A Load Balancing Model for Parallel Simulation of Fluid-Structure Interaction in Cavitating Flow

Abstract

Load unbalancing problem has a significant impact on the parallel efficiency of fluid-structure interaction simulation in cavitating flow. When the total parallelism is determined, the speedup will be seriously affected by the distribution of cores for the fluid solver and solid solver. This paper proposes an adaptive-λ load balancing model to maximally achieve the optimal parallel efficiency by generating a proper distribution scheme for the participant solvers. Our model is an optimization of the Kannan's method, which changes the original fixed-value λ to an adaptive one. Specific formulas are set up by a series of liner fittings and the parameter λ is calculated by a function of grid scale and parallel scale. A parallel FSI platform for cavitating flow based on preCICE is constructed to verify the present model. Experiments show that, compared with the traditional Kannan model, the adaptive-λ model could perform better parallel speedup and achieve wider application scope. This could help give a guidance on parallel decomposition for each participant solver in FSI applications.