Minutes of the meeting held on the 17.11.99
Present : Dieter Schlatter, Giuseppe Bagliesi, Vincent Lemaitre, Andrea Valassi,
Stefan Schael, Monica Pepe-Altarelli, Guenther Dissertori, Dirk Duellmann
- Dirk Duellmann gave a nice summary on the status of Objectivity/Db
and the related LHC++ software. Version 5.2 is in principle available,
also HepODBMS, but not all LHC++ components are already updated to be used
together with v5.2. However, what we need is there.
With this new version we don't have to upgrade all database files.
Objectivity has released this new versions for all platforms in one go.
Also Linux is there. We have to be careful with which version of RedHat
we want to link (problems with runtime library). Redhat5.1 works.
AMS has been upgraded to multi-threading, people are working on the upgrade
of the related HPSS software. So this will not be available yet.
There exist some upgrades to HepODBMS, such as a new Tag type which corresponds
to columnwise Ntuples. They can be accessed with the old interface
ExplorableCollection.
Finally Dirk gave a nice summary of the experience and problems people
have encountered in BABAR with the AMS and lockservers, and with their
analysis package. For a summary you can have a look at the minutes of the
LHC++ meeting of November 4th..
- Guenther gave a list of things to do in the near future. On the database
side, he plans to do the upgrade to v5.2, also on Linux! In this upgrade,
missing banks should be added and also the MINI banks should be defined.
So if we decide for a large scale analysis, we might write MINIs, and maybe
even apply some preselection, on the basis of the CLAS card or even some
cuts (see talk by Vincent). For a big analysis, sooner or later we have
to test HPSS.
He estimated the storage size needed to write all LEP1 data and MC. It
amounts to about 250 Gbytes for the MINIs.
In the case of a big scale production, we need somebody to babysit it.
There is interest by people from Munich (Stefan Schael reported about that),
so we could envisage to write it on their Linux cluster.
On the analysis side, some utilities have to be added, such as classes for
user-defined object-collections (Monica could work on that), or lepton ID
(Vincent and Vitaliano?) and even b-tagging.
There should also be tests carried out for implementing HTL and TAGs and
analyse them with some tool, such as JAVA or ROOT.
These efforts can in principle start now, they can be done in parallel with
the database work.
It is pointed out that we should introduce the 'cfortran.h' standard for
calling fortran from C++.
- Vincent gave an outlook of the tools needed to perform the single
W analysis. In particular, when concentrating on the x-section measurement,
the analysis is rather simple and does not rely on many tools.
He will evaluate the size of MC samples to estimate the backgrounds.
There we might have to apply some preselection, such as some pmiss cut,
in order to reject unwanted events at an early stage.
- Discussion We will have another meeting in mid-December to see
what the status is. Guenther will try to invest a week or so to study
the upgrade and the move to Linux. Then we will have more concrete ideas
in January what regards a big scale production.
Guenther Dissertori
Wed Nov 17 19:20:38 MET DST 1999
GD