Fishing for parallelisation strategies
Fishing is a major economic source for many Pacific Islands. SEAPODYM is a modelling framework designed to predict tuna population dynamics under fishing pressure and various environmental conditions (e.g., climate change). The code integrates spatiotemporal, age-structured advectionreaction-diffusion models with statistically fitted observation data. Simulations can take on the order of 10 hours, depending on the grid resolution and the number of fish cohorts.
To reduce the wall clock time, NeSI was approached by the Pacific Community to review the code and make suggestions for applying parallelisation and other acceleration strategies. Here, we present our findings, including the development of an MPI toy program that emulates the behaviour of SEAPODYM to obtain an estimate for the maximum achievable parallel scalability of the code on NeSI’s parallel platform.
ABOUT THE AUTHOR
Alex Pletzer is a research software engineer who helps researchers run better and faster on NeSI platforms. On his spare time, Alex windsurfs, paints and plays ping pong.
For more information about eResearch NZ / eRangahau Aotearoa, visit:
https://eresearchnz.co.nz/