People on Vidyo: Andrew Melo, Bo Jayatilaka, Carlos Perez Dengra, David Lange, David Smith, D Mason, Elizabeth Sexton-Kennedy, Eric Lancon, Frank Wuerthwein, James Amundson, James Letts, Johannes Elmsheuser, Justas Balcas, Laurent Duflot, Marcelo Soares, Oxana Smirnova, Riccardo Di Maria, Stephane Jezequel, Stu Fuess, Tigran Mkrtchyan, Xavier Espinal, Xin Zhao

next meeting in 2 weeks: feedback from sites to hear impact of the DL model on total cost of ownership
last meeting on video, next will be zoom

* Frank Wuerthwein (Univ. of California San Diego (US)) - Archival Bandwidth at T1s

very first estimate wrt needs (bandwidth)
numbers for CMS and ATLAS, and the sum of them
please refer slides for numbers and their calculation
for CMS, numbers given for RAW processing, RAW form T0, MINI production, and recall for analysis
numbers are soft, and need to be scrutinised and careful revision
for 3 week retention and a max of 1PB tape recall capacity per week, expected processing delays of up to 3 weeks
total 1000 Gbit/sec aggregate archival bandwidth needs across all CMS T1s
for ATLAS, 2 scenarios possible
scenario 1: maximise tape usage
RAW reprocessing and data taking not overlapping: no need to add them (as done for CMS)
MC simulation (not accounted for in CMS)
MC reco (counterpart in CMS: make MINI and NANO - not added to the total)
scenario 2: minimise tape usage
difference with scenario 1: needs different copies
ATLAS-CMS comparison provided (slides for number)
no attempt made yet to understand and reconcile differences
ATLAS sustained peak need 880 Gbps vs 750-1000 Gbps for CMS
proposal to refine numbers and present at HSF-WLCG meeting
are these numbers used in network requirements as well? for the final assessment, these numbers wouldn’t be relevant
RAW size number drive everything -> for CMS decreasing it from 6.5 to 4.4 is already huge improvement
concern that numbers per site are too optimistic
other solutions should be taken into account to deal with these numbers