White Papers

Dell Storage for HPC with Intel Enterprise Edition for Lustre sofware
4.1 N-to-N Sequential Reads / Writes
The sequential testing was done with the IOzone testing tool version 3.429. The throughput results
presented in Figure 9 are converted to MB/s. The file size selected for this testing was such that the
aggregate sample size from all threads was consistently 2TB. That is, sequential reads and writes have
an aggregate sample size of 2TB divided equally among the number of threads within that test. The
block size for IOzone was set to 1 MiB to match the 1 MiB Lustre request size.
Each file written was large enough to minimize cache effects from OSS and clients. In addition, the
other techniques to prevent cache effects helped to avoid them as well. Files written were distributed
evenly across the OSTs (Round Robin). This was to prevent uneven I/O loads on any single SAS
connection or OST, in the same way that a user would expect to balance a workload.
Figure 9: Sequential Reads / Writes Dell Storage for HPC with Intel EE for Lustre
Figure 9 shows the sequential performance of the 960TB test configuration. With the test bed used,
write performance peaks near 6GB/sec while read performance peaks near 10GB/sec. Single client
performance has reads at 926MB/sec with writes at 645MB/sec. The write and read performance rises
steadily as we increase the number of process threads up to 32 for reads and 48 for writes. This is
partially a result of increasing the number of OSTs utilized, as the number of threads is increased (up
to the 24 OSTs in our system).
To maintain the higher throughput for an even greater number of files, increasing the number of OSTs
is likely to help. A review of the storage array performance using the tools provided by the Dell
PowerVault Modular Disk Storage Manager, Performance Monitor was performed to independently
confirm the throughput values produced by the benchmarking tools.
4.2 Random Reads and Writes
The IOzone benchmark was used to gather random reads and writes metrics. The file size selected for
this testing was such that the aggregate size from all threads was consistently 1TB. That is, random
reads and writes have an aggregate size of 1TB divided equally among the number of threads within
that test. The IOzone host file is arranged to distribute the workload evenly across the compute nodes.
The storage is addressed as a single volume with a stripe count of 1 and stripe size of 4MB. A 4KB