Scalability Test of Multiscale Fluid-Platelet Model for Three Top Supercomputers

Area of Science:

• Computational biophysics
• High-performance computing (HPC)
• Multiscale modeling

Background:

• Platelet behavior in biofluids is crucial for understanding thrombosis.
• Accurate nanoscale biophysics simulations require significant computational power.
• Existing simulation methods face scalability challenges.

Purpose of the Study:

• To assess the scalability of Tianhe-2, Stampede, and CS-Storm supercomputers.
• To evaluate the performance of multiscale fluid-platelet simulations.
• To demonstrate the feasibility of complex biological simulations on current HPC.

Main Methods:

• Multiscale fluid-platelet simulations were performed on three supercomputers.
• Experiments varied problem sizes, from single-platelet to 16-platelet simulations.
• A multiple time-stepping (MTS) algorithm was implemented and compared to single time-stepping (STS).

Main Results:

• The MTS algorithm significantly improved simulation performance over STS across all tested problem sizes.
• Specific simulation rates (μs/day) were achieved on each supercomputer for different experiment scales.
• The Stampede supercomputer achieved the best performance for the largest simulation (Exp-L).

Conclusions:

• Advanced computational algorithms like MTS enhance the performance of multiscale simulations.
• Complex biological and engineering problems, such as thrombosis, are becoming computationally feasible.
• Current HPC resources are adequate for performing these demanding simulations.