Analysts and vendors say the death of hard disk drives, a non-volatile storage technology that's been around for over 60 years, has been greatly exaggerated. But as other storage technology advances, including solid-state drives, the role HDDs play in the enterprise is starting to change.
HDDs, composed of moving parts including spindles and platters, are often referred to as spinning disks or, pejoratively, as spinning rust. For 40 years, HDDs have come in two standard form factors, 2.5-inch and 3.5-inch, and have mainly used the SATA and SAS interface. The technology's density, cost and reliability carved out a place as the workhorse of enterprise storage for decades.
But that reign is being challenged. While currently more expensive, SSDs are faster, a characteristic that matches the evolution of modern workloads. Quad-level cell SSDs, in particular, pose the most significant threat, as the drives can outperform HDDS in capacities of up to 30 TB in a 2.5-inch form factor.
Here, industry experts look at the recently announced 20 TB drives and how they feed into the never-ending appetite for capacity, whether individual drive speeds matter anymore, the fate of the 2.5-inch HDD and how close the industry is to crossover cost between HDDs and SSDs.
Is bigger better? 20 TB has entered the market
In 2021, hard disk drives went from a top density of 18 TB to 20 TB. Experts believe this 2 TB bump in density, which can add up in direct proportion to the number of drives used, will affect the market, particularly for cloud and archive storage.
Ted Deffenbaugh, vice president of product line management, Seagate: The way the industry used to work is the small capacities would stick around and when we introduced a brand new capacity, it could take years before it came up in appreciable volume. With these data center customers, what they told us is, 'We want the largest hard drive that you can get us.' We ramped these things up to millions of units per quarter. We are in the early stages of seeing the big mix over [from 18 TB to 20 TB], we've got an extremely strong and positive result to our 20 TB. As we've designed it, we believe that it's got some features that will allow us to ramp it very strongly.
Ashley Gorakhpurwalla, executive vice president and general manager, HDD business unit, Western Digital: We just introduced [20 TB technology]. The industry will move to that technology going into the year. Our customers are demanding more capacity, more density, more [total cost of ownership], and so 20 TB will soon be followed by increasing capacities within the year.
Colin Presly, senior director, office of the CTO, Seagate: The growth in the market is really with the data center customers. They are driving this innovation and we don't get any respite in terms of pushing innovation because of their demand. … They're driving us hard to be as innovative and as low cost with a quality of service.
Eric Burgener, research vice president, infrastructure systems, platforms and technologies group, IDC: Archive workloads are where there is going to be a lot of HDDs sold. The cloud providers, AWS Glacier, that's deep archive, they have a ton of HDDs. … If there's any chance you might need data in an archive [and] you want to get it back within a couple seconds, then you're going to store that on HDDs. That's where those 20 TB HDDs are going to be of more interest, because that makes [archive storage] cheaper, because you need to buy fewer HDDs.
Edward Burns, research director, hard disk drive and storage technologies, IDC: 20 TB HDDs should be shipping in significant volumes by the middle of the year and begin to approach the volumes of the 18 TB by the end of 2022, making 20 TB the leading volume product shipment by the end of the year -- not for the year, but beginning to ship at a higher run rate at that time. We should see significant volumes from WD and Seagate at that capacity point.
The need, or lack thereof, for speed
In the past, HDD vendors emphasized their growing I/O operations per second (IOPs) or data transfer rate typically, in MB/s or how they stacked up against SSDs. But as software and array vendors have found ways to gain more performance out of pooling several HDDs together, HDD vendors shifted away from competing with SSDs to figuring out how to partner with them. Experts expect the push for HDD-SSD partnerships in the name of performance to continue in 2022.
Presly, Seagate: [Performance comes down to] what workloads [customers] are striving for. … It's a combination of devices and technology that really makes sense. … The bulk performance of a hard disk drive is still driven by the mechanics. A 3.5-inch form factor drive hasn't really changed in its performance for 20 years. Realistically, we've done some things around caching and the amount of DRAM you put on a drive, which helps with the write performance. But the amount of IOPS you get out of [HDDs] hasn't really changed because it's dominated by the seek performance of the actuator [to read and write the data]. The IOPS per terabyte metric is important in a multi-tenant environment, like a data center or a hyperscaler, because you have a lot of users using not any one device. It's not like a laptop where you have one user, one device. The amount of performance per terabytes that you see on a device is an important metric.
Gorakhpurwalla, WD: Most of our customers are not demanding sort of ultimate performance, to get the best out of the hard drive based on incredible latency needs. There are some that use SAS to double that speed [over SATA]. What you have in an SSD though is an NVMe or PCI Express capability, which allows you to bring the data into memory or processor faster. However, there's a curve in which all of our customers operate: There's a performance tier of data that gets used all the time, and that's sometimes worth keeping in an SSD.
Burgener, IDC: There are some applications that, if it takes 20 milliseconds each time I put an I/O request in, HDDs are fine because they're going to respond somewhere in the range of five to 20 milliseconds. If that's all your app needs, great. The cheapest way to build an 80 TB storage capacity is by an HDD. Flash becomes more interesting as you get a little more of a performance need in whatever workload you have.
Burns, IDC: Spin speeds are less of an issue because latency-sensitive applications will shift toward SSDs rather than higher spin speed HDDs in general.
The future of 2.5-inch HDDs in the enterprise
Compared to 3.5-inch counterparts, 2.5-inch HDDs come with faster RPM speeds and a smaller form factor for denser arrays, which also limits the maximum capacity per drive. Denser arrays can use lower-cost flash in the same footprint as 2.5-inch drives, but with customers demanding higher density in large form factors and greater performance in small form factors, vendors don't appear to be investing much into their development.
Deffenbaugh, Seagate: We have customers that are just absolutely wild about the [Seagate 2.5-inch HDDs]. … So even if we don't actively develop it, it actually turns out that we have a hardcore set of customers that are going to be riding that portion to the sunset, they absolutely love what it brings to their architectures, and they just want more of the same.
Gorakhpurwalla, WD: If you look at our roadmap, 2.5-inch drives are still pretty high volume, but they're not as high volume as they were three years ago or five years ago or eight years ago. ... We still advance that roadmap forward on 2.5-inch drives ... but over time that will be something that will decline in terms of market size.
Marc Staimer, president, Dragon Slayer Consulting: [2.5-inch HDDs] have been going away for a long time ... 15,000 RPM, big deal. If I want speed, I'm not going to buy a hard drive; I'm going to buy flash.
Burns, IDC: Shipments of these are small and are likely to gradually decline over time. There is still a place for them currently but not likely in the long term -- more than several years out.
Crossover cost and the threat of QLC
As for the cost advantage, HDDs offer a lower price per terabyte than QLC, the least expensive SSD. But it can't compete when it comes to performance. When QLC costs as much as HDDs, which is where QLC SSDs are headed, what will happen to the demand for HDDs?
Burgener, IDC: There are lots of reasons why [flash] is better than HDDs. ... If flash cost exactly the same as HDDs, nobody would be buying HDDs. Moves like [Vast Data adding 30 TB QLC SSDs] make flash that much cheaper. Even though it is not as cheap as HDDs, it lets more people say, 'I'm ready to go to this now.'
Presly, Seagate: Big customers have teams of people working on total cost of ownership across the whole architecture, and the [storage] devices are only a piece of the equation of TCO. … They're trying to put together this puzzle of how to put these pieces together so that none of those things are stranded, because it's all cost. It doesn't make any sense to use NAND purely as a storage device because you end up with stranded performance. If you put SSDs in a 4U system, the amount of bandwidth and performance they provide far exceeds the amount of network you could deliver. It's really a case of putting enough SSDs and DRAM above that, synergistically with hard drive storage where the attribute of each device makes sense to operate together. Otherwise, you just end up with a stranded pick your metric: stranded network, stranded compute, stranded performance, stranded capacity.
Sinan Sahin, principal product manager, Seagate: It's all about how do you normalize the cost of it in dollars per TB. Hard disk drives are still the most economical way to go with it with a dollars per terabyte. But in terms of new and flashy next year, I would say that dual actuators will increase the speed, and it'll still be focused on those dollars per terabyte in terms of an affordable storage compared to something like a flash.
Gorakhpurwalla, WD: [HDDs will] never get displaced in the enterprise. Because the capacity points and the data explosion just demand that we have multiple tiers in the data center, and the primary tier will be hard drives and magnetic recording. You have this body of two minds, if I only know the end device, I'm thinking the hard drive industry is probably declining or going away. If you're in the enterprise space, if you're paying attention to data being created through digital transformation in the world, you realize we're building more factories, and as an industry, we're putting more capacity in. Because capacity, if you listened to our customers, and then translated through our business, you realize our next five years, the number of bits we need to produce is increasing by 35% every year.
Staimer, Dragon Slayer Consulting: [HDD technology] is still cheaper than flash. At the end of the day, hyperscalers and major cloud vendors realized the utility in low cost. Even the cheapest QLC is … give or take, anywhere from five to 10 times more expensive per gigabyte, per TB. [One can argue] deduplication and compression, but those things work just as well on spinning drives. … [HDDs are] a technology that is understood, and [users] understand the risks that are pretty much mitigated with things like RAID [redundant array of independent disks] and then erasure coding. I don't see it going away anytime soon -- until you get closer in cost where the performance justifies the net difference.
Burns, IDC: Periodically, companies that have a vested interest in seeing SSDs grow -- many of these companies sell systems that are all flash, for example -- will come out with statements implying a rapid demise of HDDs due to price parity with SSDs in an unrealistically short timeframe. These types of claims have been made by various companies for the last 10 years at least. And while certain segments have shifted toward SSDs and continue to do so -- notebook PCs, desktop PCs to a lesser extent, consumer electronics, enterprise performance optimized -- overall, the HDD petabytes have continued to grow to record levels, driven by enterprise capacity, optimized drives and video surveillance segments primarily. HDD industry revenue is expected to grow overall over the next five years, and we don't expect price per GB for SSDs to come close to parity with HDDs during this decade.