Flash storage market shifting, Kaminario CEO says

The flash storage market has changed considerably since startup Kaminario released its first product in 2011. Most of the startups that designed storage arrays from the ground up for flash have been acquired or dissolved. Pure Storage and Kaminario are the main all-flash pioneers left standing.

Kaminario, the only significant privately held all-flash vendor left, secured $75 million earlier this year, raising its funding total to $218 million since the company launched in 2008. CEO Dani Golan said Kaminario's fifth funding round will fuel global expansion and product support for nonvolatile memory express (NVMe) technologies.

We recently spoke with Kaminario founder Golan about the flash storage market, from past to present, with an eye toward the future.

What's your current take on the evolution of the flash storage market?

Dani Golan

Dani Golan: When we first came to [investor] Sequoia we said, 'Guys, we believe that with what is happening with the consumption model around cloud and what is happening with flash, we shouldn't focus on a specific market like most of the technologies. We believe it's going to be everything.'

For me, the first all-flash generation really focused around performance. And then the second generation focused around specific use case. XtremIO founders came from VMware. That was all about [virtual desktop infrastructure] VDI. Pure went after, at the beginning at least, the small to midsize businesses. And we said we don't believe that's the answer. We believe you need to build the next-generation, general-purpose storage.

So, it's a very different mission that we took on. We said to Sequoia back then, 'Guys, we are going to build a more comprehensive solution on the one hand, but it will take us longer than others. And you need to be patient. This is going to take more time, more capital into the business.'

For me, the first all-flash generation really focused around performance. And then the second generation focused around specific use case.

We have been executing over the last six, seven years what we said we're going to do. And if you look at the future of storage technology, the one thing that is the next disruption -- it's the biggest disruption since flash -- is NVMe. Finally, there is a protocol worthy of flash and silicon speed. We will discover that we are really the only one that is truly ready for NVMe and, more importantly, NVMe [over] Fabrics going forward.

Why is Kaminario's architecture more suited to take advantage of NVMe than the product architectures of your competitors?

Golan: We are the only architecture that completely separates compute from storage. That allows us to grow in both dimensions, both in capacity and in compute. It's essentially scale out and scale up, and we're the only ones that can do that in adaptive block-size environments.

NVMe inherently is a much more elegant and faster protocol. But if you're putting regular controllers in front of it, essentially you're not doing anything. In our case, since you can also scale the controller, we can absolutely give the benefit of performance to NVMe.

When will your arrays support NVMe?

Golan: We are going to reveal more and more this year, but we're not at the moment giving an exact date.

How can you compete against the big vendors in the flash storage market, many of which acquired startups to bolster their all-flash arrays?

Golan: Although we are facing IBM and [Dell EMC] and HPE and all the giants day in and day out, we took a very different strategy than any of these big guys. If you look at any metric of the big guys, they are losing market share.

In the last three years, we were highly focused on a greenfield project and infrastructure to one industry. If you look at most of our customers, they are coming from software as a service [SaaS] companies and software as a service's vision in larger corporations.

So while Dell EMC and IBM and HPE have been fighting over shrinking IT budgets, we focused our energy and our go-to-market toward this fast-growing market of SaaS companies and really the public cloud. And if you look at those technologies, it is a really good fit. Out of the next-generation flash guys, we are the only one that is scale up and scale out. We paid attention tremendously to cost efficiency; and lastly, the adaptive block size in a scalable manner. All of them are a perfect fit for this market.

The cloud market is a $175 billion spend in 2019, where two-thirds of it is SaaS companies. And we're taking money not just from the infrastructure of these guys, we're taking money from the application side. Because at the end of the day, taking their crown jewels -- their money-generating applications -- and putting them in the right infrastructure for them is worth a lot. And the investors were so excited exactly on the point, the fact that we could prove to them that we're not fighting over shrinking budgets of legacy applications and legacy IT budgets. We are really, really focused on a fast-growing market and fast-growing software-as-a-service applications.

Why did your CTO leave the company, and what was your strategy to replace him?

Golan: Shachar [Fienblit] has been with us for eight years. He wanted to do new things, and so it was very amicable. He still advises the company. We stay very close. We have a tremendous bench. We've essentially groomed and promoted from within. We believe that we have the right technical depth to continue to innovate.

We recently promoted Eyal David as our new CTO. Eyal has been with the company from almost the inception. He did pretty much every role in engineering, and later on, he ran our solution management and product management and business development.

Your chief architect, Doron Tal, also recently left the company. Why did he leave, and who is filling his role?

Golan: It has been a natural progression. People come and go. Eyal Gordon, also with the company almost from the inception, did many roles in engineering and in architecture, and we promoted him recently to our chief architect.

Is Kaminario's real value in software innovation?

Golan: We're on our sixth generation, so we've been updating our platform continuously. We are completely software defined. Part of our innovation cycles and the heart of our architecture [is that] we are very detached from the hardware. Today, we're running on white-label x86 Intel servers, and we're running with a variety of SSDs. Today, we're running 3D [triple-level cell] TLC 4 TB [SSDs], and soon we'll increase that.

We're keeping current. And the reason we can do that very elegantly is because we're software defined running on commodity hardware. We have customers that have different controllers and different types of SSDs under the same Vision OS operating system, and it's seamless for them. There is no forklift upgrade. There is no controller replacement. It's all investment protection [for] what they bought; they can use [it] as long as they want.

Where does your white box hardware come from?

Golan: Super Micro. It's standard x86, two-socket, on purpose. When we choose new hardware -- whether it's controller, network, SSD -- we are very careful to always use middle-of-the-road hardware. When you choose exotic hardware -- let's say, very, very expensive CPUs -- you pay a premium that is unnecessary. So, we are always using middle-of-the-road CPUs, middle-of-the-road capabilities in networking, and we're using the most cost-efficient SSDs. We will not go to higher densities if the market is still commanding a premium for them. We want to make sure we represent our customers, and we don't charge them a premium just because we went to 32 TB disk where there is no reason to do that when they can get 6 PB in a rack of Kaminario.

Do you use Samsung SSDs?

Golan: We are using Samsung. We are using a couple of other vendors that, because of NDAs, I can't disclose. It seems that others are certainly catching up. And now Samsung has a real competition on their hands, and it's good for us. It's good for our customers.

Have you had any trouble getting SSDs due to the NAND shortage we're seeing now in the flash storage market?

Golan: We saw the shortage very, very early, and we've been discussing it for a very long time with all of our vendors. We bought options ahead of time, which secured us. It cost us a bit, but it allows us now to really benefit with stability versus the scramble right now.

Was there an accompanying price increase?

Golan: Yes, but relatively small to what we hear is happening. So, we paid a bit early versus a lot right now. It was Q2 last year we started gearing up. We didn't know for sure, but we had speculation that would be the case, and we planned ahead, bought options. They cost us a bit, but not a lot.