Concept Guide

Performance Characterization
18 Dell EMC Ready Solutions for HPC BeeGFS High Performance Storage | ID 460
The following commands were used to execute the benchmark for writes and reads, where threads was the
variable with the number of threads used incremented in powers of two. The transfer size is 2M and each
thread wrote to or read 128G from a single file striped over all the 33 targets. Three iterations of each test
have been run and the mean value has been recorded. Figure 11 shows the N to 1 sequential I/O
performance. The command used to run the test is given below:
mpirun -machinefile $hostlist --map-by node -np $threads -mca btl openib,self
~/bin/ior -w -r -i 3 -d 3 -e -O beegfsNumTargets=33,beegfsChunkSize=2M -o
$working_dir/ior.out -s 1 -g -t 2M -b $file_size
Sequential N to 1 Sequential I/O Performance
From the results we observe that there is a steady increase in both the read and write performance as the
number of concurrent threads is increased. The performance remains stable for reads as well as writes all the
way to the maximum number of threads used in this test. The peak read performance of 66.3 GB/s is
observed at 512 threads. Write performance attains the peak value of 37.4 GB/s at 64 threads.
36.0
37.4
35.2
34.8
31.4
54.1
54.8
57.5
62.1
66.3
0
10
20
30
40
50
60
70
1 2 4 8 16 32 64 128 256 512
Throughput in GB/s
Number of concurrent threads
N to 1 Sequential I/O Performance
Writes Reads