Design and optimisation of an efficient HDF5 I/O kernel for massive parallel fluid flow simulations

TL;DR

This paper presents a new HDF5 I/O kernel optimized for massive parallel fluid flow simulations, enabling high-speed data storage and retrieval on supercomputers for large-scale scientific computations.

Contribution

It introduces an efficient HDF5 I/O kernel that minimizes latency and maximizes bandwidth, supporting fast visualization and data steering in large-scale parallel simulations.

Findings

Achieved write bandwidths near theoretical peak on supercomputing clusters.

Enabled fast interactive visualization and steering functionalities.

Supported high-fidelity computations with trillions of unknowns.

Abstract

More and more massive parallel codes running on several hundreds of thousands of cores enter the computational science and engineering domain, allowing high-fidelity computations on up to trillions of unknowns for very detailed analyses of the underlying problems. During such runs, typically gigabytes of data are being produced, hindering both efficient storage and (interactive) data exploration. Here, advanced approaches based on inherently distributed data formats such as HDF5 become necessary in order to avoid long latencies when storing the data and to support fast (random) access when retrieving the data for visual processing. Avoiding file locking and using collective buffering, write bandwidths to a single file close to the theoretical peak on a modern supercomputing cluster were achieved. The structure of the output file supports a very fast interactive visualisation and introduces additional steering functionality.