Run2

Run2 developments

See master JIRA ticket ALIROOT-6564 : Muon 2016 reconstruction readiness

• ALIROOT-6562 Add MCH LV values to OCDB

• ALIROOT-6563 Add ClusteringAbort and TrackingAbort flags to the Muon ESD/AOD

• ALIROOT-6615 Reset MUON/Calib/RejectList in RAW OCDB 2016

• ALIROOT-6599 Remove unused OCDB objects

  • not started (but clearly not critical, more a spring cleaning issue)

• ALIROOT-6561 Flag the MUON/Calib/HV OCDB object as already patched for St12

  • done

• MTR efficiencies (to be able to get their effect in simulations)

• Remove in-reco-QA

Add MCH LV values to OCDB

The issue :

• some LVPS were unstable end of 2015. LV trips could not be accounted for automatically offline as the LV information is missing from OCDB

The solution :

• get low voltages stored in the OCDB, like the HV, so we can kill offline the detection elements that suffer from LV trips.

The status :

• setup of the transfert from DCS archive to OCDB (via Shuttle) was quite an epic story but finally succeeded this thursday...
• next step (TBD) is to test the rejection based on LV trips on reconstruction.

Fallback :

• remove runs with LV trips from the good run list
• that's the current situation actually

Add ClusteringAbort and TrackingAbort flags to the Muon ESD/AOD

The issue :

• for some events the clustering (or the tracking) is aborting for some reason (e.g. too many pads hit)
• we don't flag those events so we don't easily get a handle on them

The solution :

• flag the aborted events at ESD/AOD level so they can be studied / rejected later on

The status :

• not started

Fallback :

• check with the reconstruction logs that the integrated fraction of affected events is small
• close our eyes

Remove in-reco-qa

The issue :

• the QA-that-runs-within-AliReconstruction (not the one running on ESDs and giving us the runlists for analysis...) is no longer maintained

The solution :

• amputate the related parts of our code from the repo.

The status :

• Currently some DQM agents rely on some QA classes
• So first the custom DQM agents for both MCH and MTR to be ready :
  • MCH is almost OK (though some testing still to be done)
  • MTR not quite ?

Fallback :

• continue as now to use non-maintained-and-poorly-written code

EVE updates

As part of the EVE development for Run2 and Run3, Jeremi set up a table("Run3 EVE readiness") to follow the progress.

So far we are green.

Run3

Reminder of the context

In Run III, Alice will (mainly) becomes trigger-less and acquire MB Pb-Pb collisions at 50 kHz (instead of ~1 kHz currently). Input data to the readout will be huge (~3 TB/s), and has to be reduced, online, to a more manageable value of 90 GB/s to disk (and 20 GB/s to T0).

This data reduction is the top priority of the Run III reconstruction, Alice-wide.

For the muon detectors specifically the numbers are (much) smaller :

• input to read out ~ few GB/s
• output of EPN ~ same order (unless we drop the pad information completely)

So we can hope to actually perform more than a mere data reduction, i.e. do a full reconstruction with physics-grade quality. That's what the MRRTF group is supposed to deliver at some point.

Muon Run3 Reconstruction Task Force

a.k.a. MRRTF

To tackle the issue, a "task force" was setup end of 2014...

• Wrote a "self-assigned" mandate beginning of 2014 and started a twiki

• Kickoff meeting in oct 2014

• Nothing happened then for about a year

• Then tried to setup (bi) monthly meetings.

• Made a 3 days hackathon in Subatech in march this year.

Moving the snail(s)

Things are starting to move, but still quite slowly.

Conclusion I : team was set up way too early

Conclusion II : hard to get available (wo)man power for the job (i.e. there's always more pressing issues than Run3 stuff)

Conclusion III : there are lots of things to learn, so slow start might actually not be too worrysome...

Reconstruction chain "big picture view"

Reconstruction chain status...

Achieved so far

• Play with AliceHLTWrapper

• Toy device in Go

• Speedup of 100x for MCH preclustering stage

• Ability to stage filtered raw data (on SAF), i.e. keep only muon information in raw data chunks.

  • not limited (nor driven by, originally...) to Run3, but quite usefull
  • e.g. two full PbPb runs from 2011 are available on SAF

Some pointers...

... if you are interested, for your own personal use or because you want to join the game...

• TDR : Upgrade of the Online-Offline Computing System (O2)

• TDR : Upgrade of the ALICE Readout & Trigger System

• HLT Tutorial

• O2 Status and Plans (Dec. 2015)

• ALFA Framework Status (Dec. 2015)

Questions ?

Backup

Playing with the MCH filtering stage