_mail & pagers (long, short & subject-only) formats).
There became so many sites receiving e-mail-ed Original Notices
that the processing time slowed the processing/sending of the Final
Notices and Packets.  The program did not start to process the sending
of the Finals (of any distribution method) until it had finished with
sending all the various Original Notices (of all the methods).

So I figured out a way to separate the socket packet processing
from the email notice processing within the GCN system.  I restructured
the big loop that handles all "sites" so that now the email commands are
passed off into a batch file which is spawned off and executed asynchonously
(at the END of ALL sites processing) from the socket stuff.  So now
the huge amount of email processing from an Original Notice does not slow up
the Final socket packets.

The new version is running right now (has been since 02 May 99).  The results
are that the time delay (after the type 1's) is 40-55 sec.  There is still
5-15 sec additional delay due to cpu loadfactor in processing all
the Packets and Notices (even when split off into a lowest-priority batch
processing).  This 5-15 sec delay will be reduced back to the nominal
few seconds when I upgrade to a new faster computer (in a couple months).

If you have any questions about this change (or anything about GCN)
please do not hesitate to ask.

Sincerely,
Scott

Scott Barthelmy                      NASA-GSFC, Code 661, Greenbelt, MD   20771
PHONE:  301-286-3106                 (work)
FAX:    301-286-1684                 (1st choice,  -1682 2nd choice)
EMAIL:  scott@lheamail.gsfc.nasa.gov
PAGER:  1-800-SKY-PAGE, PIN 2618712  (by phone, voice menu/instructions)
PAGER:  2618712@skymail.com          (by email, 240 characters max per message)
WEB:    http://gcn.gsfc.nasa.gov/gcn