Administrator Guide

Performance characterization
46 Dell EMC Ready Solution for HPC PixStor Storage | Document ID
Figure 25 N to N Sequential Performance - SMB
From the results we can observe that write performance rises steadily with the number of threads, almost
reaching the plateau of about 19 GB/s with 16 threads and attaining the maximum performance of 20.2 GB/s
at 512 threads.
The read performance does not immediately rise as expected but stays almost flat at 2 threads. After that, it
rises with the number of threads fast and steadily, almost reaching the plateau at 16 threads, attaining the
maximum read performance of 23.18 GB/s at 512 threads and having a sustained performance of about 21.5
GB/s.
Summary
The Gateway nodes provide connectivity at reasonably high sequential performance when accessing different
files (N to N).
The solution scales out in performance as more Gateway nodes are added. The N to N sequential
performance from two Gateways nodes can be reviewed on Table 11, and it is expected to be stable. Those
values can be used to estimate the performance for a different number of Gateway nodes, keeping in mind
that numbers reported were for two Gateways, each contributing half of the bandwidth reported.
Table 11 Peak & Sustained Performance
Benchmark
Peak Performance
Sustained Performance
Write
Read
Write
Read
NFS: Large Sequential N clients to N files
18.53 GB/s
16.63 GB/s
18.3 GB/s
16.5 GB/s
SMB: Large Sequential N clients to N files
20.20 GB/s
23.18 GB/s
19 GB/s
21.5 GB/s