RAID 0 4K Random Read/Write
Each QD for every parameter tested includes 300 data points (five minutes of one second reports) to illustrate the degree of performance variability. The line for each QD represents the average speed reported during the five-minute interval.
4K random speed measurements are an important metric when comparing drive performance, as the hardest type of file access for any storage solution to master is small-file random. One of the most sought-after performance specifications, 4K random performance is a heavily marketed figure.
The HGST SSD800MH array averages 676,182 IOPS, with peaks in the 680K range. The Micron P400m array reaches 618,671 IOPS. We were advised to expect roughly 600K IOPS for 6Gb/s products, as there is still further fine-tuning of the firmware underway. Both arrays score impressive scores, and the scaling is good with a large jump at QD128. Much of the lower IOPS performance of the P400m array is likely due to the need for the controller to manage more channels simultaneously. In our upcoming Micron P400m RAID review, we will take a close look at scaling with fewer SSDs.
Our read latency chart reveals that the array latency nearly doubles in the jump from QD128 to QD256. This doubling of latency reaps only a minor performance gain in IOPS for the 12Gb/s SSD's, as shown in the chart above the latency chart. This indicates that limiting the array to QD128 with the HGST 12Gb/s SSD's would provide optimum performance in both IOPS and latency for a read-centric 4K workload in some latency-sensitive environments. The Micron P400m's 100K jump in IOPS from QD128 to 256 is enough to offset the increase in latency.
Garbage collection routines are more pronounced in heavy write workloads. This leads to more performance variability.
The HGST 8-drive array produces an average of 484,499 IOPS at QD256, or roughly 60,562 per SSD. The Micron array averages 422,231 IOPS at QD256, or 26,389 IOPS per SSD, slightly above the rated specifications. The Microns provide superb scaling with the heavy write workload, while the HGST is likely constrained by RAID controller limitations.
The write latency charts illustrated the scaling for both arrays is optimum for IOPS/latency at QD128.