Draft: Added persistreco to rare charm lines

Persist reco was enabled for the rare charm lines. After discussions, it was decided this is needed for future developments and analyses, at least in the beginning of data taking. Fyi @jbrodzic @mfontana

added Charm hlt2 selection labels

added RTA label

Hi @dmitzel,

I see your point. However, having persist reco for the whole set of lines ((2 +- 0.4) kHz rate if I'm not wrong, see !1216 (merged)) might not be sustainable. Should we move them to sprucing? I see this as the only option unless you plan to tighten the selections or move back to normal HLT2 lines very early during data taking.

What do @mstahl @lpica @roneil @fsouzade think?

I think "moving" to sprucing is not necessarily the most productive way to approach the choices in front of you.

The key question is whether a certain rate can go to disk or whether it will go long-term to tape, and be spruced to produce a smaller sample which can be kept on disk long-term. 2 kHz of persist reco is certainly a high rate to keep on disk indefinitely, as it corresponds to between 2-4% of our total allowed Run 3 data volume. So for sure if you have persist reco at rates of (kHz) a sprucing step will be needed. This is a good general rule of thumb for any lines, with some special cases with respect to calibration lines and so on.

For b-physics you can have inclusive triggers and "move" exclusive lines to sprucing using as input the events selected by these inclusive triggers. For prompt charm you (almost) always need HLT2 lines and then sprucing is there to regulate the disk/tape balance.

Does that make sense?

Many thanks for the comments. The 2kHz is the total rate for a module of >100 exclusive lines (none of the individual lines reaches ~1kHz). As you say, affording more inclusive lines isn't really feasible for prompt charm. Since we have not yet fully understand (semi)leptonic rare charm decays in all details, in particular when it comes to electrons, we would like to maintain flexibility for future studies and developments.

I am not sure what the best strategy is here. Exclusive Hlt2 lines followed by exclusive sprucing lines (essentially copies) to ab able to regulate the disk/tape balance?

I think at sprucing level you will need to be more selective on what information you want to save (e.g. only ECAL clusters for electrons, or tracks and clusters within a cone) and what information you can discard. Otherwise you could aim at selective persistence already at HLT2. @roneil is working on this for spectroscopy lines and can provide more details

In any case note that in 2022 there is no question of throwing detector data away since we have to commission this first. So this is not a problem for today, but you should focus on delivering something within an order of magnitude feasible today and planning a properly scalable solution for 2023.

mentioned in issue #421 (closed)

mentioned in issue #426 (closed)

mentioned in issue #432 (closed)

Is this ready to be tested ?

ping

Before testing the MR, we first need to make a discussion offline and decide whether this is really the way we want to go (see discussion above).

ok fine. Maybe you want to mark it Draft then ?

Sorry - I wasn't aware the MR had still been marked as ready. Now back to draft. We will come back to this once we have resolved the issue.

Any news ?

ping

mentioned in issue #444 (closed)

marked this merge request as draft

Can this MR be closed, @dmitzel? Once Charm and/or RTA have decided the best way forward (e.g. which lines should have selective persistence vs. persistreco, where will sprucing be needed), we could open a new one.

ping

closing

closed