We're using two block sizes to show sequential read and write data: the traditional 128KB test and 1MB blocks.
Our host test system connects to our enterprise network with a Mellanox ConnectX-3 VPI card that transmits and receives data at up to 40Gb. We're using a Supermicro SSE-X3348SR with 48 10GbE ports and four 40GbE ports as our core. The FreeNAS Mini is also connected to the same switch. Multi-client testing runs off the same Supermicro switch as the core, but other switches are incorporated to allow us to run 120 clients. The other switches use the 40GbE backbone and everything runs at full unblocked speed. The ConnectX-3 frame buffers (as well as the switch buffers) do give us slightly inflated 1GbE numbers, but the bottleneck is always the NAS storage device in our testing.
Working through the queue depth range, we see that most of the products under test today manage to deliver solid sequential read performance with little variation between products. This will change when we get into workload testing though.
We see a lot more variation in the sequential write test. Surprisingly, the FreeNAS Mini didn't perform the best out of the test group.
Using larger 1MB blocks to measure performance, the FreeNAS Mini managed to keep performance over 100 MB/s most of the time. It's difficult to ask for more than that, even though it's possible to reach 125 MB/s with gigabit Ethernet.