Simon Delamare, Pierre Neyron and Lucas Nussbaum
This tutorial will present how to use the Grid’5000 platform to set up scientific experiments, with a focus on those related to High Performance Computing (HPC) and Cloud computing. Grid’5000 is a scientific instrument designed to support experiment-driven research in all areas of computer science related to parallel, large-scale or distributed computing and networking. Grid’5000 is supported by French major scientific institutions (Inria, CNRS, Renater, Universities).
Key features of Grid’5000 platform are ability to conduct large-scale and reproducible experiments, while providing to users a high capacity of customization at every level of computer stack (from hardware to high level applications).
Grid’5000 gives an easy access to a variety of hardware technologies and is particularly suitable to carry out HPC (high performance computing) experiments: Users can investigate parallel algorithms, scalability problems or performance portability on Grid’5000, without restrictions usually found on traditional HPC computing centers.
In the first part, the tutorial will introduce Grid’5000 and explain its usage basics. The second part will present resources and tools available on Grid’5000 to conduct HPC-related and Cloud-related experimentations.
The official Grid’5000 website is : https://www.grid5000.fr
The tutorial website (with support material) is: https://www.grid5000.fr/mediawiki/index.php/User:Sdelamare/Europar_Tutorial
First Part: Getting Started
This session will be dedicated to the general Grid’5000 usage. After a quick presentation of the testbed, users will learn how to obtain a Grid’5000 account and connect to the platform by configuring their SSH environment. They will learn how to connect to different Grid’5000 sites and to discover, visualize and reserve Grid’5000 resources. Then the tutorial will present how to get root access on nodes by deploying and customizing their own experimental environment.
Second Part: Cloud or HPC experimentation
The second session will present the various resources and tools available on Grid’5000 platform to perform experimentations related to either Cloud or HPC.
Option 1: HPC experiments
Attendees will learn how to discover and get access to Grid’5000 hardware with specific characteristics, such as accelerators (GPU, Xeon Phi) or HPC networks (Infiniband). The tutorial will explain how to use those resources in the context of Grid’5000. Finally, the tutorial will show how to design an environment dedicated to HPC experimentation thanks Grid’5000 advanced reconfiguration features.
Option 2: Cloud experiments
Grid’5000 provides the ability to install a Cloud stack inside an experiment, customize it, and experiment on it. In this part of the tutorial, attendees will deploy their own OpenStack (Liberty or Mitaka) cloud inside Grid’5000, and learn how to customize it using Grid’5000-provided tools.
To attend this tutorial, participants should have basic knowledge of Linux command line and SSH. A laptop with SSH client is also required. A Grid’5000 account will be created at the start of the tutorial.
Participants who want to attend to second part of the tutorial only require a Grid’5000 account and basic knowledge of platform usage.
Participants to this tutorial will be able to continue to use Grid’5000 after the tutorial in the context of the Grid’5000 Open Access program.