Harshad - Handle Darshan Data
0 likes82 views
The document outlines the use of the Darshan tool for performance data acquisition on the Shaheen 2 and Cori supercomputers at the KAUST Supercomputing Laboratory. It provides a framework and instructions for accessing Darshan data, including how to connect to Shaheen, retrieve performance data from specific job IDs, and compare executions. Additionally, it emphasizes user support and encourages feedback for potential enhancements to the tools provided.
Harshad - Handle Darshan Data
- 1. Handle Darshan Data (HarshaD) George Markomanolis Computational Scientist KAUST Supercomputing Laboratory georgios.markomanolis@kaust.edu.sa 7 November 2017 1
- 2. Using Darshan tool —? Have you ever used Darshan tool? —? If the answer is “I don’t know, probably not”, and you are using Shaheen 2 or Cori, then maybe you have used it, as it is enabled automatic on Shaheen II and Cori. —? KAUST Supercomputing Laboratory (KSL) provides a framework to provide you easy access to performance data from Darshan: —? Visit web page https://kaust-ksl.github.io/HArshaD/ for instructions. The framework is supported on both Shaheen and Cori. 2
- 3. HArsaD I —? Connect to Shaheen with –X (for Windows/Linux) or –Y (for Mac) —? git clone https://github.com/KAUST-KSL/HArshaD.git —? Get the Darshan performance data from your last experiment, execute: —? ./open_darshan.sh —? Get the Darshan performance data from the job id 65447, execute: —? ./open_darshan.sh 65447 —? Compare Darshan perfromance data from job id 65447 and 65448, execute: —? ./compare_darshan 65447 65448 —? You need extra tool, check the github link for instructions 3
- 4. HArshaD II - Comparison —? In case that you want to compare the execution of two applications, execute: —? compare_darshan.sh job1_id job2_id —? One PDF file, with the Darshan performance data of both executions, is created 4
- 5. Quantum Espresso – Darshan results I 5
- 6. Quantum Espresso – Darshan results II 6
- 7. Quantum Espresso – Darshan results III 7
- 8. Quantum Espresso – Darshan results IV 8
- 9. Quantum Espresso – Darshan results V 9
- 10. Quantum Espresso – Darshan results VI 10
- 11. S3D – Darshan I 11
- 12. S3D – Darshan II 12
- 13. Conclusions —? A suite of tools to use Darshan without knowing a lot of details was presented —? Explanations of how to read Darshan output —? Use the tools! https://kaust-ksl.github.io/HArshaD Provide feedback, we can add possible functionalities 13