Administrator Guide

Appendix A Storage array cabling
18 Dell EMC Ready Solutions for HPC BeeGFS High Capacity Storage | ID 424
N-1 Sequential Performance
From the results we can observe that performance rises with the number of clients used and then reaches a
plateau that is semi-stable for reads and writes all the way to the maximum number of threads used on this
test. Therefore, large, single-shared file sequential performance is stable even for 512 concurrent clients. The
maximum read performance was 22.23 GB/s at 256 threads. The maximum write performance of 16.54 was
reached at 16 threads.
4.1.4 Metadata performance
Metadata testing measures the time to complete certain file or directory operations that return attributes.
MDtest is an MPI-coordinated benchmark that performs create, stat, and remove operations on files or
directories. The metric reported by MDtest is the rate of completion in terms of OP/s. MDtest can be
configured to compare metadata performance for directories and files. The benchmark was used for files only
(no directories metadata), reporting the number of creates, stats, and removes the solution can handle.
MDtest 3.3.0+dev with OpenMPI version 4.0.2rc3 was used to run the benchmark across the 8 BeeGFS
clients that are described in Table 2
.
To understand how well the system scales and to compare the different thread cases on similar ground, we
tested from a single-thread case up to a 512-thread case with a consistent 2,097,152 file count for each case.
The following table details the number of files per directory and the number of directories per thread for every
thread count. We ran three iterations of each test and recorded the mean values.