ESG's Mark Peters and Scott Sinclair discuss their impressions from the Flash Memory Summit 2019, held in Santa Clara CA.
Read the related ESG Blog: What’s Next and What’s Missing in the Future of Flash
Accouncer: The following is an ESG on location video.
Mark: So Scott and I are here at Flash Memory Summit 2019. I think I've come to all but one of these. Scott, you've done a fair few these days. And we just thought we'd give you some of our key thoughts.
Over the years, it's been fascinating to watch this event grow from really just a few guys, and it was guys in Birkenstocks and sort of, you know, flat tables and sketching out what they could do. And obviously, as you can see around us now, it's turned into a proper trade show. Still more focused on vendors than on end-users. But, Scott, let's start off by talking, not so much about the granularity of each individual improvement in capacity or performance, but where do you see flash these days, both at this event and in the market? Where's its focus?
Scott: Well, you know it's funny you mention that. Because when you think about flash in general, I mean, it almost is a storage market nowadays. I mean, we still have disk, and disk is still a sizable percentage. But when, you know, the systems' vendors all talk about storage, it's all flash nowadays.
So, you know, we see flash permeating the market. It's not just a segment. And we see new trends. So we see things like NVMe and NVMe over Fabrics poised to transition flash. Within the greater market, it's all flash dominated. What's interesting about this show is we see that, you know, how we can use flash in different and new ways, and maybe tune flash to specific workloads, or do things within memory that are really on that cutting edge. And that's really what this show has become.
Mark: Right. I want you to go a little more into that because I've often thought that this show would expand from being those, you know, techie component people. Then it became somewhat of a system show for a time. I feel it's more technical and component-y again, but where specifically do you see the focus on the use of flash? That's what I'm getting at.
Scott: Well, I think what we're seeing here is really for this emerging crop of workloads that are just so, just performance and latency-sensitive.
Mark: Like, what, though?
Scott: Like, artificial intelligence, machine learning, perfect examples. Higher-end analytics. I see so many of the innovations around here talking about...okay, thinking about GPUs, and how do we use that? How do we tune flash or SSDs to specific solutions? How can you get the maximum amount of efficiency out of flash? And I think these are the elements where, you know, it may be still a little cutting edge for the enterprise, but there are specifics offers of service vendors, or hyperscalers, or enterprises that are on that cutting edge, maybe IoT, looking for, "Okay, how can I get a specific high-performance infrastructure to my solution?" And these are the innovators that are trying to fill that void.
Mark: It's interesting to me because those things that you mentioned, machine learning, AI, all that sort of thing, very glibly stated, is exactly what all the systems' vendors are talking about. But they really can't do it without what's going on here.
Scott: Yeah. And what's interesting is, you know, one of the things that we are seeing is we're seeing all the innovation for the new components, right, that tailor to these workloads. But I think what's still missing is a level of intelligence within the software layer to understand the workloads. Because here, innovators are building all the tools to accelerate the infrastructure, but organizations need the intelligence to understand which tools to use, and what scenario to get the maximum benefit.
Mark: I said we weren't really going to talk about specific, you know, gradual evolutions of performance or capacity, because at the end of the day, those are only interesting when there's some sort of step change. And I'd love to tell you some of the things we've heard out, privately in the hall that will be coming. But, you know, I don't want to just talk about gradual evolutions until they're announced, but, anything else that struck you as particularly interesting?
Scott: Well, one of the things that we're still waiting for is the persistent memory revolution. You know, is Intel Optane going to deliver that? I think there's a lot of hope for that technology. Persistent memory I think is something that we're starting to see pop up here. We've seen instances of it before. It's not new. But is it going to reach the, you know, the mainstream?
Mark: Okay, well, then it's funny. I don't know if this is a fair question to ask, but I've always been struck, you just mentioned persistent memory, I've always been struck by the fact that this is FMS, as in the flash memory. Why wasn't it flash storage?
Scott: You know, it's funny. You mentioned, when you were talking about the, you know, the early days, right, where it's just a few guys in Birkenstocks at tables. And I think back then the idea was, "How do we take flash memory and use it for storage?" And now with persistent memory and Optane, people are thinking, "Okay, how do we take flash storage and use it for memory?"
Mark: Oh, now I feel like the end of time here. On that cliffhanger, we say thank you very much for watching.