BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:NumPEx tutorial: I/O data management with PDI [ONLINE]
DTSTART:20250930T070000Z
DTEND:20250930T100000Z
DTSTAMP:20260421T031700Z
UID:indico-event-14709@indico.math.cnrs.fr
DESCRIPTION:Tutorial on I/O data management with PDI\n2025\, September 30\
 , 9:00-12:00\nMaison de la Simulation\, together with the Exa-DoST and Exa
 -DI projects of NumPEx organize a free Tutorial on I/O data management wit
 h PDI. This tutorial will be held online via Zoom.\nContext\nThe increase 
 in computational power goes hand in hand with an increase in the amount of
  data to manage. At Exascale\, IO can easily become the main performance b
 ottleneck. Several I/O libraries (FTI\, HDF5\, netCDF\, etc) exist\, provi
 ding efficient management of I/O data with different focuses respectively.
  However\, the traditional way of tightly coupling the simulation code wit
 h external I/O libraries is less flexible and can have an important impact
  on the maintenance of the code base. \nTo address this issue of flexibil
 ity\, we designed PDI\, which is an abstraction layer with a plugin system
  that allows the use of external I/O libraries in a dynamic and flexible w
 ay. Users can choose different plugins\, via a YAML configuration file\, f
 or the desired dataflow without recompiling the simulation code.\nContent\
 nThis tutorial will present the main concepts of PDI with introductory lec
 tures and incremental hands-on exercises. Several PDI plugins will be cove
 red during the tutorial\, including trace\, hdf5\, pycall\, and user_code.
  \nAll hands-on exercises will be based on a simple 2D simulation written
  in C. Participants will learn how to integrate PDI into this example simu
 lation code. They will also learn how to configure the YAML file in order 
 to use different PDI plugins to manage the data generated by the simulatio
 n.\nThe main topics covered by hands-on exercises are:\n\nInstrument your 
 code with PDI.\nGet the trace information with PDI.\nUse pycall plugin to 
 generate partial images.\nUse hdf5 plugin to generate output files (one fi
 le per process\, one file per timestep\, one file for all timesteps). \n(
 If time allows) Use user_code plugin to compute in-process metrics on-the-
 fly.\n(If time allows) Demonstration of in-situ visualization using cataly
 st plugin.\n\n \nOrganization\nRegistration is free\, but mandatory. \n\
 nhttps://indico.math.cnrs.fr/event/14709/
LOCATION:Online
URL:https://indico.math.cnrs.fr/event/14709/
END:VEVENT
END:VCALENDAR