Pure Storage CEO Giancarlo: Pure Storage Platform Is Built For AI

‘As enterprises go through their normal storage refresh, they should be able to uplevel it so instead of having to build specialized environments for AI, they’re able to have their workload data, their existing data that’s sitting in existing environments, accessible for analytics and AI,’ Pure Storage CEO Charles Giancarlo tells CRN.

At its recently concluded Pure//Accelerate conference, Pure Storage did talk about storage, showing its soon-to-be-released 150-TB version of its DirectFlash Modules, which it builds its all-flash arrays around instead of industry-standard SSDs, and showed a mock-up of its upcoming 300-TB DFM.

However, for Santa Clara, Calif.-based Pure Storage, the focus of the event was on what it calls the Pure platform, which aims to tie on-premises and cloud storage into a system that allows the management of storage as if it were a cloud for all workloads including AI, all with a strong security underpinning.

The Pure platform comes from the company’s focus on using a single storage operating system, called Purity, to manage data consistently wherever it is, said Pure Storage CEO Charles Giancarlo.

Pure Storage invests 20 percent of its revenue in R&D to find ways to make storage more modern, Giancarlo told CRN.

“We’ve been working for many years on the ability to turn storage into a cloud of storage and not just arrays,” he said. “Networking has been virtualized and made universal, and compute, whether virtual machines or containers, is increasingly virtualized. But storage, at least in the enterprise, was still very bespoke, very fragmented, and tied to specific application stacks, and so not, in my parlance, not virtualized and not made like a cloud. It was high time to do it.”

By doing so, the Pure platform is API-based and operates as a cloud and as a fleet of arrays on which customers can build policies and define the system through software, Giancarlo said.

“As enterprises go through their normal storage refresh, they should be able to uplevel it so instead of having to build specialized environments for AI, they’re able to have their workload data, their existing data that’s sitting in existing environments, accessible for analytics and AI,” he said.

Pure Storage aims to lead the way in terms of how it is building AI and security into its Pure platform, turning it into more than just a storage system. Here is more from CRN’s interview with Giancarlo.

Define Pure Storage and talk about how that definition has changed in the last couple of years.

I think the definition that hasn't changed is Pure Storage still treats storage as if it’s high technology, not a commodity. That’s certainly the reason why I came to the company.

The view that all the other vendors took was storage is a commodity meant to be reaped and not invested in. Whereas we invest 20 percent [of our revenue] every year in R&D and are constantly looking to see how we can make it better and more modern. We’ve been working for many years on the ability to turn storage into a cloud of storage and not just arrays. Networking has been virtualized and made universal, and compute, whether virtual machines or containers, is increasingly virtualized. But storage, at least in the enterprise, was still very bespoke, very fragmented, and tied to specific application stacks, and so not, in my parlance, not virtualized and not made like a cloud. It was high time to do it. Also, others do not treat storage as if it needed to be integrated. ... And our view is nope. Stay focused on one operating model, one operating system, manage flash across the board, and then network all the storage. So that’s what we're doing. We’re trying to bring storage kicking and screaming into the 21st century.

Has that focus on one operating system ever been a hindrance?

Of course, in the sense that it takes longer. It certainly would have been much faster to buy a company that had disk management to get to lower price tiers. We waited until the combination of fundamental flash costs and our technology would allow us to get to a pricing level where we could satisfy lower-cost workloads. When we first came out, we were pretty much the first all-flash environment for block workloads. There are so many different use cases for storage. To invest in the same software to eventually get it to handle the majority of different use cases just takes money, time and patience. And so it’s been 10 years. I mean, I think I should admit that we’ve been in feature catch-up mode for 10 years. But we’ve done that feature catch-up not by buying other companies for the most part, but by investing in Purity. And now we’re at least close to equal if not superior to most other vendors out there, but we did that on one software environment. So that makes a big difference. But it has required discipline and investment and patience to get there.

At the Pure//Accelerate conference, the focus was on AI and security, but not so much on flash storage.

Well, ‘Coz’ [John Colgrove, Pure Storage founder and chief visionary officer] showed our not-yet-shipping 150-TB DirectFlash Module, an actual test module. He also showed a mock-up, a 3-D-printed version, of our 300-TB DirectFlash Module that we’ll produce by the end of next year. You can have a little bit of fun with flash versus disk. And the press has a lot of fun with it. But we’re so convinced that flash now will replace disk this decade that we don’t consider that news anymore. ... We’d rather have people talking about Pure Fusion, what we’re doing in AI, what we’re doing in cyber resiliency. We'd rather have the discussion around that.

On that subject, a year ago this month you said, ‘the days of hard disks are coming to an end’ and predicted that ‘there will be no new hard disks sold in five years’ as flash storage replaces hard drive storage. Do you still stand by that?

I’ll be very specific because I actually said that the last hard-drive system will be sold in five years. There will always be spares and replacements and so forth. And now, yes, we’re on that path. I have four years left.

What is the competitive environment? How has it changed in the last year?

I’d say it's probably gotten a bit more intense. It’s always been intense. But when you’re not the biggest player in the market, you have the biggest players shooting arrows at each other and at everyone else including you. We’ve gotten to be big enough now where I think most of the guns are aimed in our direction. It’s both a compliment and a challenge, of course. If anything, it has gotten a bit more intense. But I think we play by different rules. In a sense, it’s an unfair competition. Our competitors play entirely by economic rules, and we play by technology rules.

What does that mean?

We provide the best technology. We spend 20 percent [of our annual revenue] on R&D, whereas our competitors generally spend less than 5 [percent], and they just can’t keep up technologically. They’ve ceded the high ground of the best product and capabilities to us. But they’ll discount like crazy. There are all sorts of programs focused on trying to take us out of accounts that they’ve put in place. It’s not that economics don’t matter, but they don’t matter to everybody. So we’re consistently picking up market share simply because customers do want best in class.

Pure Storage is talking a lot about the idea of its technology being a storage platform. How big a change is that compared with how Pure Storage presented itself prior to this?

This speaks to what I was just talking about selling at a higher level in the account. You can’t claim to be a platform unless you can solve the majority of your customer needs in a specific area, in our case, storage. About 10 years ago, we only had one product, FlashArray, that solved one set of customer needs, that of block storage for high performance. We came out with FlashBlade, came out with file on FlashArray, and that took care of file services in both scale-out and scale-up environments. Frankly, we had to catch up on the feature set to make ourselves broadly applicable in file and object environments, but even then it was still only for high-performance environments. Flash was too expensive to offer services for, let’s say, lower-priced environments like backup or archive. And it wasn’t until a combination of our technology and the falling price point of flash chips got us to where we can now offer flash for the price of disk. And that now lets us offer our products both for low-cost as well as high-performance, and for block, file and object. And we did it the whole time on just one operating system that we call Purity. And the thing that pulls it all together in a platform is what we’re doing with Fusion, where now those arrays don’t operate as individual arrays, but operate as cloud storage.

So now you can really look at what we do as a platform. It’s consistent. It’s API-based. It operates as a cloud and as a fleet rather than as individual arrays. And customers can build policies. And applications are truly software-defined now, where customers through software can define how our systems operate and how their storage operates and can be accessed by their developers. So that’s what we mean by the platform. It’s a huge and radical shift in storage, similar to what virtual machines did to compute 15 years ago.

So what comes next? What’s missing now from that platform that you think really would help beef it up a bit?

We’re already being asked by customers for new capabilities because they see the power of what that platform can do for them. They want auto-balancing and auto-tiering. And then what’s really next is upleveling all their storage so that as they get into more AI environments, they can actually use their existing production data, the data that’s sitting on existing workloads, and have the additional performance needed to get access to it for their analytics and AI environments. As enterprises go through their normal storage refresh, they should be able to uplevel it so instead of having to build specialized environments for AI, they’re able to have their workload data, their existing data that’s sitting in existing environments, accessible for analytics and AI.

How far away from that is Pure Storage?

Basically, it’s just our ability to continue to scale and service customers over the next several years. There’s no more magic that needs to be developed. At this point, it’s really about fleshing out and filling out the feature set and capabilities and getting customers to refresh from their disk environments to the Pure flash environment.

At the Cisco Live event a couple of weeks ago, Cisco unveiled a new partnership with Nvidia and Vast Data. Pure Storage has had a long, close relationship with Cisco and Nvidia with its AIRI hyperconverged infrastructure. Does Cisco’s new relationship Vast Data say anything about its Pure Storage relationship?

No, in the sense that they announced it as a new relationship. But if you looked at the [presentation] slide that was right after that, Cisco basically added Vast Data to their other relationships of which we were on that second slide. You know, I’m communicating with Cisco almost every week. So it’s still a very, very good relationship.

What keeps you up at night?

The new transition that we’re making to storage as a cloud of storage rather than individual storage arrays. It changes the level at which one needs to sell and promote in an account. Typically, storage is sold to subject matter experts. We believe we have the right to sell at the enterprise level, meaning high up in the IT environment, the CIO level, or just below CIO level. Whether it’s for our channel or for our own sellers, it’s a new motion. Typically, they sell at the use case level. And so it’s a big change. Storage has been bought and sold the same way for the last 30 years. We’re trying to uplevel that in the accounts. And that takes a lot of cultural change again, both in our own team but frankly in the channel as well. There’s a lot of training, a lot of promotion, and a lot of encouragement needed.

How do you get channel partners to think that way? Do you see them moving in that direction? Do you have to prod them?

We have to prod them, train them, incentivize them, talk to them, speak to other parts of the channel organization that deal with higher levels of need in the customer accounts. And it’s really convincing everyone that the world of storage has changed fundamentally, that they need to challenge their assumptions and get used to looking at storage differently than in the past.

You said at Pure//Accelerate that Pure Storage expects to sign a large hyperscaler to use Pure Storage technology within the year. Given that Pure Storage has traditionally been a channel-only company, how do you sell to hyperscalers while maintaining that channel model?

We consider a hyperscaler as reselling services to third parties. So in that sense, they are like a channel. But there’s really no margin in such a sale for a channel. Actually, there’s another element: There is always a channel involved in that the large hyperscalers do not actually stage and assemble their own racks for their data centers. They use a third party, and they pay that third party to do that. And that is effectively the channel for the product.

While Pure Storage has always been a channel-only company, given that the company is now reaching higher into the enterprise, are there any situations in which Pure might actually have to take a deal direct to get one of those larger customers?

We have had customers in the past that have demanded that we sell directly to them and not through our channel, and we’ve not done so. We’ve stuck to our model. There’s always a value in having a channel. Even the world’s largest companies in many cases work through the channel. They want global support. They want global installation. There’s a variety of different things that they want. So I don't really see it as being in conflict with our business model. I think, if anything, it’s the opposite. The channel is really quite fundamental to, in most cases, customer satisfaction, so we’re very happy to work with channels.