Home > Workload Solutions > High Performance Computing > White Papers > HPC High-Performance Storage Solution for BeeGFS > Further storage bench results
For ME5, running with a lower number of threads per target increased throughput compared to ME4. The following command is used to run storage bench write:
beegfs-ctl --storagebench --alltargets --write --blocksize=1m --size=$((SIZE))M --threads=$THREADS
The following command is used to run storage bench read:
beegfs-ctl --storagebench --alltargets --read --blocksize=1m --size=$((SIZE))M --threads=$THREADS
$SIZE is a variable that is configured to write an aggregate totally of 8TB across 32 storage targets for varying thread counts. This is defined as:
SIZE=$((250000/$THREADS))
The following is a summary of results obtained by running storage bench at varying thread counts:
Threads | write-KiBs | read-KiBs | write-GBs | read-GBs |
1 | 28154211 | 29103615 | 28.83 | 29.81 |
2 | 30431313 | 29070171 | 31.17 | 29.77 |
3 | 30276385 | 30154675 | 31.01 | 30.88 |
4 | 30134254 | 30157806 | 30.86 | 30.89 |
5 | 30050541 | 29916035 | 30.78 | 30.64 |
6 | 30023428 | 29373712 | 30.75 | 30.08 |