Containerisation of NEMO Employing Singularity (CONES)
ARCHER2-eCSE02-13
PI: Dr James Harle (National Oceanography Centre)
Technical staff: Dr Chris Wood (EPCC)
Subject Area:
Figure 1: The NEMO AMM7 model domain used in this study. Variable shown is Sea Surface Temperature (°C).
Image produced by the COAsT analysis package (Polton et al 2023).
Containerisation is the process of packaging up a computer program, and everything else required to run it. In doing so the program becomes portable and easily transferable between desktop computers, High Performance Computing systems (such as ARCHER2) and even the cloud. In this study we detail a method to containerise a program of a numerical ocean model (NEMO) and examine how well it performs compared to the conventional method of compiling and running it. For the purposes of testing, we made use of the 7km resolution Atlantic Margin Model (AMM7; Figure 1) developed at the National Oceanography Centre.
We conducted a series of simulations: 3 conventional and 6 using the containerisation approach. Each of these used different compiler options and/or containerisation methods. The simulations were performed over a range of code optimisation levels and core counts (i.e. proportion of computer resource used to run each simulation) and their performances compared.
Overall, the containerisation process adds up to 10% to how long the ocean model takes to run, compared to the conventional approach. For many scientific studies this may be acceptable in the face of the benefits afforded:
- There is minimal setup knowledge required.
- It is easily transferred between different computer systems.
- The packaged program can be published.
- It makes the scientific study reproducible for others.
This work will hopefully widen the access to the NEMO ocean model within the ARCHER2 user community. It will also provide a template to for others wishing to containerise their code on the ARCHER2 system. While this approach provides an essential means to reproducible science, it can also be used to facilitate the teaching/demonstrating of ocean modelling.
Publications linked to this study:
Polton, J., Harle, J., et al: Reproducible and relocatable regional ocean modelling: fundamentals and practices, Geosci. Model Dev., 16, 1481–1510, https://doi.org/10.5194/gmd-16-1481-2023, 2023.
Information about the code
NEMO (Nucleus for European Modelling of the Ocean), a state-of-the-art modelling framework for ocean and climate modelling
Singularity (now Apptainer), a containerisation software
CoNES, GitHub builds of NEMO