Managing large scale computing for MPD experiment Krzysztof
- Slides: 20
Managing large scale computing for MPD experiment. Krzysztof Dygnarowicz Supervisor: Oleg Rogachevsky 1
What we will talk about What is MPDroot Tasks Why we Simulate Resources Results Future plans 2 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
MPD Simulations 3 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
MPDroot 4 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Why we doing simulations now? To check if our model makes sense To make reconstruction programs To know what to implement in FPGA 5 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Large data problem 6 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
NICA Resources Cluster Heterogeneous platform “Hybri. LIT” SUPERCOMPUTER “Govorun” Hybri. LIT Other clusters working in Russian Data Intensive Grid (RDIG) 7 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
HYBRILIT 8 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Software Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com 9
Monitoring of resources 10 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Monitoring jobs 11 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Other clusters in RDIG 12 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
By Andrey Moshkin Data of simulations 13 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
My tasks HOW TO GET ACCESS? HOW TO WORK WITH DIRAC AND LIT CLUSTER? HOW TO USE DIRAC ON NICA CLUSTER? 14 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
What has been done? Scripts to product events and reconstruction working on LIT cluster. DIRAC job manager scripts – in progress. Instructions how to get access and run your jobs on LIT and by DIRAC job manager. 15 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Outline of scripts 1. Get information of JOBID 2. Link generator and Mpd. Root 3. Generate data (runqmd. bash for URQMD) 4. Run them by detector (run. MC. c) 5. Reconstruct event from detector data (reco. c) 6. Copy data to safe storage 7. Erase unnecessary data 16 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Instruction How to get access How to start Jobs Where to store data 17 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Future plans Improve scripts and file storage system Extend instruction of NICA cluster and future improvements of DIRAC system Generate events 18 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Sources https: //home-hlit. jinr. ru/#/ http: //hlit. jinr. ru/en/installed_software_e ng/ http: //hlit. jinr. ru/en/for_users_eng/hardw are-and-software-environment_eng/ http: //db-nica. jinr. ru/mpdmc/stat. php https: //home. cern/science/computing/stor age http: //mpd. jinr. ru/wpcontent/uploads/2018/06/NICA_computing _TDR_1. 03. pdf http: //nica. jinr. ru/projects/mpd. php 19 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com
Thanks for your attention kdygnaro@protonmail. com 20 Managing large scale computing for MPD experiment, Krzysztof Dygnarowicz kdygnaro@protonmail. com