Normally we give a boilerplate explanation of how we secure erase the drive and make sure it is in steady-state before we begin testing. Since the ALLONE drive is always in steady state, we just moved right into our performance testing.
For random 4KB write operations, the ALLONE did a better job at staying with the pack. Once the queue depths were sufficiently high, it easily outpaced the Intel and Micron offerings.
Random 4KB reads were almost the complete opposite. At low queue depths, it was free and clear of the other drives, but stalled as queue depths rose. Once again, while the competition was putting up 300-750K IOPS, the ALLONE was stuck at 130K.
As we mentioned earlier, sequential performance is not the strong suit for the Cloud Disk Drive 101. While it was able to hang close with the Micron P420m for sequential writes, it was no match for the competition for the remaining tests. In fact, it was only 10-20% than most enterprise SATA SSDs. We are so used to PCIe-based storage products hitting 1 or 2 or 3GB/s in our sequential tests, that we were a little disappointed. Fortuantely for ALLONE, sequential operations are not the intended workloads for the Cloud Disk Drive 101.
At this point, we aren’t quite sure what to think about this drive. While it has some unique characteristics, it’s losing certain tasks by a very wide margin.
SNIA IOPS TESTING
The Storage Networking Industry Association has an entire industry accepted performance test specification for solid state storage devices. Some of the tests are complicated to perform, but they allow us to look at some important performance metrics in a standard, objective way.
SNIA’s Performance Test Specification (PTS) includes IOPS testing, but it is much more comprehensive than just running 4KB writes with IOMeter. SNIA testing is more like a marathon than a sprint. In total, there are 25 rounds of tests, each lasting 56 minutes. Each round consists of 8 different block sizes (512 bytes through 1MB) and 7 different access patterns (100% reads to 100% writes). After 25 rounds are finished (just a bit longer than 23 hours), we record the average performance of 4 rounds after we enter steady state.
- Preconditioning: 3x capacity fill with 128K sequential writes
- Each round is composed of .5K, 4K, 8K, 16K, 32K, 64K, 128K, and 1MB accesses
- Each access size is run at 100%, 95%, 65%, 50%, 35%, 5%, and 0% Read/Write Mixes, each for one minute.
- The test is composed of 25 rounds (one round takes 56 minutes, 25 rounds = 1,400 minutes)
Once again, we have to remind ourselves that we are not dealing with NAND storage. There are two things that stand out that make these graphs unique. The first is that, other than 512B, there is almost not difference between read, write and mixed workloads. Take another look at the bar chart, they are almost completely identical.
The second thing that jumped out as we reviewed our results is that the performance is the same, regardless of previous operations. We normally run our SNIA tests because it helps show how an SSD handles quick transitions between different workloads. Because of this, you sometimes see results that are lower than if you ran that individual test for long periods of time. With the ALLONE, much like rotating HDDs, the performance you get is completely independent from any previous operations. This is a trait that most SSDs would kill for.