4K Random Read/Write
We preconditioned the Intel DC S3700 for 18,000 seconds or five hours. Every second we are receiving reports on several parameters of the workload performance. We then plot this data to illustrate the drives' descent into steady state.
This chart consists of 36,000 data points. The dark blue dots signify the IOPS during the test, and the light blue dots are the latency encountered during the test period. We place the latency data in a logarithmic scale to bring it into comparison range. This is a dual-axis chart with the IOPS on the left and the latency on the right. The lines through the data scatter are a moving average during the test. This type of testing presents standard deviation and maximum/minimum I/O in a visual manner.
Note that the IOPS and Latency figures are nearly mirror images of each other. This illustrates the point that high-granularity testing can give our readers a good feel for the latency distribution by viewing IOPS at one-second intervals. This should be in mind when viewing our test results below.
We provide histograms to provide further latency granularity below. This preconditioning slope of performance happens very few times in the lifetime of the device, and we present these test results for the tested device only to confirm the attainment of steady state convergence.
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 Intel DC S3700 does provide very tight groupings of performance, much like the SMART Optimus. Intel's optimizations to their new QoS standard are more prevalent in the write tests below. We can see several dips in performance with the Intel SSD in the QD64 and QD256 range. These are indicative of the new approach with the Garbage Collection algorithms that Intel is utilizing for this SSD. This pattern is more pronounced in the write testing results.
The Intel DC S3700 does deliver an average speed of 64,533 IOPS at QD256 in steady state, with a little variability encountered during garbage collection routines. The SMART Optimus, with its native SAS connection, provides an average read speed of 93,860 IOPS at QD256.
The Intel DC S3700 garbage collection mechanisms routines are more pronounced in heavy write workloads. Much of the latency optimization seems to focus around the garbage collection routines triggering at steady intervals of 90-100 seconds. This does provide very steady performance within a nice tightly defined range, but the SMART Optimus also exhibits much the same characteristic with less variability. The scale of this chart should be taken into consideration, with the Intel varying only 5,000 IOPS from minimum to maximum. This is admirable, but a characteristic also shared by many other enterprise storage solutions.
The Intel DC S3700 averages 36,428 IOPS, while the SMART Optimus averages 43,081 IOPS at QD 256.
The histogram results are especially relevant due to the Intel SSD featuring tightly defined definitions of latency performance. We typically only publish our results from the 256QD range. We will include those below. Taking the Intel specified latency ranges with 4K random write activity into consideration at QD32 we included those results to ascertain the veracity of their latency specifications.
The Intel specification (page 2) states that the 800GB model we are testing will deliver 99.9% of I/O within 10ms, and 99.9999% of I/O within 20ms. The chart above reveals the results of every single I/O issued during our test run, and portrays them as percentages.
The Intel, with 4K QD32 random write access, gave 439,050 I/Os (or 4%) in the .4 - .6ms range, 3,796,071 I/Os (35.7%) in the .6 - .8ms range, 3,319,112 I/Os (31%) in the .8 - 1.0ms range, and 3,040,731 (28.6%) within the 1-2ms range. There was a smattering of I/O in the higher ranges, but with 6,636 I/Os in the 10-20ms range, accounting for .0624% of the total I/O.
In short, Intel met their QoS specifications for 99.9% of I/O. There was 1 solitary I/O delivered in the 20-40ms range, but the percentage is lower than our tool can calculate. It is safe to say that Intel also met their QoS specification for 99.9999% of operations conducted within 20ms. Interestingly enough the SMART Optimus also met these specifications, but within a tighter range. There were no I/Os from the Optimus over the 8-10ms range. Overall both SSDs perform beautifully in this test, but the SMART Optimus features overall lower latency.
The performance in the higher-load 4K QD256 test was similar to the results at QD32, with the Optimus beating the Intel by a slim margin. The Intel had 425,253 I/O (3.9%) in the 4-6ms range, 8,766,022 I/O (82.1%) in the 6-8ms range, and 1,307,699 (12.2%) within the 8-10ms range.
The power consumption measurements, taken during our precondition run, for SMART Optimus are higher than the DC S3700, averaging 6.51 Watts compared to the Intel's average of 4.36 Watts. Some of this higher power requirement is due to the SAS interface.
The Intel scored 8,336 4K write IOPS per Watt, and the Optimus scored 6,617 IOPS per Watt.