DEEP-SEA -- WP3 Telco 2022-07-21 -- #30 ==== ###### tags: `DEEP-SEA` `WP3` `Minutes` ## Table of contents [toc] ## Important links [DEEP-SEA WP3 Overview (index page)](https://gitlab.version.fz-juelich.de/hedgedoc/3lNokhk6QTup4qda8AD7lw) [DEEP-SEA WP3 on Gitlab](https://gitlab.version.fz-juelich.de/deep-sea/wp3) [Notes from the last meeting](/hedgedoc/BH1ZSe-YS7KkJRDdlpBpog#) [Central landing page (top-level index of WPs, misc. resources, etc)](/hedgedoc/5UPCNFrTRJOCWNsAnQZPGg#) ## List of participants | Partner | Name(s) | |:-------:| ----------------------- | | FZJ | Abhiraj Bishnoi, Norbert Eicker, Hans-Christian Hoppe, Dominik Gottwald, Ghazal Tashakor, Jan-Oliver Mirus | | ParTec | Michael Rauh, Xu Huang, Carsten Clauss, Jennifer Lopez | | Atos | Piotr Lesnicki | | CEA | | | BSC | Antonio J. Peña| | FHG | Tiberiu Rotaru| | FORTH | Manolis Marazakis, Lena Kanellou | | LRZ | | | ETHZ | | | TUDA | | | KTH | | | KUL | | | ECMWF | | | TUM | Isaias Compres | Announced absence: Simon Pickartz, Thomas Moschny ## Agenda * Update of tasks by each task leader (i.e., first plan for the start of the tasks): - Task 3.2: Memory Management (Antonio Peña) - Task 3.3: Low-level Communication (Carsten Clauss) - Task 3.4: Resource Management, Scheduling and Orchestration (Muhammad Owais) - Task 3.5: Containerisation (Michael Rauh) - Task 3.6: Continuous Integration (Abhiraj Bishnoi) ## Minutes ### Task status update - Tk3.2 (Memory Management | M9-M36 | lead: BSC) - *BSC*, FORTH, TUM, CEA - Interaction BSC+FORTH coming up in mid-September - People are invited to join MNHACK as a way to spend a week programming with applications as users of the stuff developed in this WP: - https://www.bsc.es/education/training/other-training/mnhack22-4th-marenostrum-hackathon - Tk3.3 (Low-level Communication | M9-M36) - *ParTec*: - work on documentation of pscom RMA interface extensions ongoing - ...and some interface design decisions are also still under discussion - For example, handles/objects for referencing memory regions are currently considered - ...perhaps as an addition to the flat memory model pscom currently supports - Atos: - FORTH: reducing memory consumption of RMA profiling tool. Specifically, we are focusing on this, since we would like to use the profiler on large traces (e.g. matching the scale of Graph 500). Currently modifying the way some statistics are produced by the profiler tool, so that they can be calculated "on the fly", without the need to store superfluous input data. WIP. - CEA: - Tk3.4 (Resource Management, Scheduling and Orchestration | M9-M36): - *FHG* , FZJ, ParTec, FORTH, TUM - TUM: MPICH work on malleability updates via PMIx started. - FORTH: tracking job launch sequence in slurm codebase - FHG: Tiberiu will replace Owais in this task. Started looking at the Slurm simulator and reading related papers/documentation. - Tk3.5 (Containerisation | M9-M36) - *ParTec* - nothing new to report - FORTH - no new update - Tk3.6 (Continuous Integration | M9-M36) - *FZJ*, ParTec, Atos, CEA, BSC, FHG, FORTH, TUM, ETHZ, KTH - Dependency Graph and inverse dependencies - Can create graph for the whole stack now in 1-2 minutes, no need to pre-compute and store on disk - Will compute it on the fly for each commit - Script to extract inverse dependencies is complete - Testing CI with Score-P - Ran into some bugs wherein the commit into the EB repo is not made with the same permissions as the user who started the pipeline - Potential solution: Use masked personal access tokens (limited to the repository, cannot be used outside it) - Works as intended, will share updated documentation with all stakeholders - Podman container for IOSEA - Thomas is back :) Will work with ParTec to setup similar container for IOSEA users ### Next Actions / AoB * Next WP3 meeting: **August 4^th^, 2022 at 16:00 CEST**