92 In the first of our posts comparing high-density hard disk drives (HDDs) and QLC flash SSDs for petabyte-scale unstructured data, we looked at the dimensions of performance and cost. In this blog, I’ll dive in deeper to compare high-density power consumption. TLDR: Today, power consumption is not a main criterion on which to base an SSD vs. HDD decision. Read on to learn why. Reminder: Scality products work with both SSD and HDD, so we don’t have a horse in this race. Ultimately, in all comparisons, the data and access patterns of the user/customer’s application workload will determine the best storage platform and media. As a hardware-agnostic storage software provider, Scality leverages both types of disk drives — HDDs for a wide range of customer use cases for unstructured data and high-density flash SSDs (and specifically QLC flash) for read-intensive, latency-sensitive workloads. The use of SSDs can be warranted for these workloads, despite current significant cost disadvantages of high-density QLC flash SSDs versus (LFF) HDDs. And the cost differential can be minimized if the workload is amenable to data reduction through compression and/or deduplication methods. Mythbusting power efficiency for HDD vs. flash Let’s dig into the issue of power consumption. Industry discussions of late have hinted that flash, in general, should have a distinct advantage over a spinning media such as HDD. Intuitively, this makes sense in that HDDs require electrical motors to spin mechanical platters and move actuator arms, whereas SSD flash doesn’t require these mechanisms. Surprisingly, our research here at Scality has found that high-density SSDs don’t have a power consumption or power-density advantage over HDD. In fact, we see the reverse today. This may, of course, change over time as drive densities increase in the coming years. That said, let’s first look at the findings for current drive offerings. SSD vs. HDD power consumption comparison Our findings: HDDs provide 19-94% better power density per drive than SSD based on specific workload patterns and today’s drive densities. For detail on how we calculated these comparisons, see the table below. This clearly demonstrates that the perception of high-density QLC flash SSDs as having a power efficiency advantage over HDDs isn’t accurate today. And, based on our read-intensive workload profile above, HDDs actually provide 19% better power density than SSDs. For the write-intensive workload profile, the advantage rises to 94% for HDDs. This will obviously vary with other workload pattern assumptions and is certainly subject to change as SSD densities increase in the coming years. Moreover, there are additional considerations for enclosure-level (servers and disk shelves) density and power consumption metrics, and how the cost of power affects each customer’s overall storage TCO. Setting the stage for analysis: How we derived our findings We’ll start with obvious sources of information — the drive vendor’s spec sheets. As a reminder from our first blog, since we are analyzing the suitability of drives in the context of petabyte-scale unstructured data, we are most interested in comparing high-density drives of each type: HDD: 7,200 rpm, large form factor (3.5” LFF) hard disk drives: Up to 22 TB per drive (as of today’s writing) SSD: small form factor (2.5” SFF) QLC flash SSDs: Up to 30.72 TB per drive (today) To be specific, let’s use Micron’s ION drive as the flash vendor source for QLC flash SSDs: Micron 6500 ION 30.72 TB QLC: 5 watts (idle), 15 watts (read), 20 watts (write) per drive And Seagate as our source for HDD power data: Seagate Exos X22 7200rpm 22 TB HDD: 5.7 watts (idle), 9.4 watts (active read), 6.4 watts (active write) Obviously, all drives naturally consume more power when active (reads and writes) versus their idle state, as well as some smaller differences in active read versus active write consumption. Given that, let’s model two different workload patterns to arrive at these expected average power consumption numbers: Workload 1: Read-intensive 10% idle 80% read 10% write Workload 2: Write-intensive 10% idle 10% read 80% write For each workload profile, drives are assumed to be in the specified power state for the percentage of time indicated. The average per-drive power calculations for each workload profile are as follows: Micron ION: Power consumption (read-intensive): (5*0.10 + 15*0.8 + 20*0.10) watts = 14.5 watts Power density (read-intensive): 30.72 TB / 14.5 watts = 2.1 TB / watt Power consumption (write-intensive): (5*0.10 + 15*0.10 + 20*0.80) watts = 18 watts Power density (write-intensive): 30.72 TB / 18 watts = 1.7 TB / watt Seagate EXOS: Power consumption (read-intensive): (5.7*0.10 + 9.4*0.80 + 6.4*0.10) watts = 8.7 watts Power density (read-intensive): 22 TB / 8.7 watts = 2.5 TB / watt Power consumption (write-intensive): (5.7*0.10 + 9.4*0.10 + 6.4*0.80) watts = 6.6 watts Power density (write-intensive): 22 TB / 6.6 watts = 3.3 TB / watt Note: All calculations rounded to the nearest tenth. Some notable observations come out of this for today’s drives: The range of power draw is larger for the SSD (5 to 20 watts) than the HDD (5.7 to 9.4 watts) Peak power consumption for SSD at 20 watts (write) is significantly higher than the peak for HDD at 9.4 watts (reads) For the read-intensive workload profile, HDDs have a 19% power-density advantage over SSDs, so for all practical purposes we can consider them similar. For the write-intensive workload profile, HDDs leverage their much more efficient write power consumption rating to gain a 94% power-density advantage over SSDs. The HDD power-density advantages are despite the significantly higher capacity of the SSD (30.72TB) versus the HDD (22TB) The verdict on SSD vs HDD power consumption In summary: Power consumption does not rise to a main criterion on which to base the SSD vs. HDD decision today. We do not see a significant power consumption difference between the drive types (SSD vs HDD) for this to be decision criteria. We guide our customers to the right choice of storage media based on their unique requirements — primarily workload driven, but with consideration of density, power, cooling and ultimately cost. Application workload dynamics, including data types, volume and access patterns are still the key criteria to lead the choice of high-density SSDs vs HDD. This is true for power efficiency (as above), as well as cost and performance criteria. Our own solutions are deployed with SSDs and/or HDDs based on the following guidelines: SSDs can deliver tangible performance benefits for read-intensive, latency-sensitive workloads HDDs will remain a good choice for many other petabyte-scale unstructured data workload profiles for the next several years, especially where price (both $ per TB and price/performance) are a concern.
