Unlike mechanical hard drives, SSDs contain no moving parts – which is why they are considered particularly efficient. But that’s not always true claims Scality, a company that provides storage services. An example calculation should prove this, and it doesn’t make SSDs look good: With one “write intensive application” an HDD, converted to the TByte of storage capacity, should only need a good half of the performance of an SSD.
Scality compares two enterprise storage media here: The Exos X22 from Seagate with 22 TB capacity as HDD (Data sheet, PDF), for the SSDs Microns 6500 Ion with 30.72 TB (Data sheet, PDF) at. Scality did not measure the power consumption itself, but took the information from the data sheets. And here things are not looking good for the SSD: It needs up to 20 W when writing, the HDD is satisfied with 6.4 W. When reading, the ratio is slightly more balanced at 15W versus 9.4W.
For two sample scenarios, the company then calculates how many TB of the respective memory can be operated with one watt. Result: In the read-heavy test (10 percent idle, 80 percent read, 10 percent write), the HDD is 19 percent ahead of the SSD, in the write-heavy test (10 percent idle, 10 percent read, 80 percent write), the lead even rises to 94 percent.
Speed becomes a disadvantage for SSDs
However, the comparison is not entirely fair: Scality compares a 6-Gbit SAS HDD (for Serial Attached SCSI) with an NVMe SSD with PCIe Gen4. While the former reads and writes at a maximum of 285 MB/s, the SSD manages up to 6.8 GB/s when reading and 5 GB/s when writing.
However, if we compare it with an SSD, which is also connected via the slower SAS interface, the advantage of the HDD shrinks: Samsung gives im for its PM1643 Data sheet (PDF) a maximum power consumption of 13.5 W, with Kioxia’s PM7 series (Data sheet, PDF) it can even be limited to up to 9 W. This should also reduce performance, but Scality does not consider it. According to the service provider’s bill, however, we get 3.6 TB/W for the SSD – which would put it ahead of the HDD with 3.3 TB/W when writing.
The very different data rates also distort the analysis: Ideally, the SSD writes 17.5 times the amount of data in the same time, so it would be finished much faster with the same application. However, the naïve view of Scality is that the application writes or reads the same amount of time on both drives. You can then save yourself such alleged analyzes – they simply do not provide any knowledge gain.
The only finding: The larger an HDD, the lower the advantage of an SSD in terms of power consumption. But you can derive this purely logically, because unlike an SSD, where the number of memory chips grows linearly, the HDD still only has one motor for the disk stack and read/write heads. Smaller SSDs therefore easily overshadow mechanical hard disks in terms of efficiency – at least if you ignore the energy for production and if they are not overbred representatives of the latest generation.