Administrator Guide

Performance characterization
42 Dell EMC Ready Solution for HPC PixStor Storage | Document ID
The system gets very good results as previously reported with Stat operations reaching the peak value at 64
threads with almost 6.9M op/s and then is reduced for higher thread counts reaching a plateau. Create
operations reach the maximum of 113K op/s at 512 threads, so is expected to continue increasing if more
client nodes (and cores) are used. Reads and Removes operations attained their maximum at 128 threads,
achieving their peak at almost 705K op/s for Reads and 370K op/s for removes, and then they reach
plateaus. Stat operations have more variability, but once they reach their peak value, performance does not
drop below 3.2M op/s for Stats. Create and Removal are more stable once their reach a plateau and remain
above 265K op/s for Removal and 113K op/s for Create. Finally, reads reach a plateau with performance
above 265K op/s.
Summary
The current solution was able to deliver fairly good performance, which is expected to be stable regardless of
the used space (since the system was formatted in scattered mode), as can be seen in Table 9. Furthermore,
the solution scales in capacity and performance linearly as more storage nodes modules are added, and a
similar performance increase can be expected from the optional high demand metadata module.
Table 9 Peak & Sustained Performance
Benchmark
Peak Performance
Sustained Performance
Write
Read
Write
Read
Large Sequential N clients to N files
40.9 GB/s
84.5 GB/s
40 GB/s
81 GB/s
Large Sequential N clients to single shared file
34.5 GB/s
51.6 GB/s
31.5 GB/s
50 GB/s
Random Small blocks N clients to N files
5.06M IOps
7.31M IOps
5M IOps
7.3M IOps
Metadata Create 4KiB files
113K IOps
113K IOps
Metadata Stat 4KiB files
6.88M IOps
3.2M IOps
Metadata Read 4KiB files
705K IOps
500K IOps
Metadata Remove 4KiB files
370K IOps
265K IOps
PixStor Solution – Gateway Nodes
This benchmarking was run with the file system running on the Large configuration plus four ME484s, that is
two R740 servers connected to four ME4084s and four ME484s (one behind each ME4084), with the optional
HDMD module (two R740) but using a single ME4024 array and two gateways using a native client to access
the file system, and exporting that file system using the NFS or SMB protocols.
To characterize this component, the IOzone benchmark was used to characterize sequential performance for
N clients to N files using large transfer blocks. Since the gateways can be used for NFS or SMB clients, both
protocols were used for this work.
To allow proper testing with a limited number of clients, the cluster described in Table 3 was created with
heterogeneous 13G PowerEdge servers including from R630, R730 and R730XD. Clients have different CPU
models type E5-26XX v4 with different number of cores per socket (10, 12, 18 or 22), adding a total number
of cores to 492, and speeds ranging from 2.2 to 2.6 GHz (2.2, 2.3 & 2.6).
Similarly, half of the clients had memory 8GiB DIMMs and the other half had 16GiB DIMMs, always using all
memory channels and with speeds mostly 2400 MT/s (except two clients @ 2133 MT/s). Therefore, each