Ed. note: The following question & answer story is drawn from a SiliconAngle interview with IBM Distinguished Engineer, Chief Architect for PureApplication, and newly minted IBM Fellow Jason McGee. In it he discusses the philosophy behind PureSystems, its direction, and its unifying impact inside IBM.

SiliconAngle: So first, congratulations on being named an IBM Fellow. Given that the Fellows program just celebrated its 50th anniversary, there aren’t very many of them, so it’s a real honor.

Mr. McGee: Thank you. Yes, there are only 246 Fellows including the eight new ones. It’s a pretty small group.

SiliconAngle: Maybe these questions are a little premature, but how has becoming a Fellow impacted your responsibilities? Are you becoming a senior spokesperson and direction setter inside the company?

Mr. McGee: I wouldn’t say that. One thing I’ve found at IBM is you often get a position long after you’ve been doing the job. So I was already senior spokesperson, direction setter, visionary for Pure. That’s part of the reason I got this award. As a Fellow you have a platform to speak from that’s broad & influential. So over time you get pulled into more things, and your opinion influences more decisions, and of course you have more freedom to innovate & try out new things. So I think it will have more impact in the medium-to-long term than it does today. There’ve been lots of congratulations of course , so that’s nice.

SiliconAngle: Has it had an impact on your personal life? Are the people you socialize with aware of this, & do they treat you differently?

Mr. McGee: Certainly they’re aware. Nobody’s treating me differently. I have lots of friends & colleagues who’ve seen it. My family are legacy IBMers, so my parents have a pretty good understanding of the significance of it.

SiliconAngle: So IBM PureSystems is a year old in the market. Over that year, what has its impact been?

Mr. McGee: It’s brought to together parts of IBM that are not always together. It’s brought the hardware & software sales teams together, for example, to address customer problems jointly as opposed to breaking the problem apart & working independently. I’ve got software sellers out there trying to explain hardware & hardware sellers trying to figure out how to work higher up the stack & think about what workloads will run on that hardware & decide if it makes sense to bring Pure into the picture or not. I think that’s good.

One of the powers of IBM is the breadth we have in technology. Pure is an embodiment of IBM in one solution.  You see that in the salesforce and in customers too, of course, in changes in their behavior, a real shift in thinking about the actual application workloads they will be running and less about all the piece-parts. It’s created a lot of interesting dialogue about organizational change & how this impacts what kind of skills they need, how they organize their teams, how their processes work, because it cuts across these domains that historically have been separate. So a lot of interesting changes.  I still think we’re in early days in the grand scheme of things. Enterprise IT moves quickly on some levels but many times is a pretty slow beast.

SiliconAngle: I see PureSystems in a way as a brilliant working out of the philosophy behind the AS400 — which was a tremendous product for IBM – in that it is a plug-and-play system that comes in house ready to handle a specific mission-critical compute task. Was that part of the inspiration behind PureSystems?

Mr. McGee: It certainly contributed.  I wouldn’t say that we started out saying how can we build a modern version of the AS400. But I think that’s a great analogy, especially with the work we’ve been doing with the ISV community to get Patterns around not only IBM middleware but partner packaged applications & solutions. And that’s part of what’s attracting customers. I drop this thing in & a few hours later I have all the infrastructure & software up & running. I can get my work done faster. I’ve seen interesting scenarios around remote data centers, different geographies. Big multinationals might have three big data centers, but they have a little bit of stuff in 20 other locations, & they see Pure as a way to package that, drop it into a small place, & have it easy to manage.

When we started out we were thinking more about how do we exploit vertical integration between historically separate domains as a means to simplify, reduce costs, go faster, perform better. All the domains of compute, storage, network,  middleware, & OS are trying to optimize within their departments, but they are not really exploiting across each other. That’s what started this.

SiliconAngle: PureSystems for Hadoop, of course, is the latest PureSystems configuration. Having it come in prepackaged must be very attractive. It lets companies have Big Data analysis in-house without having to install & optimize it themselves. Is that what’s happening?

Mr. McGee: That’s an aspect of it. You see that in PureApp. We have all kinds of Patterns. PureApp is a more general-purpose system [than the AS400] in the sense that it can run a variety of workloads. If you look at certain Patterns, part of what motivates people is they need that capability — they need analytics, or they need process management, or they need reporting — but they don’t have the experience in that area. A Pattern provides an easy way to get started.

But the other aspect that’s compelling is it’s not just about that initial experience. It’s about how can we exploit that environment. You take a system like Pure, where I control the hardware, I design the system, I know exactly what it’s going to sit on; not only do I make it very easy, I also can make it run better. I can tune & optimize and give you better performance out of the box than I ever could in a general system like WebSphere, for instance.

So the Hadoop system is the same – I can give you all the parts you need to run it, and I can make it scream because I control the whole stack, top to bottom. That is part of the magic of Pure, you have that level of control, so you can do things you can’t do in a general purpose environment.

SiliconAngle: Is IBM itself using PureSystems for Hadoop for instance in its Big Data cloud service, and did experience with the service provide part of the basis for design of the Hadoop configuration?

Mr. McGee: I don’t know if we are using the new systems in those environments. Certainly our experiences running those environments & with some of the software in the Hadoop space have hugely influenced the decisions we made about how we built that system. If I expand your question to PureSystems generally, absolutely we’re starting to use those systems internally in IBM. Both PureFlex, PureApp & some of the other systems like PureData for Analytics, we’re deploying & using ourselves in the corporate CIO’s office & and in some of our cloud offerings. First, that validates our technology. Second, we want the same benefits that we say our customers can get out of them.

SiliconAngle: What is the direction of PureSystems development? Given that Moore’s Law still seems to be holding, the processors that will be available for the next generation will soon have twice the compute capacity that they did when PureSystems was announced. How will that next generation use that huge jump in compute power?

Mr. McGee: Absolutely these are not static systems. They’ll evolve. We’re always looking at both what future hardware systems should look like and how we can exploit those changes through the whole stack of middleware & hardware. It’s not enough to exploit a new processor, we have to use it in a way that benefits the customer. We’re always looking at that. If you look at most applications, compute power is only part of the problem. A lot of the interesting hardware innovation is really is around storage and network, and how those influence the overall availability & performance of applications. In Pure we can look holistically at all the developments going on in the hardware and software space & how we might put them into a system & leverage them for a specific workload to some benefit.

Often new technology comes along – like RDMA – where you can get real performance benefit, but you need the right network cards, the right network switches, the right operating systems with the right patch levels, & the right middleware that exploits it. In a traditional environment some poor guy in IT has to figure all that out. Sometimes it’s so hard that they just don’t bother, so often you don’t see that technology get exploited very well except in niche scenarios.

But in a system like Pure, my engineering team can work that all out, & I can deliver the whole thing end-to-end. Now every application that sits on that box automatically exploits that functionality with no additional work from the customer. So we are looking at where we can exploit the full system design to give customers the full benefit out of a new piece of technology that they might never realize themselves because of the complexity of putting all those pieces together.

SiliconAngle: IBM just made a large public commitment to flash storage. I presume that will impact the next generation of PureSystems. Will we be seeing servers with large amounts of solid-state storage built into them?

Mr. McGee: We already have large amounts of solid state storage in the system. So in PureApp its 50-55 Tbytes of storage, 10% of that is SSD. So we have that technology in the system; we’re leveraging that automatically on behalf of the user for workloads. You will see more of that as we look at next generation designs & how we can exploit some of these technologies. In general I have way more CPU than I need. What I don’t have is I/O and storage bandwidth. So that’s really where the inhibitors are today. So flash plays a very important role going forward.

SiliconAngle: I get frustrated about systems where data has to be written first to a disk array and then fed back to flash rather than the reverse. It’s the wrong way.

Mr. McGee: Part of that is driven by costs, that the cost is so high that it’s a precious resource. Therefore you have to use it in more specialized ways. Part of that is what drives everything, that you could go direct, but you would have to make some change to the application or the architecture. So when are you going to fit that change in? So there are a lot of cases where you could get a benefit, but you’re not going to have the time to exploit it directly. So an automatic tiering system makes sense. You get some benefit, but it’s not as efficient as if you went straight.

I think as flash becomes more prevalent, as the price curve changes & you can have more of it, then those things will change, & you will be able to use it in a different way.

SiliconAngle: Is PureSystems for Hadoop the start of support for NoSQL databases in general? Will we see other NoSQL databases on PureSystems as well in the near future?

Mr. McGee: I won’t comment specifically but that space is not lost on us in IBM.

SiliconAngle: Thank you.