CRN Exclusive: Pure Storage CEO On Fast-Changing All-Flash Storage Business And Competing With Giants

All-Flash Star

As far as storage vendors go, Pure Storage is more of a midsize company. But when it comes to the all-flash storage business, the company sits squarely in the top five along with its much larger competitors including NetApp, Hewlett Packard Enterprise, IBM, and the leader, Dell EMC.

That Pure Storage can go toe-to-toe with its much larger competitors reflects its focus exclusively on all-flash storage, a fact CEO Scott Dietzen says is key because, unlike those competitors, it does not need to protect a legacy disk-based storage business because it never had one.

CRN recently sat down with Dietzen to discuss the company's business, future industry and company developments, and how to compete with storage industry giants.

Pure Storage is competing in a very competitive environment, especially against larger and longer-established storage vendors who have moved quickly into all-flash storage. What makes Pure Storage unique?

First of all, we're serving three markets at the same time. One market is cloud infrastructure. That's selling into Software-as-a-Service, consumer internet, and Infrastructure-as-a-Service providers. The second, new, market we're playing in is the edge, the Internet of Things, big data, analytics and deep learning. AI is one of the key themes that we see across that marketplace. And the third is traditional enterprise IT.

How do you compete in these markets?

In the first two markets, we don't see our traditional competitors. Dell EMC, HPE, NetApp, IBM, their 25-year-old infrastructure designs are simply not credible for cloud computing, not credible for AI. They don't have the business model, they don't have the ease of use, and most of all they don't have the performance. If you put flash inside a system that was designed 25 years ago for mechanical [storage], you can't exploit all the parallelism that you have in flash. ... With our partner Nvidia, we won [a contract for] one of the premier deep learning AI platforms in the industry. That application is driving tens of Gigabytes per second of bandwidth, and millions of IOPs. And it's literally a nonstarter on any storage that wasn't designed entirely for flash. You can't retrofit the ability to do AI. You can't retrofit the ability to do cloud.

How about in the enterprise market?

We do face fierce competition in the enterprise portion of our business. But we're growing, while if you look over the trends of the last two years, our competitors are overall in decline. Dell EMC shrunk 17 percent year over year. NetApp did turn in its first quarter of growth, but they've been in two years of decline. They need to maintain that growth to show that they can acquire new customers rather than just shift their installed base. People are throwing away old FAS systems and adding all-flash FAS systems in their place. That's not growth. Growth is new customer acquisition. NetApp said that, for all-flash, they were attracting a new customer a day. We had five new customers a day.

What time period?

This is sustained. I don't know what [time period] theirs is over.

In that situation, they're serving mostly a shrinking installed base. So they're share-shifting a pie slice that's shrinking to all-flash. Now, it's stopped shrinking for the most recent quarter. Now the question is, does it stay? Or does it continue to shrink? I think the challenge they face is, they're not playing in cloud, and they're not playing in AI.

You say that NetApp is growing by shifting their existing customer base to new technology. Why is transitioning that base not a positive for NetApp?

I think it does fairly count for all-flash. They're selling more all-flash storage. We always knew that was going to happen. That more and more of the market was going to turn to all-flash. But [this shows they have] a growing business inside of a shrinking one. … What happens is, the competitors' margins have been continuing to decline. As their all-flash business has grown, their margins have been shrinking. We think that's reflective of their doing their all-flash business at very poor margins. Because their designs are inefficient. So they can't make enough money to support their current business models as the majority of their business turns to all-flash. I think that's what's happened to some of NetApp's competitors, and NetApp is at risk from the same thing.

But are NetApp's flash storage margins shrinking any faster than Pure Storage's? Or are you talking about their overall storage margins including flash decreasing?

One hundred percent of our business is all-flash. Our margins are at 65 [percent or] 66 percent. Our competitors are running margins well below that. And their margins have been declining.

You're talking about your competitors' all-flash business?

We don't know. As their all-flash business has ramped, their overall margins have dropped. You have to draw what conclusions you want from that.

I could see where this means their overall margins are decreasing because their disk-based business is taking a hit.

It's certainly possible.

Pure Storage says flash storage technology will not fit into older storage technologies. But companies like HPE and NetApp counter that their storage operating systems were designed to accept new technologies like flash without disruptions. How do you respond?

First, the way they use flash as their devices get bigger, they have no path to use 5-TB or 10-TB SSDs because they perform like mechanical disk inside their designs. You have to shed all of the SCSI, all of the SAS, all of the SATA, out of the storage. That's the easy part. That's about 20 percent of the work.

Eighty percent of the work is changing the data structures and algorithms so that you can drive lots of parallel access to the same device. With NVMe, you get 64,000 parallel queues into each device. Their storage was designed for parallelism across the devices, but not within devices. And that has got to change pretty dramatically.

I think an even harder change is to fit into the cloud. Cloud demands 100 percent DevOps. … Everything has got to be completely automated, and the surface area has got to be super simple.

You mentioned NVMe. While Pure Storage may be among the first to offer NVMe in all-flash arrays, competitors can be expected to follow soon, right?

Here's the problem: Once they declare it as the future, and say, 'Yes, we're going this way,' everyone that buys their product between now and then, they don't have investment protection. Everyone who's ever bought Pure Storage gets access to NVMe without having to rebuy anything they've already purchased. We always protect our customers' investment. For HPE, for NetApp, for IBM, for Dell EMC, every customer that buys between now and this NVMe future [will] have to throw everything away and rebuy everything again. So basically, part of the reason they're dancing and saying this is still three years off is because they're scared to death that people are going to say, 'Why should I buy something I know I'm going to have to throw away because you're not moving as fast as the rest of the industry?'

You mentioned NetApp not being able to add the latest SSDs to its systems. Wasn't NetApp the first to start shipping the high-capacity 15-TB SSDs?

The problem isn't that they can't ship large-capacity SSDs. The problem is, in their system, those large SSDs perform like mechanical disks because they're still using the same SCSI and SAS. So here's a simple way to think about it. When we got started with flash, the common form factor was 256 Gigabytes, right? Between that and a 15-TB SSD is 60X. If [I'm using] a straw and drinking out of this, and I increase this [capacity] 60-fold, and have the same straw, my ability to access what's inside of that device has shrunk dramatically. That's what's happening. You may have to replace the hardware and the software architecture to be able to accommodate the parallelism.

How about future technologies?

This is just the first start. What's coming next is TLC [triple-level cell]. And then QLC [quadruple-level cell]. You can't treat flash like a mechanical disk if you want to be able to use the most cost-effective flash in your device. You've got to change the way you write it. The whole reason [our competitors] use SSD form factors is that they didn't want to rewrite their software to manage the flash directly. They wanted to be able to treat it like it was a hard drive. That's the investment they all still have to make if they want to bring their new solutions forward to the new world.

And that's what Pure has done?

I would say only Google has more data center flash expertise in the world than Pure Storage has. And you see this with our DirectFlash modules for FlashArray, and with FlashBlade we're able to do things with flash that our competitors aren't able to.

Can you give an example?

In a 4U form factor, we get 15 GBs per second sustained bandwidth. We have seen legacy solutions [where] two racks can't deliver that much bandwidth. It's the ability to leverage all of the parallel paths into the flash that lets you do these modern things like deep learning and predictive analytics with much higher performance.

Where does flash fit in terms of cloud infrastructures, both private and public?

First of all, we sell to the cloud. We arm the Software-as-a-Service and the consumer internet providers to build their own cloud platforms. And that's again something that we don't have to compete with our traditional competitors because their technology is not credible in the cloud. Something we [just announced is technology] to make it easier for multi-cloud deployments, for a customer to run some of their applications in their own data center infrastructure but then seamlessly integrate with public cloud offerings. A classic use case is, you want to be able to take data and move it up to public clouds. You want to be able to do disaster recovery. If you have a data center that goes down, you want to be able to fail over to the public cloud.

What else?

Being able to move data back and forth and even schedule automatic policies for how data moves from container to container, to the cloud and out of the cloud. Those are services we added.

I would say where we're pushing the envelope way beyond the rest of the industry, though, is we're actually doing co-processing with the public cloud. [We have demonstrated] an edge application [where] we're running deep learning on the edge and then we're accessing Google's AI services in the cloud. So we run deep learning on the edge to select particular pieces of data to then ship up to the public cloud and run on Google. Google has the best AI in the world to do deeper learning … and collaborative processing between the Google cloud and Pure on the edge.

The ability to move data seamlessly between the cloud and on-premises, isn't that what NetApp is doing with its Data Fabric?

One difference is, we don't put our own software in the public cloud. We're not trying to provide a separate sandbox inside of AWS in order to receive our data. When we move data into AWS, we want it to be capable of working in the AWS native formats, and running on all of the AWS services. What we've consistently heard from customers is, 'If you're providing me a separate sandbox inside of AWS, then I'm not even at AWS. If I'm at AWS, I want to use AWS features. If I'm at Azure, I want to use Azure features. And if I'm at Google, I want to use Google features. Don't try to redefine what those clouds mean for me.'

Our legacy storage competitors are not fully grasping that. They're trying to extend their footprint into the cloud by providing a separate container for applications. That's not what their customers want. At least what our customers want.

How has the shortage of NAND memory components impacted Pure Storage's business?

It hasn't. We've not changed pricing. Product margins have stayed the same. We've maintained great growth. Customer satisfaction is sky-high.

How have you been able to get away without changing prices?

I would say we've had to be a bit more careful in how we discount. So we've not changed list pricing, but we've been careful in discounting in order to stay in our operating sweet spot. But we've had huge advantages. We multi-source our flash. We can use consumer-grade flash. And we have a two- to five-times data reduction of anyone else, so we get two- to five-times the value per flash cell. So you can argue that a constrained flash environment helps us because our competitors are much less efficient than we are. And so it impacts their business performance way more than ours.

When we look at the last quarter, Pure Storage's year-over-year growth is what?

North of 30 percent, in the most recent quarter.

And is that something you can sustain?

At our analyst day, we said we expect to maintain 30X percent year-over-year growth for the next three years. And we'll deliver … over $1 billion in revenue this year, and over $2 billion in 2020.

At what point do you see Pure Storage being a profitable company?

We've not officially stated this, but most analysts' models have us delivering profitability in Q4. We just missed it by a point-and-a-half in Q4 of last year.

Is that a big deal for you?

We'll see. There's a class of investor that wants to see a profitable business. So, I think as we cross over to profitability, we'll bring new investors on.

Anything missing on the software side to give Pure Storage at least the same capabilities as any of your competitors' offerings?

Replication for FlashBlade is the main significant gap. And it's something that we are working on.

When can we expect to hear more about replication on FlashBlade?

I'm not going to say anything now. [We just released] snapshots, and snapshots is a pre-requisite of replication.

What's next for Pure Storage?

Helping customers get ever more value from their data. We're really excited about the evolution of predictive analytics, real-time analytics, machine learning, and being able to mix and match on-line transaction processing with real-time analytics so even if data sets are changing really quickly you can still get those insights to help you change and optimize your data. That's the whole theme of data platform. It's not just storage. We're helping you get full value from your data. Deepening our value proposition. And making it easier to manage data through the life cycle as you're ingesting data that's changing really quickly, like Internet of Things. You're able to do real-time analysis on that data and make split-second decisions on how to better optimize your business.

I know that's vague. But we have learned not to tip our hand too much to our competitors about what we see as things that are going to differentiate us next year.