-----Original Message-----
From: [log in to unmask] [mailto:[log in to unmask]] On Behalf Of Kim, Peter C.
Sent: Monday, January 28, 2008 5:09 PM
To: atlas-department-l
Subject: FDR-1
Dear all,
Here is a brief summary of the BNL-FDR meeting today.
Comments and corrections are welcome.
Cheers,
Peter
* FDR-1 starts in the week of Feb. 4. Raw data size at
1.6 MB/event leads to 70 TB for the week.
(There was a mention that the event size could be 2.5 MB,
in which case the total becomes 110 TB.)
Raw data will be kept at Tier-1 sites on tape.
* ESD and AOD will be processed at Tier-0, and then sent
to Tier-1 sites. Arrival of AOD at BNL is expected with
a 3-4 day delay; first batch arriving around Feb. 8.
Data size for all ESD and AOD in FDR-1 are 43 and 8 TB,
respectively. Tier-2 sites will then get the AOD data
each site subscribed, and have them available for users.
* Average AOD event size is 200 KB, though we heard a number
320 KB last week. Need to clarify this...
* DPD data set definitions, processing (when and where),
distribution schedule are all in a flux at the moment.
Rough estimate for the sum of DPDs is about the same as
AOD data, 8 TB.
* FDR-1 will be followed by 3 weeks of Combined Computing
Readiness Challenge (CCRC). Exactly the same set of data
as in the FDR-1 week will be generated again and again.
No need to keep these data, just to test the data flow
and distribution.
* MC production will continue during FDR-1 and CCRC.
* GRID analysis jobs (pathena) on special queues are being
tested at some Tier-2 sites.
* Tag database are desirable to have at the same time as the
AOD transfer; users can go through the data more efficiently.
As for the SLAC readiness as a Tier-2 site, we seem to be in good shape.
We should subscribe to all AOD data => 10 TB disk space
CCRC exercise and spare room => 10 TB disk space
Wei says we have more than 20 TB available.
Some ESD data or DPD later could be downloaded per needed basis.
Some coordination to be done between MC production jobs
on GRID, PATHENA analysis jobs on GRID, and SLAC-locally
submitted jobs, in the batch queue priorities and slots.
We need more disk space for FDR-2, as the event size will be
larger and the lumi increased by an order of magnitude.
SCS plans to order ~240 TB for ATLAS.
================================================================
|