The thesis aims to establish a simulation environment for data transfers and processing large amounts of data on existing supercomputer networks. In some cases, it is better to transfer input data to another location for processing, while in others, it is faster to process the data at the current location despite slower hardware. The simulator enables a practical analysis of existing and future computational infrastructure. It is a tool for testing new algorithms for the optimal distribution of computational loads, which can contribute to better utilization of existing hardware. According to existing computational loads, virtual computational jobs were imported into the simulator to be run using the existing scheduling algorithm and the actual hardware. The simulator serves as an alternative way for testing new algorithms, which isn't possible on live infrastructure, enabling improvements in the distribution of large computational loads worldwide.
|
