ROOT I/O Meeting

Friday 28 Sept 2018, 16:00 → 17:40 Europe/Zurich

32/1-A24 (CERN)

Brian Paul Bockelman (University of Nebraska-Lincoln (US)), Philippe Canal (Fermi National Accelerator Lab. (US))

Attendees: Philippe, Danilo, Jim, Zhe, Guilherme Oksana

Philippe: Let’s delay discussing the ROOT Users’ workshop feedback until next week.

Oksana: This week I was able to run zlib with an updated CRC and have some numbers with look improved (See DIANA meeting on October 15th).  I am also started to look at the Intel zlib library.  Next step: JIRA issue related to fix performance LZ4 (JIRA ROOT-9633) ; it will take some times.

Zhe: Working on bulk I/O PR for Brian.

Philippe: Working TTree I/O customization scheduling.  I have fixed the CMS uses cases.  Need to check another independent report (rule not executed).   Also solve race condition in new StreamerInfo reading optimization; Danilo could you check the performance in case of many thread opening the same file at the same time.  I will also debug a new ‘configuration problem’ seen by ART.

Guilherme: RAS for I/O. Working on cmake.

Danilo: I fixed the missing dictionary checker for std::unique_ptr.  I still need to add support for std::array and std::tuple.

Danilo: TChain fails late if one of the file is corrupted.

Philippe: Maybe the TTreeReader should be such that it ’warns and skips’ bad files/entries.

Axel: Sorry, I have to move on. My report: MacOS 10.14 support, and IMT auto-on shows several IMT-related failures

Philippe: Will need to discuss goal/interface of TMPIFile …..

Danilo: We need to make sure we carefully test this before incorporate (and take ‘ownership’ of the code).

Philippe: Yes Taylor and Yunsong will be take TMPIFile further.

Danilo: I attended the ATLAS I/O review.  No surprise/tragedy.   I am not sure how public the information is.  There are action on us:
  * Description/primer of the parameter that can alter the I/O performance (eg. CacheSizes, AutoSave, etc.)
  * Asynchronous prefetching is not stable enough to even test.
  * Point made about being able to run on HPC.
  * Peter V.G. was a good bridge between ATLAS and ROOT worlds.

Philippe: We will have a meeting next week.

16:00 → 16:20 Round Table