Concept Guide

Performance Characterization
14 Dell EMC Ready Solutions for HPC BeeGFS High Performance Storage | ID 460
4 Performance Characterization
To characterize the performance of the Ready Solution, the testbed used is shown in Figure 9. In the figure,
the InfiniBand connections to the NUMA zone are highlighted. Each server has two IP links and the traffic
through NUMA 0 zone is handled by interface IB0 while the traffic through NUMA 1 zone is handled by
interface IB1.
Testbed Configuration
The hardware and software configuration of the testbed is specified in Tables 1, 2 and 3. In order to assess
the solution performance, the following benchmarks were used:
1. IOzone N to N sequential
2. IOR N to 1 sequential
3. IOzone random
4. MDtest
For all the benchmarks listed above the testbed had the clients described in Table 4. Since the number of
compute nodes available for testing was 32, when a higher number of threads was required, those threads
were equally distributed on the compute nodes (i.e. 64 threads = 2 threads per node, 128 threads = 4 threads
per node, 256 threads = 8 threads per node, 512 threads = 16 threads per node, 1024 threads = 32 threads
per node and 2048 threads = 64 threads per node). The intention was to simulate a higher number of
concurrent clients with the limited number of compute nodes.