STAR CRS %tage occupancy: 100%
STAR CAS %tage occupancy:  95%
We will take the lower-end 95% far occupancy efficiency as rule of thumb.
RealTime/CPUTime efficiency showed to be > 98% (100% in some cases). We will use 98% efficiency

Based on production=P16ij,filetype=daq_reco_picodst,storage=HPSS, we get from the catalog

The current MuDST to picoDST size ratio was therefore take as = 7. By the time we add the information the PWG have asked, this ratio will drop to a lower number. We assumed 6.5 (10% for sure for the EMC, a few % for the FMS). Our calculation consider this.

This will give the following summary table
 

I do not have yet the split of option A and B but they change little (the BES years are the challenging years for STAR computing - we will end up with similar requirements). The plan above is for 5 and 4 B events while the guideline is for 6 and 2 (so 9 versus 8). Not sure there is a need for more precision and I would like to merge both 19+20 for the resource estimates as it changes little.

Since we have 8.6 PB of space in Xrootd, we will still not be able to keep everything on disk without additional disk space. BUT if we can transform all MuDST on Xrootd to picoDST, the starting number 8.50 (currently used) would become 1.3 PBytes then we add onl picoDST ... and we can make it through 2018 and keep al datasets.  I will add text to this regard (but ideally, we are not at that stage).

Production readiness for the past years can be summarized as follows [on second thoughts, completion should be given in %time rather tan %tage events as we have done most of the time challenging ones]
 

PHENIX farm utilization by STAR jobs, past year