Performance of ODROID-MC1 for scientific flow problems

摘要

In late 2017, Hardkernel released the ODROID-MC1 cluster system, which is based on the ODROID-XU4 single-board computer. The cluster consists of four nodes, each equipped with a Samsung Exynos 5 Octa (5422) CPU. The system promises high computational power under low energy consumption. In this paper, the applicability of such a systems to scientific problems is investigated. Therefore, flow computations using a lattice-Boltzmann method are employed to evaluate the single core, single node, and multinode performance and scalability of the cluster. The lattice-Boltzmann code is part of a larger simulation framework and scales well across several high-performance computers. Performance measurement results are juxtaposed to those obtained on high-performance computers and show that the ODROID-MC1 can indeed compete with high-class server CPUs. Energy measurements corroborate the ODROID's energy efficiency. Its drawbacks result from the limited amount of available memory, the corresponding memory bandwidth, and the low-performing Cortex A7 cores of the big.LITTLE architecture. The applicability to scientific applications is shown by a three-dimensional simulation of the flow in a slot burner configuration. (C) 2019 Elsevier B.V. All rights reserved.