Exclusive Interview From Scott Dietzen, CEO of Pure Storage

Scott Dietzen, 53, CEO of Pure Storage since September 2010, was formerly president and CTO of Zimbra, now part of VMware. Zimbra was originally acquired by Yahoo!, where he served as interim SVP of communications and communities. Prior to Zimbra, he was CTO of BEA Systems, where he helped craft the technology and business strategy for WebLogic that drove BEA from $61 million in annual revenues prior to the WebLogic acquisition to over $1 billion. He came to BEA in 1998 via the acquisition of WebLogic, in Java and web application server technology. Prior to WebLogic, he was principal technologist of Transarc (acquired by IBM), developer of distributed transaction and file sharing systems. He also serves on the board of Cloudera since June 2010. He earned his Ph.D. and M.S. in Computer Science and B.S. in Applied Mathematics and Computer Science from Carnegie Mellon University. Hobbies: ski, bike and scuba dive. He has two young children. “I’m a dad and I work, that’s all I do.” Dietzen also said that total of his salary and bonus is $350,000 for the year. “Any other figure is stock-based compensation, but I’ve not been selling shares.”

StorageNewsletter.com: How do you rank your company in the worldwide market of all-flash subsystems?
Dietzen: Well, you’ve seen us ranked between number two or number three in overall market share and EMC is generally given credit for being number one.

Pure Storage’s revenue always increased but the company never was profitable and has huge losses. What’s the problem?
We’re the fastest growing storage vendor, that we’re aware of, in history. The company is following the same recipe that NetApp did in its early days. We take the dollars from growing the business, and we hire sales and support people in order to grow very quickly. The business was actually cash flow positive in the fourth quarter of last year, $30 million on $150 million in revenue. So we’re generating cash even though we are not yet profitable.

So when will Pure be profitable?
We have not specifically made a promise on when will we be profitable. All of the fastest growing companies in tech history have deferred profitability in order to drive growth because it takes money to fund growth at this pace, but we have highly differentiated products margins and we’re selling our storage at a better price than our competitors are and we’re cash flow positive. We defer more of our revenue for subscriptions than our competitors do because we’re offering higher quality support.

So your goal is more to take market share than to be profitable?
We aspire to be the number one brand in storage worldwide and we believe that the strategy we are pursuing has us on that path to be number one.

Will you never enter into hybrid arrays?
No, because it will not be a market for much longer. Flash now cost less than a 15,000rm disk drive and next year an 8TB disk gets more expensive to own than flash. The CEO of EMC storage business and the CEO of NetApp both said in their last earnings calls that all the primary data, hot data, goes on flash and not mechanical disk. So they are copying the Pure Storage recipe … well trying to.

Tell us more about your new high-end product.
FlashBlade is for unstructured data, big data workloads. It complements flash array which is very good at structured data and virtual machines, at less than $1/GB.

Coho Data said about the new Pure Storage, FlashBlade, that it is a bad idea to bet against commodity hardware and the product not being scale-out. What do you answer to that ?
The product is scale-out. We already supports 15 blades inside a chassis and each blade is like a little mini flash array. So it is scale-out as it’s defined. You could not deliver 1.6PB in 4U using traditional SSDs, and you could not deliver 15Gb/s of bandwidth using traditional SSDs. We simply couldn’t have put so much flash at such a low price into the right hardware form factor without innovating in hardware.  

Why didn’t you put your products for SPC-1 testing?
Because we do deduplication we’re disallowed from benchmarking. And we don’t turn deduplication off in our products.

For FY16, only 21% of revenue came from international markets. When do you intend to invest outside USA?
The reason for that is that our U.S. business has been growing as fast as our international business, so the 21% has held roughly constant over the last 18 months / 2 years. No one has ever in history grown a storage business as fast as Pure.

But Europe is more than 21% percent of the worldwide market.
Eventually we can’t continue to grow at two and a half times year over year in the US, so as growth will slow at some point below 100%, I think that is when international will take more. Ultimately if we’re gonna be number one. International will be the majority of Pure business but we’re extremely happy that all of our businesses are growing two and a half times year over year at a $600 million revenue runway. It’s never been achieved before.

When you sell a subsystem, what’s the percentage corresponding to acquired SSDs?
I would say it varies depending on the size of the flash drives and so on. If you compare to an all-flash VMAX or an all-flash XtremIO, they will need two and a half to three times the flash, three to five times the memory and CPU, ten times the interconnects. We generally end-up in a situation where the cost of hardware in an XtremIO or VMAX configuration, if its more than a 100TB, cost more than Pure street price.

Eventually could Pure be acquired?
It’s public knowledge now that vendors have tried to acquire Pure Storage and we said “No, thank you.” I can’t think of a company that’s been acquired when they were growing 150% Y/Y. Our public market capitalization is approaching $3 billion. At two and a half time Y/Y growth you would have to offer our shareholders a lot of money to make it interesting. Public companies don’t want to spend that much on top, and so companies that are growing as fast as Pure almost always stay independent because there is so much more for shareholders to keep the business independent. To protect the business from hostile acquisition we have a dual-class share structure: current shares, pre-IPO shares, have ten votes and when a share changes hands post-IPO it reverts to a single vote. So you’d have to buy more than 90% of the shares to drive a hostile acquisition of Pure.

You have three SSDs suppliers, Toshiba is one but who are the other ones?
Samsung is also a public supplier of flash to Pure as well as another supplier. We are working with three of the top flash suppliers.

What is your opinion on subsystems based on SSDs vs. flash chips ?
One of the critical technologies in this is NVMe. It replaces the SAS protocols that customers and vendors are using to interconnect flash to CPU because SAS is not the right protocol for flash. We used SAS for a period of time because we found way to engineer around the constraints of the protocol but we are convinced NVMe is the future. I think for a hardware form-factor like the flash array, it is going to be a fine interconnect with regular SSDs. For FlashBlade though, we simply cannot get the density or performance that we need, even with NVMe, out of SSDs and that’s why we built the new hardware FlashBlade design.

What is the percentage of your sales through channel and direct?
100% through channel, zero direct.

Your biggest reseller?
I don’t believe we’re publishing that data. Most of our VARs are regional and focus. There is some, Datalink, World Wide Techology (WWT), that have international businesses, but most play in their regions. In Europe specifically, Arrow is our largest distributor.  

EMC apparently wants to destroy your company by suing it on the ground of hiring former EMC employees and infringement of patents. What’s your reaction?
We believe in fair and open competition in the marketplace, we think it drives innovation and customer value. Our disruption ultimately will prove a good thing for EMC because I think it’s the reason they acquired XtremIO, DSSD, it’s the reason they launched an all-flash VMAX… We’ve replaced hundreds of VMAX systems around the world with our flash array and I don’t think that happened in EMC history, that somebody so successfully challenged the VMAX which has had 20 years on top of the storage industry. I think the reason we have litigation is because of that success in the marketplace.

Your main customers?
Seven of the top twenty SAS vendors, including Workday or ServiceNow, are using Pure Storage. We don’t sell to Google or Amazon but we are present in most of the other clouds. Some of our largest customers include Nielsen, LinkedIn, Workday, Barclay’s Bank …

Do you think SSD will be used for archiving?
The story from the current disk vendors is very similar to what the  tape vendors said when Data Domain came along. If you remember they said: “Disk-based backup will be important someday but you should stick with our tape technology and we’re gonna add disk to it.” Some customers believed that that was the right path, but others said: “No we should take technology that was purpose-built for disk, like Data Domain, rather than live on this 20 year old tape technology.” The same history will repeat. Disk and tape have much more in common than disk and flash do. We’re gonna see a complete change of the information technology. If you take power, cooling, swap costs, flash is cheaper next year and is dropping faster. All hot and warm data is going to go on flash. For cold data I think we will see a mix of disk, tape and optical media. The one reason to put cold on flash, is that some of the next generation backup vendors, like Rubrik and Cohesity, provide restore in place so they are very excited about using FlashBlade as a backup target because you can reconstitute and run the application right where it is with the same kind of performance as a tier 1 system.

Please a comment in one sentence on these companies also in all-flash arrays:
Dell: With Dell and EMC you have nine distinct all-flash storage offerings and I think we will see substantial consolidation in their products.
Hewlett-Packard Enterprise: The challenge for 3PAR is that it was the best disk array arguably ever built but the technology is ill-suited to flash. They have not been able to catch-up on data reduction. Their chance to get to NVMe is going to be very tough.
Hitachi Data Systems: A solid disk product, but investing very late in the shift to flash, so it has missed data reduction, copy protection, variable block architecture, NVMe … so many of the critical innovations in flash.
IBM/Texas Memory Systems: IBM is fundamentally challenged as a company. Their business is shrinking substantially in the US. Much of their flash success has been tied to their proprietary OS. We see them actually much less than we see other competitors. I think that may hurts their business longer term.
Kaminario: Very rarely see them. Small vendor that we think is growing much more slowly than we are. I think that makes their road tough.
NetApp: Very strong history in storage, but ONTAP is 24 years old. We shall see if SolidFire was a good acquisition. If you look at how many nodes are required if you’re gonna do a half petabyte of storage on SolidFire today, it’s something like 30 to 40 distinct nodes. And they still rely on mirroring, which on flash doesn’t make cost effective sense to keep multiple copies. Until they get more efficient in hardware, it will be a tougher road.
Nimble Storage: Good company, has had great success down market and in the small business. I think the challenge is that public cloud and hyper converged have made that market tougher and so they’re now trying to reinvent themselves as an enterprise and all flash company and I think that’s a lot of work.
Tegile Systems: See them very rarely. I think it is a challenge that they are based on ZFS which is ultimately the intellectual property of Oracle. And because Oracle is doing their own stuff internally you don’t have any of the new innovations. It is  also a very poor file system for flash. We evaluated it in our own early days and concluded that it would have to be rewritten to take advantage of flash.
Tintri: Very good VMware integration, like Tegile design for mechanical disk rather than flash. They will be challenged by VVOLS, which I think commoditize much of their innovation.
Violin Memory: See them very rarely. They’ve innovated in hardware but have traditionally come up short in software and I think that has hurt their penetration of the market.