Distributed multiscale computing with MUSCLE 2, the Multiscale Coupling Library and Environment

Abstract

We present the Multiscale Coupling Library and Environment: MUSCLE 2. This multiscale component-based execution environment has a simple to use Java, C++, C, Python and Fortran API, compatible with MPI, OpenMP and threading codes. We demonstrate its local and distributed computing capabilities and compare its performance to MUSCLE 1, file copy, MPI, MPWide, and GridFTP. The local throughput of MPI is about two times higher, so very tightly coupled code should use MPI as a single submodel of MUSCLE 2; the distributed performance of GridFTP is lower, especially for small messages. We test the performance of a canal system model with MUSCLE 2, where it introduces an overhead as small as 5% compared to MP

Multiscale computing in the exascale era

“…Over the last decade, and with many collaborators, we have developed the so-called Multiscale Modelling and Simulation Framework (MMSF) for designing, programming, implementing and executing multiscale applications [3,26,27,[35][36][37][38][39][40][41][42]. This framework has been successfully tested on applications from several fields of science and technology (e.g.…”

“…This framework has been successfully tested on applications from several fields of science and technology (e.g. fusion [35,43], computational biology [35,44,45], biomedicine [17,35,36,[46][47][48][49][50][51][52], nanomaterial science [9,13,35], and hydrology [35]). The MMSF offers many benefits: a clear methodology, software and algorithm reuse, the possibility to couple new and legacy codes, heterogeneous distributed computing, and access to unprecedented computing resources [26].…”

“…It describes the scales and computational requirements of submodels and any scale bridging components needed. A coupling library like MUSCLE2 [36] ensures that communication between heterogeneous components is possible, with minimal and local changes to the single scale code. In the last step of the pipeline, submodels are executed on suitable computing infrastructure.…”

“…This relates to a multiscale model for in-stent restenosis, an unwanted response of coronary arteries after balloon angioplasty and stenting. [36,37,51,57,58] The 'SMC' refers to a slow dynamical process, the response of Smooth Muscle Cells in the coronary arterial wall, with a typical timescale of days. Drug Diffusion (DD) simulates diffusion of inhibitory drugs from a stent into the arterial wall and has a timescale of minutes, and the blood flow (BF) has a timescale of one heartbeat, so a second.…”

Coupled Simulations in Plasma Physics with the Integrated Plasma Simulator Platform

“…MUSCLE 3 is a coupling library dedicated to multiscale applications [13] following a componentbased approach. It provides its simple API in Java, C/C++, Fortran, Python, MATLAB and Scala, which supports developers while implementing their own components.…”