Solving parallel I/O challenges with dynamic workload management
When:
12 December 2023, From 09:30 to 18:00
Meeting Place:
- Online: A connection link will be sent to registered people (see registration below).
- In person: BSC-Repsol Floor -1 (S1-2-1) Barcelona Supercomputing Center – Centre Nacional de Supercomputació
Registration
Use the following form: https://forms.gle/t9MU9rnPYZZrEMxr6
Content / Tentative Program
- Morning session (9:30 – 13:00) duration 3:30
- Introduction
- General overview of ADMIRE (Jesus Carretero, UC3M)
- Exascale I/O
- Ad-hoc Storage GekkoFS (Marc Vef, Mainz Uni.)
- Ad-hoc Storage Hercules (Javier García-Blas, UC3M)
- Capio (Massimo Torquati, UNIPI)
- Lustre Quality of Service (QoS) (Jean-Thomas Acquaviva, DDN)
- Job Malleability
- FlexMPI (David E. Singh, UC3M)
- Introduction
- Afternoon session (14:00 – 17:30) duration 3:30
- Real-Time Monitoring at Scale
- ADMIRE Metric Proxy: Enabling real-time measurement at Supercomputer Scale (Jean-Baptiste Besnard, ParaTools SAS)
- Job and I/O Scheduling
- In-transit and In-Situ (Ramon Nou, BSC; Yi Ju, MPCDF)
- ElastiSim (Taylan Özden, TU Darmstadt)
- Application and I/O Modelling
- FTIO: Predicting I/O Phases Using Frequency Techniques (Ahmad Tarraf, TU Darmstadt)
- Extra-P: Extending Extra-P with I/O Modelling Capabilities (Ahmad Tarraf, TU Darmstadt)
- GreatNestor (Barbara Cantalupo, UNITO)
- Application Co-Design
- Overview and challenges (Rocco Sedona, Juelich)
- Applications some examples
- Real-Time Monitoring at Scale
- Meeting closing (17:30 – 18:00) (Jesus Carretero, UC3M)