 The Cube at IBM Impact 2014 is brought to you by headline sponsor IBM. Here are your hosts, John Furrier and Paul Gillan. Okay, welcome back everyone. This is live in Las Vegas at IBM Impact. This is the Cube, our flagship program. We go out to the events, extract the signal from the noise. I'm here with Bernie Schieffer, IBM Distinguished Engineer on the software side. Based in Canada, almost 30 years with IBM. You've seen a lot of the action. Welcome to the Cube. Thank you, nice to be here. So I really want to chat with you about, obviously power systems is all the rage and tough from a hardware perspective, from a leadership perspective, not from a software perspective, but in general the role of software because without going into the speeds and feeds of the hardware side, we're in a software defined era right now in a whole new way. And obviously you can almost, we were saying really, oh, mainframe's at the cloud. It's a cloud as the mainframe basically distributed in different parts. But in general, that's where the market's going. You're seeing cloud being a great resource. On-premise isn't going anywhere. The data center still needs to be somewhere. There'll be an on-prem always. People are on-premise. So you're seeing the new distributed software model come back at systems again. So I want to get your take on that. So like from a big picture standpoint, power is enabling new capabilities. What is the software, if you had to kind of summarize the key software model and innovation around power, that enables it besides the open source stack, but what is the software paradigm? Well, that's a big question. So let me start by the fundamentals, which is that you've seen the new power rate that was just announced is a building block, a rack-based building block that you can then grow into a larger, complete computing environment, suitable for on-premise as well as for the cloud. And now you have to make the maximum use of that capability in the software. And that comes through different dimensions. Some of them are sort of low-level processed, the number of cores and the frequency of the chip and so on. And for the most part, the software will exploit that transparently. Then you move into higher-level value propositions, things like the simultaneous multi-threading, where if you have not designed your software right, you're going to get stuck. You're not going to be able to use all those hardware threads. Now beyond that, there's sort of new differentiated capability and just before we got started here, I was reminiscing that it was almost exactly four years ago today that I started the first internal meetings between the lead technologists on the power side working with software group on what are the new differentiating capabilities of power rate and how is the software going to exploit them? Four years ago, congratulations, you graduated college and that's time span because that's like going to college. So take us through that. So four years ago, you had a vision, hey, we should put software on this in a way. Take us through quickly, what happened in those four years? What are some of the highlights of the innovation? Well, part of it gets driven bottom up in terms of what will the process technology allow you? How much can you squeeze on a chip and what's the right balance of cores and caches and memory bandwidth and IO? And all of those are important. And then it's these new features, they all consume real estate on the chip. So how much value do they add? How much do they cost? Are they going to help you in the transaction processing business? Which is still an absolutely fundamental, foundational part of the business. But as we've heard at this conference, there's shifts, things are moving in more into analytics, into more insight over these last quantities of big data. And those have different requirements. So I got to ask you the question. So we were talking to someone last night about Google and the iPhone and how when they bought Android, Larry Page bought Android, any room in those sidekick was like a different, different kind of mobile device. And they were building it out. And then Steve Jobs in 2007 showed the iPhone, basically he said to Larry, oh shoot, we better go back and redraw the software. Mainly, and that was the beginning of now, what now looks like in 2007, as Steve, the watershed moment of the whole mobile revolution, smartphones and all the great stuff. Did you guys have one of those moments where in the power group, we were like chugging along and was there a technology that hit you over the head like, okay, let's take a big step back to take a monster step forward. Was it flash, was it something else? Did you have a, was there a moment or a point that made you guys sestat or was it just smooth sailing? I mean, because flash hit the scene pretty quickly. It did. And flash is good for, I come from the database side and databases love flash. It can make up for a lot of other sins. What I would say as a watershed moment was really that next to the mainframe, the large power servers really are untouchable by any other system out there. There's no other large SMP that comes close to what exists. But what we've seen is the evolution of the cluster of smaller computer building blocks that really make a difference with respect to being able to incrementally, horizontally scale your systems. And so you can see a kind of watershed moment where the first and most visible power rate system launch is actually a two socket server that it's a small system, not a big gargantuan system, and it's optimized for Linux. And you- But there's no one-trick pony here. They actually use the power systems in concert with a lot of scale-out little boxes, right? That's exactly it. So the S822 and S824 are two scale-out boxes. They're only about yay high and they pack an incredible amount of compute capacity in them and that can connect out through PCIe Gen3 out to flash and external flash. So there wasn't anything that hit you guys over the head in terms of, okay, we got to kind of rethink it or something that just accelerated the process. Or was there? I think there were multiple things. Like I said, there used to be primarily a thought around large scale-up systems. And secondly, it was very focused on AIX and you've seen a lot of focus here on Linux, the new announcement about canonical Ubuntu. Who would've thought Ubuntu on a UNIX mainframe? So things have changed and the industry has shifted. We've seen the shift to cloud. You've seen now the first steps towards that on the power system side with Watson coming to the cloud as well. Yeah, I think the power system really highlights that this HPC high-performance computing has really become much more mainstream outside of the little niche of the compute world. That's actually more general. I don't want to say general purpose because it's not general purpose because you can mix and match, but some of the things around analytics, like storing big data, you don't have to make decisions on you can store a boatload of data. But when you want to jam on it, you're going to have to turn up some horsepower, right? Is that kind of the mindset that I oversimplify it? Well, I certainly completely agree with you that HPC, which we used to think of something that you would use to simulate nuclear explosions and weather simulations and things like that, has now broadened into analytics, into lots of new kinds of commercial applications where people are taking a flood of data and trying to make sense of it all. It's almost as if the power system is going to be the new standard in the data center because in a way it's not just one box and you solve a problem and you leave it there and you manage it. It's actually part of the operating environment now in the data center. Do you see that? It is, and it continues to have that reliability and scalability that is still valuable. So having a system that stays up is valuable even if you have lots of them and where you have some kind of fault tolerance where you can have another system takeover, there's still that moment where nothing's happening while you figure everything out. Okay, Bernie, I'm going to give you the last word of the segment. I really want you to tell the folks out there what is so important about the power systems from a software standpoint and why at this point in time in the industry is it something to take note of? Well, first of all, it is a seamless transition from what's come before. You can move from earlier power systems, no must, no fuss, you don't have to do major new replatforming of your applications, of skills and so on. And secondly, you're opening it up through a whole range of new capabilities and through the open power consortium, you now have connections into NVIDIA and all kinds of new accelerators that were previously unthinkable in the world of an open systems mainframe computer. Final question for you. Four years ago when you had that initial, probably email, you blast it out to someone else called a meeting, your expectations were probably, hey, let's explore this. Are you surprised? Share with us your personal feeling of where it is now and are you blown away? Did it meet your expectations? Exceed your expectations? Is it completely radically different than you thought? Give us some personal insight there. Well, it's always interesting how it's the little things that make the journey difficult. And one of the things that's been, I'd say a change for me in my many years within software group is IBM is a large company and I've had other colleagues and competitors comment on the fact that IBM hardware and IBM software are almost like separate companies. And to me, the journey for Power 8 has been a major new step in terms of a unified collaboration of hardware and software coming together to deliver Power 8 software optimized for the new Power 8 hardware. Bernie, it's been a pleasure to have you. I wish we had more time. You're what we call a tech athlete, someone who's been a great, just like a pro athlete in tech. He's been almost a 30 year veteran, distinguished engineer. Great to hear about your success and we love this environment and thanks for enabling it. So this is theCUBE. We'll be right back after this short break here in live in Las Vegas for IBM Impact.