RE: Data Mining Part of MDC-1

["Douglas Olson" <dlolson@lbl.gov>]

Hi Bruce,

Here is a quick summary of what we accomplished.  We will be analyzing the
log files and have some quantitative results in the next days.
Arie as provided a fairly detailed description of the different
tests carried out (thru Sunday), that you can peruse at
http://www-rnc.lbl.gov/GC/email/archive/msg00600.html.

We managed to run a number of baseline tests for various queries
run individually and sequentially.  This was followed by a series of
measurements in which the queries overlapped, in time as well as
in files touched.  For these cases we had two versions of the
policy module, one called no policy that does nothing to
coordinate multiple queries, and the one that does coordinate
which event ID's are returned to different queries based upon which
files are in the cache and shared by those queries.
I think there aren't any big surprises in the results,
just more the assurance that what you think ought to happen, does
in actual measurements.  

Most of the queries were carried out using a simulator of actual 
analysis code, in which we could adjust the amount of cpu time 
per event.  We did succeed in running one of the STAR physics 
analysis codes in STAF and reading data from Objectivity for 
some of the queries. This at least provided a nice software
connectivity milestone, even if there are not yet the
representative 10% of physicsts beating on Objectivity
for analysis.

I would like to thank STAR for their tape drive over the
past several days, and PHENIX as well for letting us have
exclusive use of their drive from Saturday night until Monday 
morning.

The RCF staff, and John Riordan in particular for making adjustments
to HPSS, were helpful and I would like to thank them as well.

All in all I would say it's been a pretty successful shakedown run.

Doug

> -----Original Message-----
> From: Bruce G. Gibbard [mailto:gibbard@bnl.gov]
> Sent: Monday, October 05, 1998 6:54 AM
> To: Bruce Gibbard <"gibbard "@bnl.gov>; Mark Baker; Barbara Jacak;
> Charlie Maguire; Dave Morrison; Torre Wenaus; Terry Healy; Shigeki
> Misawa; Andy Como; Donald McLeod; Razvan Popescu; John Riordan; Tom
> Throwe; Flemming Videbaek; Edward Nicolescu; Doug Olson; Edmundo
> Garcia-Solis; Peter Jacobs; Jeff Porter; Grace Tsai; Yuri Fisyak; A. K.
> Holme; David Zimmerman; Krzysztof Worniak; Andrzej Olszewski
> Subject: Data Mining Part of MDC-1
> 
> 
> I would like to have each of the groups, which have been performing data
> mining during MDC, send me a brief description of the exercises they
> have performed and some of the results (quantitative if possible)
> obtained.  If you happen to have more lengthy descriptions, I would be
> interested in those too.
> 
> Thanks,
> Bruce
>