There's been a lot of churn and some new technologies in the storage drive space. Overall, capacities per drive are increasing but drive shipments are falling. Demand is migrating to SSD, and while cheaper SSDs are becoming much more reliable, their higher densities drastically reduces the drive count in an appliance. In addition, compression is shrinking raw capacity demand, with resulting pressure on drive needs and box counts.
All of these transitions have major implications for the logistics of the storage industry. They will alter the balance of power in the drive vendor base, at one level, while altering buying and selling patterns for the storage appliance and software segments.
The most profound changes are already apparent at the consumer end of the market. Desktop PC sales are tanking, as the mobile paradigm and applications based in server clouds pre-empt local applications and data storage. This is clearly driving down consumer-class HDD sales, but these drives are further challenged by a move to differentiate the remaining PCs by using faster SSDs. As a class, consumer disk drives’ days are numbered.
In the enterprise, the performance benefits of SSDs, coupled with much lower prices, are conspiring to kill off the HDD array. The market looks to settle out with a new primary storage tier of all-flash arrays, coupled with SATA drive appliances for secondary cold storage.
The SATA drives in the secondary storage are in many ways the last hurrah for HDDs. With SSDs at a price premium four times of HDD bulk drives in distribution, it’s hard to make a case for all-SSD solutions today. This will change in the 2017 timeframe as 3D stacked flash die become denser and new life extension technology makes TLC and QLC viable in the datacenter.
It is also worth noting that SSDs make data compression techniques possible, cutting raw capacity demand by as much as five times in that secondary storage.
HDDs won’t go away quickly. This is a somewhat conservative industry and change can be slower than rational thinking would explain. Still, the HDD is in big trouble and HAMR and large platter drives are unlikely to save the ship.
The implications begin at the component level, of course. Vendors strong in HDD components, but not playing much in SSDs, will need to retrench. From silicon to platter production to metal parts, business will drop at an accelerating pace. Flash foundries will see an upswing, which may drive some volatility in prices, but smartphone sales are flattening as phones top out on features, and that should free up foundry capacity.
The flash foundry business requires large capital investments, with the result that capacity is limited. HDD makers have hidden behind the theory that flash chip supply cannot meet demand as HDDs get replaced, but this is based on a static usage pattern and we are seeing mobile unit capacity needs dropping in the enterprise due to the shift to cloud-based apps. Add compression into the mix and mobile SSD demand should drop drastically on a per system basis. This should allow SSD-based notebooks to compete at HDD price-points, but it also reduces flash die needs.
The enterprise will see the demise of the RAID array over the next few years. Sales are already dropping fast, due in part to the fact that RAID arrays can’t handle many SSD drives due to bandwidth limitations in the controllers. More likely, we’ll see small appliances with around a dozen drive slots, importantly using a COTS motherboard as a controller. Data integrity is achieved by redundancy at the appliance level, rather than with individual drives in an array.
Whitebox products in this context are manufactured by quality suppliers, typically the ODMs selling to Google and AWS in huge quantities. These vendors are experts at designing COTS systems and are beginning to sell their brands in the broader U.S. market.
The technical barrier to entry for these small “whitebox” appliances is low, especially as we move to all-SSD solutions, with off-the-shelf open-source software. The implication is a fragmentation of the storage system supplier base and a price war for storage hardware. This pushes traditional storage system vendors into a software-and-services model, rather than their current hardware-centric approach to the market.
The channel should see a large surge in demand for whitebox appliances, opening up distributor sales for appliances and after-market drives, while also creating consulting and integrator/fulfillment needs.
The storage revolution doesn’t stop with just new appliances, though. The so-called hyper-converged system approach uses those small appliances as combined servers and storage, heavily virtualizing both. This software-defined infrastructure approach may eat into server sales from traditional vendors and so also make a new market for distribution and integrators.
The changes in demand patterns and technologies in storage have profound implications extending beyond the storage logistics chain into the whole of IT. Companies that deny the coming changes or who are slow to respond will find a tough road forward.