Desynchronization of simulation and optimization algorithms in HPC environment

Abstract

The need for the scalability of an algorithm is essential when one wants to utilize an HPC infrastructure in an efficient and reasonable way. In such infrastructures, synchronization affects the efficiency of the parallel algorithms. However, one can consider introducing certain means of desynchronization in order to increase the scalability. Allowing certain messages to be omitted or delayed can be easily accepted in the case of metaheuristics. Furthermore, some simulations can also follow this pattern and thereby handle bigger environments. The paper presents a short survey on the desynchronization idea, pointing out already obtained results, or sketching out future work focused on scaling the parallel and distributed computing or simulation algorithms.