White Papers

Dell HPC Lustre Storage solution with Intel Omni-Path
Figure 13: File Metadata Operations
Figure 13 illustrates the file metadata results using MDtest.As shown in this graph, file create metadata
operations start with less than 1100 OPS at 1 thread and scale to over 27K OPS with 240 concurrent
threads. This may be due to the Lustre locks needed on the MDT, but also those on the OSTs, since
using a stripe count of 24 had a significant decrease in performance. At 240 threads, we had 2
directories (-b 2) and 2.2 million files were created.
File stat metadata operations is overwhelmingly the lightest metadata operation of the three observed.
A single thread test yield over 2K OPS and scale to more than 500K OPS with 240 concurrent threads.
The increase in performance could be due to improvements made in Lustre version 2.7 metadata
operations. Also of note is the use of 15K RPM drives in the MDT volumes.
Removal of files is also limited by accesses to OSTs, similar to the create operation. However, the
remove operation has an advantage over the create operation when increasing in total threads, starting
with over 1.9K OPS at 1 thread and scaling to over 144K OPS with increasing concurrent threads.
5. Performance Tuning Parameters
There are a number of performance tuning parameters that can be configured for an optimal system
depending on intended workload patterns to be used. This section will detail the tuning parameters
that we configured on the Lustre testbed system in the Dell HPC Engineering Innovations lab.
5.1 Lustre specific tunings
[root@node057 ~]# lctl set_param osc.*.max_pages_per_rpc=1024
0
100000
200000
300000
400000
500000
600000
1 2 4 8 12 16 24 32 48 64 72 96 120 128 144 168 192 216 240
OPS
Threads
MDtest Files Metadata - Dell HPC Lustre Storage with Intel Omni-Path
File Create File Stat File Remove