Skip to main content
FlexPod

Test results

Contributors Noah-Hardy kaminis85
PDFs

Previous: Testing approach.

Testing consisted of running the FIO workloads to measure the FC-NVMe performance in terms of IOPS and latency.

The following graph illustrates our findings when running a 100% random read workload using 8KB block sizes.

A graph of FlexPod AFF A800 8KB random read IOPS versus latency, reaching 1.25M IOPS at 0.76 ms

In our testing, we found that the system achieved over 1.2M IOPS while maintaining just under 0.35ms of server-side latency.

The following graph illustrates our findings when running a 100% random write workload using 8KB block sizes.

A graph that shows FlexPod AFF A800 8KB random write latency rising with IOPS

In our testing, we found that the system achieved close to 300k IOPS while maintaining just under 1ms of server-side latency.

For 8KB block size with 80% random reads and 20% writes, we observed the following results:

A chart of FlexPod AFF A800 8KB 80:20 latency rising to 0.95 ms at 1M IOPS

In our testing, we found that the system achieved over 1M IOPS while maintaining just under 1ms of server-side latency.

For 64KB block size and 100% sequential reads, we observed the following results:

A chart showing FlexPod AFF A800 64KB sequential read latency rising from 0.2 to 0.78 ms as IOPS increase

In our testing, we found that the system achieved around 250k IOPS while maintaining just under 1ms of server-side latency.

For 64KB block size and 100% sequential writes, we observed the following results:

A chart showing FlexPod AFF A800 64KB sequential write latency rising with IOPS

In our testing, we found that the system achieved around 120k IOPS while maintaining under 1ms of server-side latency.

Next: Conclusion.