Open Beilinson opened 5 years ago
With Tomer - parallel the code in the following way: around 50 runs (maybe 100, to check what is the optimum in the current state of the cluster)
With SBATCH parameter and with dividing of the data by snapshots. Meaning the same snapshot on the same node, and maybe no deletion between them.
With Tomer - parallel the code in the following way: around 50 runs (maybe 100, to check what is the optimum in the current state of the cluster)
With SBATCH parameter and with dividing of the data by snapshots. Meaning the same snapshot on the same node, and maybe no deletion between them.