Mixed Workload And Steady State
80 Percent Sequential Mixed Workload
Our mixed workload testing is described in detail here, and our steady state tests are described here.
Mixed-workload tests give us a clearer picture of how a drive will perform in your computer. A computer system rarely just reads or just writes data, unless used as a secondary volume to store mass data. Under this view, the Reactor shows fare performance, but it still falls well short of the Samsung 850 EVO 250GB.
80 Percent Random Mixed Workload
The same 80 percent mix of read-to-data writes with random 4KB data shows the same picture. The Reactor 256GB performance is right in the middle, where we expect to find it based on price data.
Sequential Steady State
Under normal conditions, users will never achieve full steady state with a client-grade SSD. Client workloads are not stressful enough to drive this level of performance. As the drive fills up, performance drops off because the drive is struggling to find clean cells that will keep write performance high.
We also get to see an exaggerated bathtub curve here of mixed sequential performance and how some drives handle mixed data better than others.
Random Write Steady State
The 4KB random-write test in steady state shows us two important measurements that have very little to do with the actual output. The first thing we look for is performance consistency—a nice steady line of IOPS with a very narrow gap between high and low points. The SM2246EN controller was clearly not designed for enterprise workloads, so the peaks and valleys are far apart. We also look at the low IOPS, because putting a pair of these drives in RAID 0 to equal the performance of another drive is a step in the wrong direction. We don't recommend that you use the Reactor 256GB in RAID 0 for performance benefits.