UL HPC School Program
The sessions will take place at the Kirchberg Campus, in B13 and A02.
Important You are expected to bring your personal laptop for all sessions since there will be no workstation available on site.
All tutorials proposed as practical sessions are available on GitHub.
Agenda - March 13th, 2015
| March 13th | Presenter | Session title | Location |
|---|---|---|---|
| 10h00-11h00 | S. Varrette | Keynote: HPC platform @ UL: Overview and Usage | B13 |
| 11h00-12h00 | S. Varrette | PS1: Getting started: basic usage of the platform (ssh, scp, rsync, OAR, modules) | B13 |
| 12h00-13h30 | LUNCH | ||
| 13h30-14h15 | H. Cartiaux | PS2: HPC workflows with sequential jobs (C, Python, Java) | A02 |
| 14h15-15h00 | V. Plugaru | PS3: HPC workflows with parallel jobs (MPI, OpenMP) | A02 |
| 15h00-15h45 | M. Schmitt | PS4: Software environment deployment (RESIF and Vagrant) | A02 |
PS = Practical Session using your laptop
Detailed Program for the practical sessions
Practical Session 1
Getting Started (ssh, scp, rsync, oar, modules, builds, screen), by: S. Varrette
This tutorial will guide you through your first steps on the UL HPC platform. We will cover the following topics:
- Platform access via SSH
- Overview of the working environment
- File transfer
- Reserving computing resources with OAR and job management
- Usage of the web monitoring interfaces (Monika, Drawgantt, Ganglia)
- Using modules
- Advanced job management and Persistent Terminal Sessions using GNU Screen.
- illustration on Linux kernel compilation
Practical Session 2
HPC workflow with sequential jobs (C, Python, Java), by H. Cartiaux
For many users, the typical usage of the HPC facility is to execute 1 (sequential) program with many parameters. On your local machine, you can just start your program 100 times sequentially. However, you will obtain better results if you are able to parallelize these executions, either across several threads or several nodes of the HPC platform.
During this session, we will cover 3 use cases:
- Using the serial launcher (single node, in sequential and parallel mode) with a C application.
- Using the generic launcher to distribute the (independent) executions on several nodes of a python application.
- The advanced usage of a complex Java framework JCell, designed to work with cellular genetic algorithms (cGAs)
Practical Session 3
HPC workflow with parallel jobs. Ex: OSU MB/HP, by V. Plugaru
Online instructions for OSU Microbenchmarks and for HPL
The objective of this session is to compile and run MPI programs over various reference MPI implementations (Intel MPI, OpenMPI or MVAPICH2, all available on the cluster). In practice, this session will focus on the following MPI applications:
- the OSU Micro-benchmarks which permit to measure the performance of an MPI implementation
- the High Performance Linpack (HPL) which is used to rank supercomputers across the world for the Top500 list
A discussion of MPI, OpenMP and MPI/OpenMP hybrid applications will conclude this section.
Practical Session 4
Software environment deployment (RESIF, Vagrant), by M. Schmitt
The objective of this session is to demonstrate the usage of the UL HPC RESIF tool to deploy a software environment:
- locally on your base workstation (Linux / OS X)
- in a Linux Virtual Machine (created by Vagrant with the VirtualBox provider)
- under your account on the UL HPC platform