Pure Storage CEO assesses NVMe, flash storage market

SAN FRANCISCO -- All-flash storage pioneer Pure Storage holds its Accelerate user conference this week to shine a spotlight on the flash storage market it helped create.

Although Pure isn't expected to introduce new hardware, it will showcase software upgrades designed to diversify the use cases for its flagship FlashArray and high-performance FlashBlade, which is made with proprietary nonvolatile memory express (NVMe) flash modules.

Analyst firm IDC ranked Pure Storage fourth in 2016 revenue among all-flash array vendors, trailing Dell EMC, NetApp and Hewlett Packard Enterprise. SearchStorage sat down with Pure CEO Scott Dietzen ahead of Pure Accelerate 2017 for his take on where the flash storage market is headed and his company's quest to crack the $1 billion revenue mark in 2018.

Pure Storage recently launched FlashBlade arrays that integrate custom PCI Express-based NVMe flash modules. There is speculation that NVMe flash gradually could supplant traditional SATA and SAS SSDs. What's your view on the evolutionary cycle of NVMe flash?

Scott Dietzen: We believe 80% of NVMe storage is based in software. You need to design algorithms and data structures to support the massive parallelism that's afforded by NVMe. Using just one of the 64,000 parallel queues opens up a very wide aperture to pack highly dense flash in small space.

Scott Dietzen

With NVMe, you're extending the storage device across the network all the way to the compute. With a very fast network and NVMe over Fabrics, you have the case of storage on shared storage devices moving about as far away from the CPU as storage on a local chassis. It blows up the notion of what it means to have compute and storage inside the same device.

Is talk of an all-flash data center all hype, or do you see a realistic time frame for it going mainstream?

Dietzen: We have some customers that are already [headed to all-flash data centers], including some notable large companies. Facebook is building data centers where the only mechanical devices are fans for cooling. Our view is that flash will continue to move broadly across the data center and, over time, compete on total cost of ownership with disk and even tape.

The big push in the flash storage market is to predictive analytics and deep-learning-type applications. Data-driven apps need much higher performance density than you get from electromechanical media. They are built to run on solid state.

Pure has integrated third-party secondary storage and backup providers in FlashArray. Is flash-based backup no longer a crazy idea?

Dietzen: We are certainly starting to see that, although it's still in the early days. We have customers that use FlashBlade as a backup and archiving target, for several reasons. One, they want to be able to restore in place and actually run the data set where it lives to get reasonable performance. You can't restore in place with electromechanical disk, unless you can accept a massively degraded performance.

One other use case is that data sets are changing so rapidly that disk-based backup can't keep up. The deltas are coming too quickly for anything other than flash to handle the change rates.

A third driver is that flash prices are falling, approaching the price of slow disk. Once you factor in data reduction, density, and power and cooling, we believe flash will rival slow disk over the next couple of years in terms of total cost of ownership.

What are the typical workloads customers place on Pure Storage FlashArray and FlashBlade storage?

Dietzen: For the most part, structured data workloads are targeted for FlashArray. That includes various software stacks -- Oracle, SAP HANA, Microsoft SQL Server and a lot of VMware workloads. Open source structured databases like My SQL and Postgres [also] will most often fit on FlashArray.

We target FlashBlade for unstructured and semi-structured bigger data sets. That can be things like software development, gaming, moviemaking, video automation, video capture.

You claim high repurchase rates for FlashArray. What are you projecting for FlashBlade?

Dietzen: We have enjoyed phenomenal success with FlashArray repurchase rates. The average Pure customer will triple their purchase of Pure Storage within two years of buying their initial footprint. That is largely a FlashArray metric, but we expect to see repurchase rates at least [equal] for FlashBlade. Across our top cohorts, like cloud providers and Fortune 500 companies, it's more like $12 for every $1 during the first 18 months.

We have some customers that are already [headed to all-flash data centers], including notable large companies. Facebook is building data centers where the only mechanical devices are fans for cooling.

What impact is the Dell-EMC merger -- and, to a lesser extent, the Hewlett Packard Enterprise-Nimble Storage merger -- exerting on the flash storage market? Has Pure Storage peeled off any of those vendors' all-flash customers?

Dietzen: There is no question that we benefit somewhat from organizational confusion. The combined Dell EMC has seven different all-flash storage offerings. Customers aren't always clear which one to use for which applications. They aren't sure which platforms will live on and which might be mothballed.

In the case of Dell EMC, we've also been able to capitalize on a changing relationship with Cisco. Cisco and EMC had a partnership prior to the Dell acquisition, but Cisco doesn't want non-Cisco gear going into EMC storage refreshes. As an EMC competitor, I think that [shakeout] has brought us a bunch of new business.

Has going public helped Pure Storage grow in the flash storage market?

Dietzen: A key reason we went public is that most of our customers are public companies. It put us in a position to showcase our growth, with full accounting scrutiny. We can transparently show customers that we have more than a half-billion dollars in the bank. We will sustain positive cash flow later this year. [Going public] removes a significant amount of risk for customers around our business. We're not a startup anymore.

Do you expect to hit your $1 billion revenue goal for this year?

Dietzen: We've guided Wall Street to a bracketed range from $975 million to $1.025 billion. But anything short of $1 billion will be deemed not to be successful.