In the enterprise server-class market, SSD and flash are causing major changes to sales of drives and storage-related gear. The performance boost that a SAN gets from adding all-flash arrays means that old practices of having many drives, but just using the outer data tracks to get performance, are no longer necessary.
The impact is that fewer drives are needed, and low-priced bulk storage drives are replacing fast drives in new arrays. At the same time, arrays with all that extra capacity are being repurposed as bulk secondary storage as users turn to the new tiering model of flash primary and bulk secondary tiers.
This change of view is reflected both in the sales of high-end hard drive, which are declining rapidly, and sales of new arrays. One result is that array prices have fallen, while another is that low-cost mid-market arrays are picking up speed as a cost-effective secondary tier vehicle.
Further pressure on the array market is coming from portable software stacks, such as Red Hat’s Inktank Ceph and Windows Storage Server, which are being applied to consumer off-the-self (COTS)-based hardware to deliver enterprise class solutions at much lower price-points than traditional Big-Iron storage providers have historically offered. The Big-Iron companies have had to react to this price pressure, and the result is that array pricing has fallen further.
There's still plenty of room for further price erosion. The wide open COTS market place allows price comparison of components, and this is leading to drive prices close to Internet retail levels, which is a long way from the $5,000 per terabyte of a few years ago.
Erosion of big iron can be expected to continue at a growing pace. The COTS-based nature of modern storage is having an impact on storage modularity. This is an area where there is rapid evolution. We see a variety of solutions, but a general trend is to compact, high-performance systems of around 12 drives that can achieve very high performance and capacity by scaling out.
We can expect similar appliance modules to appear from a number of Chinese ODMs. These will use commodity components throughout, and differentiation will be realized through software features. ODMs already deliver boxes in high volumes to the largest CSPs, such as Amazon and Google, and a cutting-out of the “OEM” middleman can be expected, with HP and others taking a hit. This will trigger a race for the bottom on prices, but it should increase margins for distributors and resellers.
Putting all the pieces together, we can see that storage prices will be under siege as the price war heats up. New vendors will become prominent — most of the top SSD/flash players had zero storage presence six years ago. Compact form factors will be common and open or near open software will be a standard part of the game. In some ways, this is the Unix/Linux revolution again, and, just as then, there will be many opportunities to participate with new products and modified distribution channels.
Likely, most customers will look to VARs or integrators to configure and test COTS whitebox units in the short-term, but we'll reach appoint where fully-loaded boxes are being shipped. Supermicro already offers this, for example.
Looking at strategic implications over a five plus year horizon, we can expect erosion of traditional vendors' storage sales, with a resulting change of emphasis on their part to software and value-added services.
NAS and object storage segments have a lower barrier to entry compared with Fibre-Channel SAN storage. NetApp can and will fight back as having powerful NAS software offerings, but will see price pressure, and may sell hardware-agnostic storage apps as a result. Other NAS suppliers will not be so successful, especially as object storage begins to erode NAS sales.
EMC is acknowledging the transition to cheap hardware. Their new ViPR storage gateway is designed to handle whitebox storage, and large hard-drive arrays have been declared old-fashioned, with a secondary storage function. The inertia of the FC world will keep sales going for quite a while, but price erosion can be expected.
The net of this is that VARS and integrators will have a “Unix versus the Mainframe” dilemma of whether to support new cheap storage solutions or favor the traditional big-iron. How quickly the transition occurs will be a function of market stability, but the long period of churning seen in the UNIX/Linux transition away from mainframes isn't going to happen, for the simple reason that the ODMs already make the product for both CSPs such as Amazon and Google, and OEMs such as IBM, HP, and so on. The ODMs are the volume producers in the industry already!