Core Minutes 9/11/2007
Present: Joanne Bogart, Anders Borgland, Johan Bregon, Toby Burnett,
Richard Dubois,
Tom Glanzman,
Navid Golpayegani, Heather Kelly,
Michael Kuss,
Bryson Lee, Chuck Patterson, Igor Pavlin,
Leon Rochester, Eric Winter
- Meeting reports:
- International Finance Committee aka "Begging for money": (Richard)
Success!
We were promised everything we asked for: 250 TBytes of disk and 400
computing cores. It seems BaBar overestimated its needs by about a factor
of 2 and we can have some of the excess; exactly when is less clear.
- ASI Science Data Center (Richard)
There was a meeting to discuss
Web tools to be used for displaying catalogue information, ASP results and
multi-wavelength campaign data. Gino Tosti will work with Paolo Giommi and Roberto Primavera
(the engineer doing the actual work), installing and maintaining
tools at SLAC. A version serving only the public data will go through the
GSSC. A delegation should come to SLAC within a month. [Thanks to
Richard for filling in the many ?? in my notes.]
- CHEP (Toby) He'll give a more complete report next week. Here
are some highlights:
- EVO: Chatting with various people, Toby found EVO is not universally
admired; other groups have the same sorts of problems we've seen. He didn't
hear the official EVO talk.
- ROOT: Dramatic changes; will be summarized in more comprehensive report
for next week's Core meeting. No longer require that all objects inherit
from
TObject
; use of abstract interfaces; leaner and meaner
(applications end up being smaller).
- Power and cooling: Several talks on this, a prominent issue for many.
(Richard) SLAC has commissioned David McFarland to look into this issue
here; they're coming up with a 2-year plan. The power load will go into
black boxes in the parking lot, allowing an overhaul of facilities in
the main building. It should all be very nice at the end of the 2 years;
getting there may be less pleasant.
- Memory and cpu architecture: Multi-cores are IN. Memory can be a
bottleneck.
- Grid: Still big, but no longer the king of hype.
- Dan's talk: Went well; got mentioned in Summary
- Gaudi support of VS2005: Toby had a chance to ask Pere Mato about this.
The response was less than satisfying: there isn't any support, and no
plans to do anything about it in the near future.
- GlastRelease (Heather)
- OBF
GR v12r3 is the first tag with the new
OnboardFilter code and related updates
for the TDS and so forth. Have seen abnormal terminations in System Tests
after many thousands of events. Anders tried a data run which failed after
about 1000 events. Diagnostics are pointing to the new OBF as the culprit;
Tracy is investigating. [late-breaking news: OBF crash is gone in GR HEAD]
- Geometry The ACD geometry is waiting in the wings; Eric C.
is eager to see Sys Test results for it. If the OBF problems don't get
resolved soon, may back that out and put the new ACD geometry in.
- Trigger Config Svc (Anders) would like to see some reorganization.
Would like a job option to decide whether to use GltWord or GemWord
for calculating the trigger engine; and
a second job option to apply (or not) prescales and engine inhibit.
These things should have
reasonable defaults. Martin leaves on vacation tomorrow; maybe he
can get it done before he goes.
[As of this afternoon, done and tagged.]
- EbfWriter (Heather) Running real data through GR,
EBFwriter will reject events with nor trigger
bits or just ACD LO. She has asked Brian Wiener to provide a way to turn
this filtering off; it's not what we want for real data.
- mootCore (Joanne) The immediate goal (for OktoberTest)
is to make GEM configuration
available to Gleam in a moderately realistic manner. Martin Kocian wrote code
to parse the relevant LATC files some time ago.
mootCore handles the connection to the MOOT database
and packages up queries needed by Online and Offline. It does everything
or nearly everything required for this exercise. The one substantial piece
still mostly missing is the Gaudi-conversant part which will handle storing
information in the TDS. See this Confluence page for more details and related
issues.
- Gaudi History Service (Heather) GlastSvc
has been updated to use this service, but it's not yet in a GR tag. She'll
promote it soon; Tom is an eager customer.
- Event collection (Heather, for David Chamont)
As we discussed at last week's core meeting, David requested to modify how
we refer to the event collections, preferring to use the name composite
event list. This idea has been blessed by Eric C. David is now migrating
the code from metaRootData into an unused package
named rootUtil. He hopes
to have an initial version committed to CVS this week. The
rootUtil
package will contain general ROOT I/O routines, useful for the skimmer, and
all code that is interested in manipulating the composite event lists.
[Thanks to Heather for supplying this text. ]
- MRStudio: (Heather) Riccardo suggests we
have another meeting late this week [looks like it will be Friday
at 11 AM PDT], at which point he expects to have a new tag fixing
observed problems on Windows and Linux.
- SCons: (Navid)
Overriding of packages is working. He would like to make it a little nicer and
easier to use, though. (Heather) Navid will give a more complete report
on status next week.
- AOB
- New hires (Richard)
The candidate interviewed recently for the Goddard
slot looks good
and an offer went out. No response yet; it's known that he's considering
other positions in addition to this one.
A prospect for Larry Wai's
old position with Science Ops was interviewed. He's enthusiastic and
looks very good.
R. thinks an offer is being generated.
- C & A news (Richard) The move
of TVAC to NRL will generate at least 6 weeks of delay. That
would give us a launch date in late March.
There was a request from
Jonathon Ormes for some new variables (McIlwain B and L parameters) to be
added to AnalysisNtuple. They're already in the history tuple, but not
currently stored for every event. Responsible parties will take action.
- YACh (Yet Another Checklist)
(Richard) for fall. Note that the OktoberTest is not
"moving to the right", regardless of what happens elsewhere.
- Pipeline help ..is on the way, in the form of Francesco,
Nicola and Johan Bregon.
- Systests (Leon) Observations after a preliminary look at
last night's run:
- No plots!
- Everything ran to completion
- Tests are going in and out of SAA. Is this what we want?
- Heather discovered a difference in the directory structure. Where
there used to be a directory named Linux there was one called
Solaris
J. Bogart, Last Modified:
01-Jun-2010 15:46:32 -0700