Combined adaptive load balancing algorithm was tested on the HPC provider computing resources. Algorithm is based on domain decomposition and master-slave algorithms. Its core scheduling adaptive mechanism handles load redistribution according obtained and analyzed data. Selection of distribution algorithm, based on collected parameters and previously defined conditions, proved to deliver increased performances and reduced imbalance. Results of simulations confirm better performance of proposed algorithms compared to the standard algorithms reviewed in this paper.
Simulations on up to 224 CPU cores proved its validity and better efficiency than standard domain decomposition and master slave algorithms. In addition, simulations have shown that there are no large losses due to the increase in the number of cores on which the simulation is performed. More information on the experiment goals and the algorithm is available in the following reference: L. Filipovic, B. Krstajic, and T. Popovic, “Combined adaptive load balancing algorithm for parallel applications”, 8th International Conference on Electrical, Electronic and Computing Engineering IcETRAN 2021.