Kaminario Goes Software-Defined

Kaminario recently announced that the company is moving to a software-based business model. Rather than build and support storage arrays, Kaminario will partner with Tech Data for logistics and essentially focus on the software part of their offering. Is this an admission of defeat or a sign of the times?

A Hardware History

Kaminario has had a long history of hardware-focused products. Early solutions included DRAM and PCIe, before the market started to settle on the SSD format. The use of an essentially bespoke design allowed Kaminario to focus on performance as their USP. The company still holds position #6 on the Storage Performance Council SPC2 benchmark. Of course, dragster-like performance started to be less of a differentiator and like all maturing markets, features, functionality and cost became important. As an Enterprise solution, K2 keeps up with the competition in terms of features and was early to TLC NAND and the $1/GB price point. (Disclosure, I did some work for Kaminario a few years ago).

Competition

However, competition has become tougher. The major IT vendors have good enough all-flash solutions, so there’s no real market for acquisition any more. The alternative choice is to go it alone, in the style of Pure Storage, but Kaminario has never been that size of company. The storage array market is contracting. Storage appliance sales are relatively stagnant partly because the market has fragmented so much. Storage choices now cover HCI, SDS, appliances and public cloud. We should also remember that storage growth is in unstructured data, not the type of storage that supports enterprise workloads. So if you’re selling to the enterprise, your products need to be both competitive and compelling.

Software First

In a modern commodity world, software is where storage IP is located these days. Disaggregated solutions are showing that a centralised architecture has bottlenecks. If you don’t need lots of features, disaggregated NVMe offers a high degree of performance. Kaminario has already been heading this way, with K2.N, their version of a disaggregated architecture. In the K2.N architecture, storage controllers and disk enclosures are able to scale independently, all connected via Converged Ethernet. This is an NVMe over Fabrics solution that still retains the ability to deliver services, while not being hampered by a controller bottleneck.

With an architecture based on commodity networking, servers and enclosures, where’s the benefit in being a hardware vendor? Building and shipping hardware means holding inventory and operating a support organisation. Better to let someone else do that and focus on the core asset, the software.

The Architect’s View

As NVMe becomes widely available as a storage protocol, NVMe over Fabrics allows high-speed communication between servers, controllers and storage. The logical conclusion is that we will see hardware distil down to racks of enclosures and servers with high-speed networking in between. The Software Defined Data Centre will allow these hardware components to be aggregated into storage arrays, disaggregated solutions or HCI as required. In this model, a software-based approach makes sense for Kaminario, because there’s significant cost investment in keeping hardware current. The value becomes (and is) the software.

We need to look at Kaminario’s plans in a longer-term context. The company isn’t stopping selling hardware products. Instead, they are selling through Tech Data and leveraging their experience to optimise the build, delivery and support process. However, there are some warning signals here. In an interview with The Register, CEO Dani Golan indicates that the transition to software will provide the ability to get to $0.50/GB. His aim seems solely to be cheaper than the competition or AWS. This is a race to the bottom that isn’t worth winning, because now you have to sell more product just to keep afloat. It would be much better to focus on value and building in features that make shared storage more attractive, not simply cheaper. Isn’t this the point of being software-focused?

Another issue is that of pricing and margin. Traditionally, storage vendors have enjoyed big margins on their products. This provided the ability to invest in R&D for the future. Moving to a pure software model risks having to give away a lot of that margin, especially when customers are supplying their own hardware. At that point, value becomes more important than ever. In that context, solutions that focus on the benefits to the application and the business will be the ultimate winners. This could be where the next storage battle will be won.

A Hardware History

Competition

Software First

The Architect’s View

Further Reading