Getty Images
Pure Storage CEO talks FlashArray//XL, Fusion and QLC
In this Q&A, Pure Storage CEO Charles Giancarlo talks about pandemic adjustments, the effects of the supply chain shortage and how the company is positioning itself for 2022.
Pure Storage CEO Charles Giancarlo said that his company is going strong despite the COVID-19 pandemic's ongoing disruptions to businesses. But he isn't satisfied to let the company sit on its laurels.
Coming off a strong quarter, where he reported more than half a billion in revenue to investors, a 37% year-over-year increase, Giancarlo reflected in this Q&A on the challenges of 2021 and what he predicts will be the storage trends in 2022.
He also touched on the new Tier 0 FlashArray//XL, Pure's latest on-premises offering that rounds out its flash array portfolio, its Pure Fusion management product that works to simplify the company's product line and where he believes QLC is heading.
2021 has seen the continuation of the pandemic, a larger interruption in the supply chain and a more spread-out workforce. How has this changed the way Pure Storage does things?
Charles Giancarlo: I've never seen a supply chain like this my entire career, and I went through 2001, which was a difficult supply chain environment. But this one is quite amazing, very dynamic. Fortunately, we've been able to insulate our customers from this; we've not missed a shipment. Lead times have stayed pretty normal, for the most part. At the same time, our operations team and our engineering team have been kept busy dealing with the changes in availability of components in particular. And overall costs of the supply chain have gone up on. On average, it'll be about 10% total increase in costs to us. We've not raised prices on our customers.
Has the continuation of the pandemic changed how you're getting work done? How has it affected productivity?
Giancarlo: Like every other company, for the most part, our people are working from home. Somewhere between 15% and 20% of our staff in our main headquarters comes in every day. But in our area, there's a mask mandate, so when they come in, they have to wear masks even though everybody who comes in is vaccinated because of the county requirements.
We would say that, and I would be surprised if you didn't hear something similar for most companies, while individuals will report that their individual productivity has gone up, no question team productivity is down. And you might say, 'Well, how can that be the case?' People feel like their productivity is up if they get their work done and if they can put in the full eight hours a day.
But for team productivity, you need creativity, you need to be able to quickly figure out how to do things in a better or more efficient way. And that happens when people come together.
Where do you see Pure Storage sitting as the world moves to a more cloud-like experience?
Giancarlo: We're enabling it and were the first of the storage vendors that announced capabilities in this area. What we've provided with the combination of Portworx for next-gen, container-based apps and Pure Fusion for traditional apps is two really critical abilities.
First, it provides the ability for data center operators to construct data service classes that meet their compliance and regulatory and operational objectives, and then advertise those to their developers across their companies within business units. Developers then get access to those services solely through code and GUIs. Now it looks like a cloud to those developers. They can look through a catalog that's created by the IT department; that catalog will have performance, capacity pricing, but also things like resiliency, data sovereignty capabilities. And the developer just says, 'I want X number of terabytes of class C,' and it's delivered to them on demand. They don't have to wait for it, it doesn't have to be configured by hand, and it's provided. That class C data storage capability might be on premises, but it might be in the cloud. That's all been set up. And that's all now virtualized, as far as the developer's concerned. They don't need to know or don't know where it [lives]. The IT department has set it up by policy. What developers know is they're getting a certain service-level agreement from their IT organization.
Do you see Fusion as a way for Pure to compete with the major cloud providers or is it a way to work with them?
Giancarlo: It's not competition at all -- maybe vaguely. What [Fusion] really does is it allows the IT organization to provide … the same experience to developers [on premises] that they get on the cloud. For enterprises, in particular, they may be using all three of the [major public] cloud vendors for different things. But they'd like to have a more consistent environment that their developers operate on; they'd also like to extract that environment from developers. The developer gets an abstracted view of this, and IT worries about where it sits. That's what Fusion provides.
Is Fusion a direction you see the company going more into versus supplying on-premises hardware?
Giancarlo: Both. We provide Cloud Block Store that operates on AWS and Azure. We have something called Pure-as-a-Service where, from the customer standpoint, it's one contract. If the customer chooses to put the data on premises, then we have a solution there.
With Pure-as-a-Service, we own the equipment, we deploy it, we manage it. If the customer wants the data on premises, we put equipment on premises with Pure-as-a-Service and deliver to the customer in an SLA. If they want the data on the cloud, we put it on Cloud Block Store, or it gets placed on Cloud Blob Storage on AWS or Azure. From a customer standpoint, they're just paying for a service-level agreement; we're already there.
Customers are increasingly looking to the cloud, but some need on-premises speed, control and security. Is the Pure FlashArray//XL the last piece to the puzzle for on-premises needs?
Giancarlo: Yes, basically. We hit the full range of performance capabilities with //XL on prem. There are lots of reasons why customers might be on prem: There are regulatory, performance and cost reasons. It's expensive to put data in the cloud. Most advanced customers have figured out you want your core data to stay on prem. Because the last thing you ever want to do is try to pull it out of the cloud -- you want to feed data into the cloud, you don't want to take it out -- because that's where they really charge you from an expense standpoint.
Pure FlashArray//XL is positioned as Tier 0 and Tier 1. Do you see performance needs jumping so high that may not be the case for long?
Giancarlo: High tech wouldn't be high tech if performance bars were not constantly increasing. But they're increasing along several metrics at the same time. With //XL, the performance metrics are increasing as far as latency is concerned. Every day we hear about new applications that require even lower latency. High speed trading is a great example of that. Latency is absolutely critical.
But then there are also areas of performance where it's not about the latency, it's about the throughput. Machine learning is a great example of that: You're feeding in thousands of images a second into a machine learning algorithm in order to create the proper weightings for neural networks. That's less about latency and much more about throughput.
Pure uses hot-swappable modules versus premade SSDs. FlashArray//XL has nonvolatile random-access memory (NVRAM) built in. How does that set you apart from others?
Giancarlo: We call the modules DirectFlash Modules. The new modules have NVRAM now built into them, which helps with the overall performance of the system and makes it more robust. The performance comes from the ability to scale more linearly: If you put NVRAM on the CPU and you add more storage, the non-volatile RAM doesn't increase. This makes it highly scalable with the amount of storage that customers decide to put into place.
Secondly, having the NVRAM, in order to be able to give an acknowledgement that the data has been securely written, it can't be in DRAM [as it is not persistent]. It has to be in in NVRAM. We're going to be moving in one place or another. To be able to get that done, though the extra latency of going across the backplane is not high. We believe that we get the best of both worlds by putting it right next to the storage itself.
Another technology you have been using is quad-level cell NAND (QLC), which allows for higher density. Initially, there was some hesitancy in using QLC from vendors around its perceived drawbacks. How has Pure found ways to use the technologies strengths while avoiding pitfalls?
Giancarlo: [QLC] goes into our FlashArray//C product. The FlashArray//C has been tuned specifically for QLC. While the software is the same, it runs a different algorithm for QLC that allows us to overcome the higher error rates that you get in QLC. The cost of that [algorithm] is performance.
We classified it as a separate product because it's designed to go into secondary storage environments, which are less latency-optimized and more cost-optimized. It has higher performance than disk-based systems, even hybrid systems, but it's the same, if not lower cost than these systems now for Tier 2.
The hesitancy in the enterprise around QLC was the high error rate and, more importantly, that the lifetime was so low at the time. We're the only vendor that manages the flash on our own, with our DFMs, not SSDs. Every other vendor's storage uses SSDs in place of hard disks. They don't manage their flash; their software manages the drives. This can create wear on QLC, not to mention its lower performance and higher error rate were creating quality problems, either lifetime or bit error problems for those vendors. But with us, because we modified our software specifically to use QLC, we were able to manage that in the software.
As the bugs are worked out of QLC, do you see this as an answer to the growing capacity needs or will Pure lean on its deduplication more for that?
Giancarlo: Dedupe and data compression are a very important part of making flash price competitive with disk at the system level. The reason why we've been able to penetrate a lot of the Tier 2 market with flash has been the dedupe and the data compression. At the same time, QLC is improving in performance. I could see a time when most of our products are using QLC, even the high-performance ones. As QLC improves its overall quality and performance, I think you will see that [QLC] will continue to gain.
Editor's note: This interview was edited for brevity and clarity.