DigitalGenetics - stock.adobe.co
We are now entering the new era of flash-based storage, leaving spinning disks in the dust. But what’s next after flash?
Matt Oostveen, the Asia-Pacific (APAC) CTO of storage hotshot Pure Storage, believes flash storage still has a runway of at least two decades before new storage technologies, such as crystal- and DNA-based storage, could potentially take its place.
In an exclusive interview with Computer Weekly on the sidelines of Pure Accelerate 2019 in Austin, Oostveen unpacks key storage trends in the region, and explains why the scale-out FlashBlade array is more widely adopted in some APAC markets than others.
Why is the adoption of FlashBlade higher in some Asia-Pacific markets such as South Korea?
Matt Oostveen: There are probably dozens of reasons. First, if we look across Asia-Pacific, there are different blends of economies, with Australia being more services-centric with a strong mining sector while Singapore is a strong financial services and shipping hub.
In North Asia, South Korea is strong in high-tech manufacturing and you can say the same about Japan and Taiwan as well. These countries have a depth of talent with high levels of education at the PhD level, typically in the Stem [science, technology, engineering and mathematics] fields.
FlashBlade is an extremely high-performance scale-out storage architecture that works very well in analytical workloads and is extremely useful for artificial intelligence [AI] applications. So, its uptake tends to be higher in those three countries where there is more Stem talent applying AI and machine learning tools in industry.
What about in places like Singapore? You mentioned manufacturing, but not everyone knows that manufacturing makes up 20% of Singapore’s economy. And AI has been gaining a lot more traction in Singapore in the past few years.
Oostveen: Singapore is probably going to be the next domino because it has a high-quality education system and a depth of talent. We’re now seeing different use cases for high-end analytics as well that go beyond the construction of chipsets to, for example, managing money and the movement of funds that’s being done by machines rather than humans.
Another use case that we’re seeing is image analysis involving constellation networks of possibly tens of thousands of cameras generating massive amounts of data. This is an edge use case, so it’s not something that sits neatly within the core of a datacentre or the cloud.
So, what we’re seeing with these installations is the concept of data gravity, where applications and infrastructure are being moved closer to where the data is generated. In these types of cases, customers are looking for high levels of processing capability, as well as remote management capabilities.
You don’t necessarily want to have an IT team located wherever you happen to have massive amounts of data being created from outside the corporate bounds. So, the use cases that we’re seeing in Korea is very much aligned with massive amounts of data generation and the need to process data at an edge location, or to pre-process and curate data before it’s sent back to the cloud or datacentre.
As the CTO for the region, what do you typically work on? Is it more on advising customers and partners on implementation scenarios and what is the best way to implement your technology?
Oostveen: If I were to summarise it in three words, it’s to listen, talk and think. The first thing I do is to listen to what our customers are doing out here in the region, because it is often different from what we do in Silicon Valley.
As an organisation, we are conscious of the fact that customers are unique around the world. When I walk through the casinos in Macau, the way that they view technology, combined with local constraints placed on them by policy and governance, is very different from the way a mining firm might operate in the remote Australian outback. I’m constantly trying to help our customers with their unique deployments, with the office of the CTO giving an international perspective on the best and worst practices.
That’s important because customers compete not just within Australia, Singapore or Japan, but on the global stage. The subtle differences they make inside their organisations can fundamentally change their competitiveness on a global stage. It’s not just about setting metrics inside a company for how fast you can deploy an application.
Today, an organisation’s ability to innovate is predicated on its ability to deliver the technology that supports innovation. So, I’m very much focused on helping customers be more innovative and differentiate themselves through better use of data.
Is there a feedback loop where your discussions with customers are being considered in Pure Storage’s overall product strategy and roadmap?
Oostveen: I visit customers to share what we’re doing in the future. This might be technology that we won’t see for at least another year or more. I talk to some of our most important customers about what that technology looks like and whether it’s going to solve their challenges, or whether they think we're completely wrong.
We allow our customers to have very early intervention in the way we develop our product strategy. Now this is the most important part of the job: once I have the information collated, it goes directly to corporate engineering and the office of the CTO. Then, we unpack all the feedback and invite some customers to expand on the feedback they’ve given and what that means. That’s the feedback loop.
Do you see any limits to current storage technologies that we're using today? What’s your thinking around on what researchers are working on, such as nano crystals?
Oostveen: Let’s look at the era that we were in before we gaze into the crystal ball. Our first media was probably punch cards, though some people might say it’s carvings on stone. But let’s say it’s the punch card, before we moved to magnetic tape.
And then we went to the spinning disk era which we think has ended, though some might say it’s still going. But it’s fair to say that we are now in this new modern era of flash-based storage. The question is, how long is this era going to last?
My view is that we have still 20 to 25 years of this era, where the technology will continue to improve. We’ve only just begun – flash storage will get faster, it will become more power efficient, and we will be able to store even more information on it. That curve is going to continue for quite some time.
Now what you’re talking about, which is the fun stuff, is what’s beyond flash? You’re right that crystal technology is quite interesting. Holographic storage is the other one I’m looking at. The third one is DNA.
Let’s compare crystals to DNA. There’s technology that lets you store very small packets of information on DNA, but it’s about 170 million times slower to write to DNA than it is to write something on flash storage. So, we’re talking about a technology that really needs to be to be worked on. And we're not sure if we'll ever get to a point where it will be useful for us.
We’re also challenged by our ability to read data from this technology. If you were to have your bitcoin information stored on a DNA liquid that you could wear on your neck – and it’s important, maybe during the apocalypse – who’s going to read that? You’ll probably need to take it to a laboratory, get it read and the response time could be a day or two before you get your information.
Matt Oostveen, Pure Storage
So, there are lots of challenges that await us on the DNA front. While it’s fascinating and I do love the long-term nature of DNA storage, I’m not quite clear on what it would look like.
As for crystals, they are a type of holographic storage that’s also long lasting. Tape and other physical media will degenerate, but crystals can last for potentially hundreds of thousands of years. This will allow us to perpetually archive a lot of important information for humanity.
But the technology that we need to read crystals is very physical. We need a laser that must be pointed at a specific angle and at a certain pulse rate. And we need a particular type of laser beam to be able to read a piece of information.
That we need to have a physically moving device that’s trying to read information from a solid-state crystal is a challenge we need to overcome. I’m not saying it’s impossible, but we’ve just moved to the flash storage era, away from the moving parts and complexity of spinning disks. So, with holographic storage, I can’t help but feel that we’re going back to that era.
But if you were to ask which one is more plausible? The technology is quite real in the world of hologram, but there’s also a lot of promise in DNA. If we expand the concept of DNA storage, that it’s essentially a chemical storage medium, we could use a different variant of the technology – rather than the double helix structure of DNA – to store information.
How much of Pure Storage’s R&D spending goes into these next-generation storage technologies versus advancing current technologies? Are you working with research institutes around the world?
Oostveen: There’s a lot of work involved in selecting the right type of NAND. We also work closely with large manufacturers to ensure they understand where we're heading and what our roadmap looks like.
And what our customers want is fundamental. Releasing new products requires large amounts of testing, so much of the investment we make in R&D goes towards selecting the right path forward from a flash storage point of view. We are the absolute market leaders at that, and we want to maintain our leadership position.
When it comes to keeping an eye on the future, we have a CTO office that loves to look at these sorts of things. We’re also part of industry groups, so we don’t do this in isolation. There are storage associations that we meet every year. We have staff members participating in forums where they interact with PhDs from universities, whether that be MIT or some of the other leading thinkers in the field of storage.
Read more about storage in APAC
- The data deluge and compliance requirements are shaping how Asia-Pacific firms are approaching storage management issues in the age of cloud.
- NVMe storage is becoming popular among Asia-Pacific enterprises that want to reduce latency and speed up application performance.
- The Australian arm of global engineering firm Laing O’Rourke signs up for Nutanix to run its core applications on a private cloud.
- With storage systems becoming more heterogeneous and siloed, some organisations in Asean are turning to storage analytics to reduce costs and address storage issues before they occur.