CHEP Rehearsals

22 Jun 2018, 09:55 → 26 Jun 2018, 19:00 Europe/Zurich

53/R-044 (CERN)

Friday 22 June

10:20 → 10:40 Exploring polyglot software frameworks in ALICE with FairMQ and fer

Speaker: Dr Sebastien Binet (IN2P3/LPC)

2018-07-09-chep-alice-fer.pdf

slides

Here are some short notes from the discussion:

• Please put slide numbers;
• It is not clear if you will have possibility to use Web presentation, please prepare a PDF version;
• Use the ALICE logo to have minimal "ALICE look";
• Too much and too long introduction to Go;
• Possible question: do you have users and real code?
  • The MUON detector is planning to use Go devices.

10:40 → 11:00 ALFA: ALICE-FAIR new message queuing based framework

Speaker: Mohammad Al-Turany (CERN)

ALFA_CHEP_2018.pdf

Here are the short notes from the discussion:

• Date and slide numbers are not visible, please change the style;
• The quality of the screenshots is bad;
• Slide 9 "cheating"picture gives wrong message;
• Please synchronise with Giulio the descriptions, abbreviations and topologies;
  • We got the impression it was different representation of the same thing.

11:00 → 11:20 Grid services in a box - container management in ALICE

Speaker: Maksim Melnik Storetvedt (Western Norway University of Applied Sciences (NO))

ServiceboxRehearsal.pdf

Here are the short notes from the discussion:

• Slide 4 - too long, it looks like a repetition;
• Use 16:9 layout;
• Performance graphs: split in two separate pictures and use bigger fonts;
• Clarify the conclusions:
  • The statement "similar performance, less overhead" is misleading;
  • Say that we are ready for the future when the sites are moving to containers;
  • Compare with the virtual machines overhead;
• Slide 6 - autofs is a problem, cvmfs is not;
• Docker Swarm is dying,use Kubernetes;
• When you speak, put accents on the key messages.

11:20 → 11:40 GPU-based Online Track Reconstructions for the ALICE TPC in Run 3 with Continuous Read Out

Speaker: David Rohr (CERN)

2018-07-10 CHEP 2018.pdf

Here are the short notes from the discussion:

• The 2 ms timeframe is what is shown since generating realistic 20 ms timeframe is not easy on a laptop;
• Put slide numbers;
• Define what is "mu" since not everybody from the computing community knows;
• Add which entropy encodings we consider;
• Clarify the sentence about 40 cores vs 800 cores and probably add text with arrows on the performance picture;
• Figures @ track fit (?);
• Why is GPU working so well for ALICE?

Monday 25 June

10:00 → 10:20 FairRoot and ALICE O2 multithreading simulation

Speaker: Ivana Hrivnacova (Centre National de la Recherche Scientifique (FR))

aliceo2_mt.pdf

Here are the short notes from the discussion:

• May be we do not need slide 2 with the outline since the talk is short;
• The talk was longer than 12 min.;
• The test slide took longer than needed;
• Use "O square" instead of "O two";
• On sl. 13 - TGeant3 and MT is misleading in the tutorial.

10:20 → 10:40 A scalable and asynchronous detector simulation system based on ALFA

Speaker: Sandro Christian Wenzel (CERN)

CHEP18TalkDraft_SandroWenzel.pdf

Here are the short notes from the discussion:

• Slide numbers on the "usual place";
• Slide 3 - no quotes;
• Slide 6 - message queue;
• Slide 7 - remove it is "very easy";
• Slide 12 - legend.

10:40 → 11:00 JAliEn: the new ALICE high-performance and high-scalability Grid framework

Speaker: Miguel Martinez Pedreira (Johann-Wolfgang-Goethe Univ. (DE))

394-JAliEn.pdf

394-JAliEn.pptx

Proceedings-394-v5.pdf

Here are the short notes from the discussion:

• Layout - create your own PDF;
• Slide 2 - the 36-38% growth: better use "significant growth", or "factor 10 from the beginning". You can probably remove the slide; 
• Slide 3 - O2 facility in Run 3;
• Slide 8 - the colours of the text should be adjusted;
• Slide 9 - MySQL cluster?

11:20 → 11:40 Readout software for the ALICE integrated Online-Offline (O2) system

Speaker: Filippo Costa (CERN)

201806_chep_readout_v3.pdf

Here are the short notes from the discussion:

• Slide 6 - add picture?
• Side 8 - split in 2 slides?
• Slide 10 - the picture is not visible, add picture of the machine;
• Align the slide numbers.

12:00 → 12:20 The ALICE Analysis Framework for LHC Run 3

ICHEP talk

Speaker: Markus Bernhard Zimmermann (CERN)

ICHEP_talk.pdf

Here are the short notes from the discussion:

• Use Run 2, Run 3 instead of Run II and Run III;
• Slide 2: data set -> typical data set which goes through a train;
• Slide 2: put the best guess about the number of data sets and the number of files;
• Slide 2: human readable formats of numbers;
• Slide 6 - fast decompression, see Root IO workshop and the adoption of LZ4.

12:20 → 12:40 Towards the ALICE Online-Offline (O2) control system

Speaker: Teo Mrnjavac (CERN)

deck-chep.pdf

Here are the short notes from the discussion:

• Slide 2: abbreviations, terminology. Use O2 facility. Specyfy the number of FLP and EPN;
• Reminder slide what is the ALICE control system doing;
• Contemporary -> latest or modern;
• Slide 8: you use Mesos to build custom scheduler;
• Mesos is a tool to solve the tasks of the control system, underline that one day it might be replaced;
• Do not stick to only one (Mesos) solution;
• GO slide is kind of "preaching", just mention what we use GO for;
  • You can point to the presentation of Sebastien Binet;
• Accent on the goals and not on the tools;
• DDS is not mentioned. We are searching for a scheduler, but DDS is not. The schedulers may have deployment functionality, but how the control system interacts with the devices? Using DDS or directly (slide 12)?
  • It is explicit decision since we are evaluating different solutions;
  • Underline it is R&D of the Run 3 Control system;
  • The confusion comes from slide 12 and 13, where we have the OCC plugin and O2 executor.;
  • Underline the flexibility;
• Slide 12: use simply "Process" and "Device" on the bottom;
• Slide 12: where would DDS fit in the picture?
• Slide 14 or Conclusions:
  • "See also DDS poster for alternative approach to deployment on large clusters..."

16:40 → 17:00 Online reconstruction of tracks and vertices using the Upgraded Inner Tracking System of ALICE at the LHC

Speaker: Mr Matteo Concas (INFN e Politecnico di Torino (IT))

CHEP2018concas.pdf

• In the conclusions, please underline you have one algorithm and two implementations, not two algorithms.
• Add units in the table.
• In the conclusions mention that the results are identical from GPU and CPU.
• Please add your poster ID somewhere on the poster. It will be used in the vote for best poster.

17:00 → 17:20 Using Generative Adversarial Networks for fast simulations in the ALICE Experiment

Speaker: Tomasz Piotr Trzcinski (Warsaw University of Technology (PL))

trzcinski_chep18.pdf

• Slide 3 - visualisation does not work with PDF. How do you show it? - Power point.
• Slide 6 - make it clear that the noise is used to initialise the GAN;
• Do you generate the charge? - it is a parallel chain in development;
• How do you take into account the magnetic field? - we suppose constant field with fixed value (0.5T). We plan to use it as input.
• The single track simulation does'n not account for cluster merging, shared clusters, etc.
• Please add a side with the work in progress and plans.

17:20 → 17:40 ALICE Overwatch: Online monitoring and data quality assurance using HLT data

Speaker: Raymond Ehlers (Yale University (US))

overwatch.v1.pdf

• Slide 10 - you spent a lot of time as you were apologising you did this works;
• Slide 7  and 8 - too much time for explaining the sliders;
• Slide 3 - put the year when the project started;
• Slide 4 - give some information about the HLT algorithms, the data transport and the types of objects you are sending (histograms). In addition this is a good example of sending ROOT objects using FairMQ and can be considered as O2 prototype.

17:40 → 17:45 DDS - The Dynamic Deployment System

Speaker: Andrey Lebedev (GSI - Helmholtzzentrum fur Schwerionenforschung GmbH (DE))

poster_chep2018_lebedev.pdf

• Put the poster ID to allow voting in the competition for best poster.
• Make the "user perspective" look as the top (may be use more visible box).
• Move the command line interface to the right and add prompt.
• In the motivation - use "in context" instead of "in frames".

17:45 → 17:50 The Security model of the ALICE next generation Grid framework

Speaker: Volodymyr Yurchenko (National Academy of Sciences of Ukraine (UA))

chep18-poster-yurchenko-v3.pdf

• Add the references at the end of the poster, if you have any.
• Put the names of the authors in bold.

17:50 → 17:55 A simulation tool for ALICE storage and computing resource usage

Speakers: Armenuhi Abramyan (Yerevan Physics Institute), Armenuhi Abramyan (A.Alikhanyan National Science Laboratory (AM))

CHEP_2018_poster.pdf

• In 2017 we did not have Pb-Pb data, so it is not good to use it for extrapolation for Run 3. Take 2015 as reference Run 2 period.
• Put the poster ID to allow voting in the competition for best poster.
• The red "...under discussion stage." -> "...under discussion."
• Move the ALICE logo on top.
• Put the names of the authors in bold.
• Crosscheck the tables on bottom left.

17:55 → 18:00 Shared Memory Transport for ALFA

Speaker: Alexey Rybalchenko (GSI - Helmholtzzentrum fur Schwerionenforschung GmbH (DE))

CHEP_Poster_draft_v2.pdf

• Put the poster ID to allow voting in the competition for best poster

18:00 → 18:05 RDMA-accelerated data transport in ALFA

Speaker: Dennis Klein (GSI - Helmholtzzentrum fur Schwerionenforschung GmbH (DE))

CHEP Poster Draft v3.pdf

CHEP Poster Draft v4.pdf

• Put the poster ID to allow voting in the competition for best poster

18:05 → 18:10 ALICE O2 setup for TPC read-out tests for the LHC run 3

Speaker: Johannes Lehrbach (Johann-Wolfgang-Goethe Univ. (DE))

chep18_a0_draft.pdf

chep18_a0_draft_v2.pdf

• Put the poster ID to allow voting in the competition for best poster.
• Complete the poster and make the new version available for comments by Wednesday.

18:10 → 18:15 Online track fit for the ALICE TPC detector in Online-Offline framework

Speaker: Sergey Gorbunov (Johann-Wolfgang-Goethe Univ. (DE))

GorbunovCHEP2018.pdf

• Put the poster ID to allow voting in the competition for best poster.
• "...round-of problem" -> "...rounding error".
• Use bigger font, especially for the pictures. Make it readable from 1 m distance.
• Put summary.

Tuesday 26 June

10:00 → 10:20 Evolution of the ALICE Software Framework for LHC Run 3

Speaker: Giulio Eulisse (CERN)

2018-7-chep-framework.pdf

Friday:

• On slide 2 or 3 put reference to the presentation of David;
• Speak slower;
• The introduction is too long;
  • Comparison with the multithreading (TBB) approach;
  • Scheduling;
  • Throughput optimisation;
  • Move the code example to the backup slides.

Tuesday:

• Slide 7: put DPL abbreviation there
• Slide 16: add link to the DDS poster
• The talk is ready, do not forget to submit it to CHEP2018 Indico page one week before the conference

10:20 → 10:40 A prototype for the ALICE Analysis Facility at GSI

Speaker: Kilian Schwarz (GSI - Helmholtzzentrum fur Schwerionenforschung GmbH (DE))

GSI ALICE AF PT CHEP 2018.pdf

• Avoid vertical text;
• Use 16:9 layout;
• Slide 3: Virt. Plat. -> Virtual Platform or VM. Remove the hostnames, use generic names. Avoid the username. Total capacity 15;
• Slide 7: Use link for GitHub commit. Simplify the text;
• Slides with orange boxes: make the font bigger;
• Slide 11, 12: make the font on the axes and titles bigger, make the legend more visible. Multiple squares per data point? Outlier on slide 12 - check it;
• Slide 13: overlapping points - make sure the values are not plotted twice;
• Shorten the presentation by 1.5 min.

All requirements have been fullfilled. New slides have been uploaded

 

10:40 → 11:00 The ALICE Analysis Framework for LHC Run 3

Speakers: Dario Berzano (CERN), Dario Berzano (Universita e INFN)

20180709 - CHEP2018 - ALICE Analysis in Run 3.pdf

• Slide 2: 5 MB/s is per job, while Slide 3: 200 GB/s is integrated, which means 2MB/s with 50 K cores;
• Slide 3: local fast storage;
• Slide 3: for the others the writing is limiting factor. We do not have this limitation;
• Slide 8: 10 GB time frame is not after the readout;
• Slide 9: "We plan to use SoA in Run 3" or keep just columnar, SoA is a detail;
• Slide 10: (ROOT, Apache Parquet) several prototypes investigated;
• Slide 14: The decompression is slowing down, but is not limiting factor. What is the analysis you run?

11:00 → 11:20 Software training for the next generation of Physicists: joint experience of LHCb and ALICE

Speaker: Dario Berzano (CERN)

11:20 → 11:40 Data handling in the ALICE O2 event processing

Speaker: Matthias Richter (University of Oslo (NO))

mrichter_CHEP2018-alice-o2-epn-processing_v0.2_180628.pdf

Here are the short notes from the discussion:

• Complete the presentation;
• What is the main idea and the message?
  • Data format?
  • Software framework?
  • Efficient data organisation?
  • "For efficient message passing the data has to be organised in a specific way..."
• E-mail address: use the CERN one?
• Scope of the presentations
  • from raw data headers to very high level things will be difficult to cover in 12 min;
  • not more than 7 messages;
  • do not introduce the RDH fields.
• Check the order of the talks in the track, so that only one person introduces ALICE.

11:40 → 12:00 ALICE Track visualisation options for LHC Run 3

Speaker: Julian Wojciech Myrcha (Warsaw University of Technology (PL))

CHEP18-Tracks.pdf

Here are the short notes from the discussion:

• Underline that OpenGL is not safe option;
• Say what is Vulcan;
• Slides 5-7 are "too dry", may be some picture of XY and YZ projections?
• Run 3 information is missing: challenges, hierarchical navigation, possibility to debug algorithms and detectors;
• Too much on the technology, not enough interesting pictures;
• Do not discuss the ALICE experiment;
• Drawing performance is only one aspect of the visualisation;
• Mention that we will need real time filters;
• The time dimension is more important in Run 3;
• Spilt the presentation in performance and visualisation parts.

12:00 → 12:20 Data distribution and load balancing for the ALICE Online-Offline (O2) system

Speaker: Gvozden Neskovic (Johann-Wolfgang-Goethe Univ. (DE))

2018-06-25_CHEP18_Rehearsals_Data_Dist.pdf

Here are the short notes from the discussion:

• What is the goal? - To describe the current research;
• Network technology is to be defined, while you present Infiniband as the only solution;
• RDMA is one option, please list also the others;
• Slide 9 - if the load balancing is related to the network, this is to be stated;
• This is R&D, state it clearly;
• New iteration on the presentation is needed, probably in the WG, then send it;
• Fix the layout, fonts and the pictures.

12:20 → 12:25 A new cluster finder for ALICE MUON arm for Run3

Speaker: Mr Sean Murray (University of Cape Town (ZA))

poster_portrait_1missinggraph.pdf

poster_portrait.pdf