Understanding the Scalability of Molecular Simulation Using Empirical Performance Modeling

Abstract

Molecular dynamics (MD) simulation allows for the study of static and dynamic properties of molecular ensembles at various molecular scales, from monatomics to macromolecules such as proteins and nucleic acids. It has applications in biology, materials science, biochemistry, and biophysics. Recent developments in simulation techniques spurred the emergence of the computational molecular engineering (CME) field, which focuses specifically on the needs of industrial users in engineering. Within CME, the simulation code ms2 allows users to calculate thermodynamic properties of bulk fluids. It is a parallel code that aims to scale the temporal range of the simulation while keeping the execution time minimal. In this paper, we use empirical performance modeling to study the impact of simulation parameters on the execution time. Our approach is a systematic workflow that can be used as a blueprint in other fields that aim to scale their simulation codes. We show that the generated models can help users better understand how to scale the simulation with minimal increase in execution time.

Offsite Autotuning Approach

“…As an alternative to runtime tests, analytic performance models can be applied to either select the most efficient variant or to reduce the number of tests required by filtering out inefficient variants beforehand. In general, two categories of performance models are distinguished: (i) black box models applying statistical methods and machine learning techniques to observed performance data like hardware metrics or measured runtimes in order to learn to predict performance behavior [15,20], and (ii) white box models such as the Roofline model [8,24] or the ECM performance model [12,21] that describe the interaction of hardware and code using simplified machine models. For loop kernels, the Roofline and the ECM model can be constructed with the Kerncraft tool [11].…”

Offsite Autotuning Approach -- Performance Model Driven Autotuning Applied to Parallel Explicit ODE Methods

“…As an alternative to runtime tests, analytic performance models can be applied to either select the most efficient variant or to reduce the number of tests required by filtering out inefficient variants beforehand. In general, two categories of performance models are distinguished: (i) black box models applying statistical methods and machine learning techniques to observed performance data like hardware metrics or measured runtimes in order to learn to predict performance behavior [16,18], and (ii) white box models such as the Roofline model [21] or the ECM performance model [9,17] that describe the interaction of hardware and code using simplified machine models. For loop kernels, the Roofline and the ECM model can be automatically constructed with the Kerncraft tool [8].…”

Semantic Interoperability and Characterization of Data Provenance in Computational Molecular Engineering

“…To demonstrate the viability of the present approach, this is applied to phosgene (using the model by Huang et al 25 ), building on previous work by Rutkai and Vrabec; 83 there, the same problem was addressed without employing a dedicated WMS, and without characterizing the provenance of the EOS parameterization as well as the data obtained by molecular simulation. The present implementation addressing this class of problems uses sampling of state points and fitting with the method developed by Shudler et al 13 The corresponding workflow can be implemented using the ms2 simulation program. The data flow and steps to be performed are depicted in Fig.…”

“…Typical challenges hence include the management of a great amount of individual tasks, the organization of the results as well as the setup and execution of simulations on diverse and heterogeneous computer system environments and architectures. 52 The TaLPas WMS addresses these problems. Its overall architecture is shown in Figure 2.…”

“…Furthermore, certain aspects of the data uncertainty (and uncertainty propagation), such as the sensitivity with respect to specific model parameters or the choice of a particular solver implementation, can be quantified by varying individual values, parameters, or elements of a workflow; 50 e.g., round-robin studies can be conducted, where various simulation software environments are employed to carry out the same (or closely related) algorithms in combination with the same models, comparing the outcome. 51 Other technologies that can profit from well-defined workflow semantics include high performance computing (HPC) and scheduling environments where computational requirements may be automatically predicted 52 and optimized by workflow autotuning and task-based parallelization. 53 In computational molecular engineering, two major organized efforts toward achieving an agreed coherent semantictechnology framework have been conducted: With a focus on process simulations, CAPE-OPEN was developed in computer aided process engineering as an open interface standard.…”