Research Article: DENIS: Solving cardiac electrophysiological simulations with volunteer computing

Date Published: October 16, 2018

Publisher: Public Library of Science

Author(s): Violeta Monasterio, Joel Castro-Mur, Jesús Carro, Alexander V Panfilov.

http://doi.org/10.1371/journal.pone.0205568

Abstract

Cardiac electrophysiological simulations are computationally intensive tasks. The growing complexity of cardiac models, together with the increasing use of large ensembles of models (known as populations of models), make extensive simulation studies unfeasible for regular stand-alone computers. To address this problem, we developed DENIS, a cardiac electrophysiology simulator based on the volunteer computing paradigm. We evaluated the performance of DENIS by testing the effect of simulation length, task deadline, and batch size, on the time to complete a batch of simulations. In the experiments, the time to complete a batch of simulations did not increase with simulation length, and had little dependence on batch size. In a test case involving the generation of a population of models, DENIS was able to reduce the simulation time from years to a few days when compared to a stand-alone computer. Such capacity makes it possible to undertake large cardiac simulation projects without the need for high performance computing infrastructure.

Partial Text

Mathematical models of the heart’s electrical activity are a valuable tool for improving our understanding of cardiac electrophysiology. In particular, models of the electrical processes in cardiac myocytes can be used to understand the cells’ functioning under normal conditions, or under alterations such as those produced by diseases or drugs. Fig 1 depicts an example of the typical voltage variation across the cell membrane of a ventricular myocyte (thick line), known as the action potential (AP), simulated with the Carro et al. model [1].

DENIS uses the resources of a network of volunteers to perform computational operations requiring a huge quantity of computing power. The problems to be solved by DENIS are divided into small parts, and each part is sent to a volunteer’s computer (a host) to be carried out. The reader is referred to our previous work [10] for a full description of the DENIS architecture. The remainder of this section summarizes the details necessary to understand the experiments described in the following sections.

With DENIS, the time to complete a batch of simulations was greatly reduced in comparison with the stand-alone alternative. The 100% completion time was divided by a factor ranging from three to 44, for the shortest (600-s long) and test-case simulations respectively, in comparison with a stand-alone computer. Contrary to stand-alone computers, in which completion time increases proportionally to the length of the simulations, the completion time with DENIS did not significantly increase for longer simulations, which poses a great advantage.

This paper demonstrates the capabilities of DENIS, a cardiac electrophysiology simulator based on volunteer computing. DENIS greatly outperformed regular stand-alone computers, dividing the time to complete large batches of simulations by a factor ranging from three to 44 in different experiments. Such capacity makes it possible to undertake large cardiac simulation projects without the need for HPC infrastructure.

 

Source:

http://doi.org/10.1371/journal.pone.0205568

 

0 0 vote
Article Rating
Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments