Home > Workload Solutions > High Performance Computing > White Papers > Dell Validated Design for HPC pixstor Storage—Joint Solution with Kalray > PowerEdge R750 Metadata performance with MDtest using 4 KiB files
This test is almost identical to the previous test, except that we used small files of 4 KiB instead of 3 KiB files, so part of each file does not fit in its inode, and another devices (two R750 nodes) are used to store the rest of the files. The following command was used to run the benchmark, where the Threads variable is the number of threads used (1 to 512 incremented in powers of 2), and my_hosts.$Threads is the corresponding file that allocated each thread on a different node, using round robin to spread them homogeneously across the 16 compute nodes.
mpirun --allow-run-as-root -np $Threads --hostfile my_hosts.$Threads –map-by node --mca btl_openib_allow_ib 1 --oversubscribe --prefix /usr/mpi/gcc/openmpi-4.1.2a1 /usr/local/bin/mdtest -v -d /mmfs1/perf/mdtest -P -i 1 -b $Directories -z 1 -L -I 1024 -u -t -F -w 4K -e 4K
Stat operations showed good numbers, reaching a peak value with 9.8M op/s at 64 threads and read operations reaching 3.3M op/s at 128 threads. Remove operations attained the maximum of 317.4K op/s at 32 threads and create operations achieving their peak of 137.4K op/s, both at 128 threads. Stat and read operations have more variability, but when they reach their peak value, performance does not drop below 2.5M op/s for stat operations and 1.7M op/s for read operations. Create and remove operations have less variability; both keep increasing until reaching their peak values, slowly decreasing after that.
Because these numbers are for a metadata module with one PowerEdge R650 NVMe metadata pair and a PowerEdge R750 NVMe pair for data, metadata performance will increase for each additional PowerEdge R650 NVMe metadata pair; also, each PowerEdge R750 NVMe pair for data will help increase the performance of some operations like create, remove, and read. The larger the files, the greater the effect of additional PowerEdge R750 NVMe pairs used for data.