Rongrong:
Presented estimates of disk space needed in the near future for MTD picoDst
Primary vertex - selected from all vertices in the event, not only the default one
- comment on Vz-Vpd difference - it could be asymmetric, take a look at the distribution
- slide 3 - Qx, Qy variables for flow analysis - will be discussed with flow experts
- slide 6 - events with HFT saved will all tracks
disk space - will be difficult to have more disk space at RCF due to lack of funding this year, but that's not impossible ( that's Rongrong's impression from S&C  yesterday), Zhenyu commented that should be no problem with saving all picoDsts at PDSF (based on discussion with Xin)
motivation for all tracks in the event - to have an option to run KF vertex over picoDST
- PDSF - everyone STAR member can get an account and it is fast, we can have issues with disk space for user analysis, conveners will follow up with the management to establish the disk usage policy
- picoDSTs - question from Jerome - how much faster is the analysis using picoDst over MuDst? Xin will try to find the candidate for such test.
- electron identification with BEMC in picoDst -> use StPicoEmcPidTraits
- status of the preparation - code being review by S&C group, should be done soon, MTD group is ready

Zhao - J/psi v2
- v2 study with all run 11 data
- slide  9 - what is the centrality bin - info will be added later
- slide 15  - larger event numbers but lower lower J/psi yield
a few possible issues:
run10: there was HLT trigger running without recording MB events, there was a central trigger, there was an online Vz cut for HT triggers in run 10, so the sampled luminosity is larger in run 11, but luminosity within the Vz cut is c comparable
- no syst. error so far
- why all run 11 points are hight than run 11? Quantify the difference, one possible choice is z-test
- add reference to method used for run 10 and run 11 data combination

Jana - J/psi in U+U
- slide 7: significance is overestimated: should be defined as Sig = S/Delta(S)
Fit looks not really good because there are many parameters. Suggestions: use Gaussian only, fix parameters in the crystal ball function (for example using embedding)
- slide 9: add uncertainty on the signal

Liang: KF vertex:
- for RefMult > 200 GeV, all results very similar
- for RefMult < 50 (peripheral) - significant improvement for KF vertex results compared to original vertex (from Miniuit)
- why there is a difference at the first place? Need to be checked with Yuri, but this could come from a minimization procedure used to find a vertex (for instance chi2 vs simulated annealing, different definition of quantity being minimized))
- suggestion; produce D0 trees with refitted KF vertex for peripheral events (let's say gRefMult < 100) Liang will contact Yuir, make sure that KF vertex finder parameters are optimized and then contact Mustafa for the D0 trees reproduction (original picoDst stay untouched). Liang will forward conversation with Yuri to the HF list.
- comment from Wie - concerns about quality of peripheral data, we should to make sure that we have  embedding for the R_AA if R_CP is not good enough
- comment from Mustafa - let's reproduce all data to not create a mess with different data structures and have default or re-fitted vertex (standard vertex for gRefMult>100, KF vertex < 100).