The commoditization of storage platform hardware, coupled with the availability of unbundled, powerful storage software is changing the who and why of the vendor base. We used to go to EMC and NetApp for storage, but the vendor base broadened as the technical issues became easier to overcome and drive and array technology began to commoditize. Buyers now have plenty of choices.
The vendor spectrum is still evolving, with today’s trend of component providers laying claims to territory belonging to their own customers. Early examples of this were Intel, one of the first producers of solid state devices (SSDs), and Seagate, who bought Xyratex and began making array cabinets for OEMs.
Then we had the rare case of an end-customer integrating along the supply chain. Google turns out to be a large producer, albeit for their own internal consumption only. They buy flash die and build their own accelerators. Only someone of Google’s scale would dare to try that, but Google is THE expert on using original device manufacturers (ODMs) to manufacture, which brings up a key issue.
The computer manufacturing chain has changed immensely. OEMs do very little manufacturing or assembly themselves. Instead, boards and boxes are made by low-cost, low-margin ODMs, mainly in Taiwan and China, then shipped to fulfillment operations in the country of delivery for final installation (usually drives and DIMMS). Most branded units are rarely handled by their putative maker.
Google learned to deal directly with these ODMs, saving a huge amount of money in the process. The other giant cloud providers, Azure and AWS, followed suit and in the process made the ODMs by far the lowest cost providers.
With almost all computer gear now standardized commercial off the shelf (COTS), we are seeing a trend to displace the middlemen, with ODM boxes offered through distribution at as much as 60 percent of traditional ODM pricing. At the same time, ODMs are entering the market directly with sales to large users, who are now capable of handling COTS gear themselves. Taken together, distribution and large-account selling is becoming an element in private cloud TCO calculations.
With the demarcations between supply chain structures becoming blurred, we are now seeing verticalisation efforts by drive makers. Why be limited to selling just drives, which are on thin margins? The answer is to look at emerging growth markets such as all-flash arrays, smart storage appliances and hyper-converged platforms.
All of these units replace the traditional RAID array, providing acceleration, object storage nodes and hosts for software-defined storage, the leading edge technical trend in storage. They all use COTS-based elements, including drives and DRAM DIMMs, with typically Intel X86 motherboards, so configuration is like using Lego (all parts fit together easily but firmly). The software needed is open-sourced (Ceph object storage, Gluster File system, etc.) or commercial, but unbundled, software (Windows Storage Server, Nutanix, Simplivity, etc.)
Currently, we have SanDisk (now part of drive market leader Western Digital) making all-flash solutions. Micron has just announced an all-flash array, too, one moreover that delivers leading-edge RDMA connectivity via the new NVMe-over-fabrics protocol.
These are very serious contenders in the all-flash market segment. Performance, especially of the Micron unit is on the high end of the market, but the vertical nature of flash foundry ownership, drive assembly and cabinets suggests a competitive edge over traditional suppliers. Time of course will tell, and the all-flash market is crowded, but the Micron designs have a good deal of flexibility and I would expect them to evolve to include versions of hyper-converged and object storage nodes in the near future. It’s just a matter of changing drive models and adding different software!
Seagate is in a more difficult situation. Xyratex built a RAID array cabinet business, but that is a declining market. At the same time, Seagate failed to jump into the flash business, with the result that it is a bit player there. A key software of systems acquisition might help this, but the candidate list is looking this, with the best deals already made.
Drive makers enjoy one huge advantage over traditional OEM vendors. The price of the drives in a storage solution make up the majority of the cost and it has been a tradition to make the required drives proprietary in some minor way, allowing high mark-up with list prices of as much as 15X the cost of the drive from the drive vendor. This is one of the “alignment” issues for Dell-EMC where Dell uses a significantly lower ratio.
Drive vendors selling complete solutions can underprice the traditional vendors by a large factor just by selling drives at distribution prices. This should disrupt the market and result in a supply-chain rethink.
Quite possibly, the traditional suppliers will cave to the pressure and sell standard drives at low markup ratios. This won’t happen overnight, but, given that we all know the products are identical inside, charging Rolls-Royce prices for a Toyota isn’t sustainable.
The chain implications of this price collapse and race-to-the-bottom are multifold. First, drive makers become much more important and having solid relationships counts, especially if supply shortfalls occur (as is currently happening). Second, margins will likely shrink, especially for high-end drives. Third, and more important, is that end-user selling will become less attractive as a way to market, with a fallback onto distribution for all but the largest customers.
As the game plays out, the A-List of storage will shift considerably. End-users and distributors both should benefit from the changes.