Erschienen in:
Geoscientific Model Development, 14 (2021) 12, Seite 7391-7409
Sprache:
Englisch
DOI:
10.5194/gmd-14-7391-2021
ISSN:
1991-9603
Entstehung:
Anmerkungen:
Beschreibung:
Abstract. Coupled reactive transport simulations are extremely demanding in terms of required computational power, which hampers theirapplication and leads to coarsened and oversimplified domains. Thechemical sub-process represents the major bottleneck: itsacceleration is an urgent challenge which gathers increasinginterdisciplinary interest along with pressing requirements forsubsurface utilization such as spent nuclear fuel storage,geothermal energy and CO2 storage. In this context we developedPOET (POtsdam rEactive Transport), a research parallel reactive transport simulatorintegrating algorithmic improvements which decisively speed upcoupled simulations. In particular, POET is designed with amaster/worker architecture, which ensures computational efficiencyin both multicore and cluster compute environments. POET does not rely on contiguous grid partitions for the parallelization ofchemistry but forms work packages composed of grid cells distantfrom each other. Such scattering prevents particularly expensivegeochemical simulations, usually concentrated in the vicinity of areactive front, from generating load imbalance between the availableCPUs (central processing units), as is often the case with classical partitions.Furthermore, POET leverages an original implementation of thedistributed hash table (DHT) mechanism to cache the results ofgeochemical simulations for further reuse in subsequent time stepsduring the coupled simulation. The caching is hence particularlyadvantageous for initially chemically homogeneous simulations andfor smooth reaction fronts. We tune the rounding employed in the DHTon a 2D benchmark to validate the caching approach, and we evaluatethe performance gain of POET's master/worker architecture and the DHT speedup on a 3D benchmark comprising around 650 000 grid elements.The runtime for 200 coupling iterations, corresponding to 960simulation days, reduced from about 24 h on 11 workers to 29 minon 719 workers. Activating the DHT reduces the runtime further to2 h and 8 min respectively. Only with these kinds of reducedhardware requirements and computational costs is it possible torealistically perform the long-term complex reactive transportsimulations, as well as perform the uncertainty analyses requiredby pressing societal challenges connected with subsurfaceutilization.