Mixed Workload and Steady State
Our mixed workload testing is described in detail here, and our steady state tests are described here.
80 Percent Sequential Mixed Workload
Mixing reads and writes certainly opens a gap in sequential performance between our charted SSDs. We see even more separation than what we observed in the 100 percent read test.
Longsys' S500 doesn't fare as well as it did in the previous sequential metric. Again, we put the most value on queue depths of two and four.
80 Percent Random Mixed Workload
Now we're mixing random 4KB block sizes, with an emphasis on the low to middle range. Even at queue depths of two, four and eight, we see a fair amount of separation, with Samsung's 850 EVO leading the way. The Trion 100 lags behind the other models, including Longsys' Foresee S500, which shows up in the middle.
Sequential Steady State
For most of you, our sequential steady state test won't mean much. It takes a workstation-class customer manipulating massive audio and video files to derive true utility from this benchmark.
Subjected to 80 percent reads, the S500 performs better than any other drive in the chart. But by the 70 percent read mark, it drops to the middle. The S500 is not an SSD you want if you're working with large files on a workstation. Performance can drop to sub-mechanical storage levels under those workloads.
Random Write Steady State
We use the random steady state tests to identify SSDs that might work well in RAID arrays. We're looking for high and consistent random write performance. The less variation observed at the high and low points, the better.
Both the Longsys S500 and Adata SP550 achieve admirable peaks. At the same time, they dip to disappointing lows as well. This deviation makes the SMI SM2256 controller a poor choice for RAID arrays.