White Papers

Dell HPC Lustre Storage solution with Mellanox Infiniband EDR

 echo 3 > /proc/sys/vm/drop_caches

In addition, to simulate a cold cache on the server, a “sync” was performed on all the active servers

(OSS and MDS) before each test and the kernel was instructed to drop caches with the same commands

used on the client.

In measuring the performance of the Dell HPC Lustre Storage solution, all tests were performed with

similar initial conditions. The file system was configured to be fully functional and the targets tested

were emptied of files and directories prior to each test.

4.1 N-to-N Sequential Reads / Writes

The sequential testing was done with the IOzone testing tool version 3.444. The throughput results

presented in Figure 10 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 had

an aggregate sample size of 2TB divided equally among the number of threads within that test. The

block size for IOzone was set to 1MB to match the 1MB 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. The 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 10: Sequential Reads / Writes Dell HPC Lustre Storage with Infiniband EDR

Figure 10 shows the sequential performance of the 960TB test configuration. With the test bed used,

write performance peaks slightly over 15.1GB/sec while read performance peaks near 18.5GB/sec.

Single client performance were consistent at around 1GB/s to 1.5GB/s for writes and reads

1 4 8 16 24 32 48 64 72 96 128 256

Throughput in GB/s

Number of concurrent threads

IOzone Sequential - Dell HPC Lustre Storage with Infiniband

EDR

Write Read