By Charles King May 28, 2014
By virtually any measure, it has been a busy spring for IBM’s Systems and Technology Group (STG). On April 7, the company celebrated the 50th anniversary of its flagship mainframe (System z) platform. Then a week or so later, IBM launched POWER8, which is the latest generation of the microprocessors underlying its market-leading Power Systems solutions.
But that same event served as a launching point for the technologies for the OpenPower Foundation, an organization supported by IBM and four initial partners – Google, Mellanox, NVIDIA and Tyan – to develop open source frameworks and methodologies around the POWER microprocessor architecture. Along with publicly revealing products and initiatives developed by Foundation members, the anniversary event also announced that twenty more companies have joined the OpenPower effort.
Tom Rosamilia, SVP of IBM STG and IBM Integrated Supply Chain, has deep insight into these and related issues backed by decades of Big Blue experience. Rosamilia joined IBM in 1983 after graduating from Cornell with majors in computer science and economics. After working first as a software developer for the MVS operating system, Rosamilia held a variety of positions in the S/390 (mainframe) organization, becoming VP of S/390 Software Development in 1998.
In the following years, Rosamilia was promoted to other leadership roles in IBM’s Software Group, including VP of WebSphere, VP of Development for Data Management, GM of IBM’s Silicon Valley Laboratory and, eventually, GM of IBM’s WebSphere software division. In 2009, he joined IBM’s Systems & Technology Group as GM of System z and, soon after that, Power Systems. Rosamilia then became VP of Corporate Strategy and GM of Enterprise Initiatives where he was responsible for IBM’s strategic direction, as well as for developing the company’s path forward in the new era of computing.
As SVP of IBM STG and IBM Integrated Supply Chain, Rosamilia has global responsibility for all aspects of the company’s semiconductor, servers, storage, and system software businesses, all of IBM’s supply chain and the company’s Global Business Partners organization. We recently had a chance to sit down with Tom Rosamilia to discuss IBM’s signature System z and Power Systems efforts, the Open Power Foundation, the company’s current market position and IBM’s path forward in the new era of cloud, social and mobile computing.
Pund-IT: If you look at the computing industry before IBM’s S/360 mainframe came along and how each system was essentially custom made, the appearance of the S/360 was similar to the arrival of the Ford Model T. You went from a market where hundreds of manufacturers all made what were essentially custom cars to a company that figured out how to place the benefits of vehicles into the hands and garages of pretty much everybody.
Rosamilia: That’s a good analogy. I keep telling people that the mainframe isn’t 50 years old – it’s just been around for 50 years. One of the things that was critical was that unique promise of value; that you could upgrade the system and not have to change all the software as a result. We’ve come an incredibly long way in terms of standardizing IT compatibility.
Pund-IT: So how do IBM’s next gen POWER8 processors and the OpenPower Foundation fit into that bigger picture?
Rosamilia: The sheer performance of POWER8 gives us some terrific performance benefits over the competition. We see 2X to 4X improvements out of the box depending on the workload. When you get into what we can do in big data, the improvements are even greater. Then when you look at what you can do with Flash-attached systems, you start to get into the 10X performance boosts. This obviously impacts IBM AIX (IBM’s Unix) and IBMi or AS400. In fact, we’re delivering major benefits with a new version of the IBMi OS that takes advantage of POWER8 technologies.
Pund-IT: How does this all flow into OpenPower?
Rosamilia: In fact, the Foundation reflects a complete change in our prior strategy of, “Unix, Unix, Unix!” to fully going after the Linux marketplace—which we dabbled in before but have never really been serious about in Power. But the biggest change in direction has been to pursue this in an open fashion from a hardware perspective, not just in software. This is also the first time that any vendor has opened its server grade hardware and said, “Let’s do this as a licensing offer.”
Pund-IT: Looking back, IBM introduced Chiphopper in 2005 to port x86-based Linux applications over to Power. And you followed that up the year after with Power.org, which focused mainly on the embedded market.
Rosamilia: Right. But none of those efforts involved the opening of intellectual property (IP) in the way that OpenPower has.
Pund-IT: So How is OpenPower going to help IBM materially?
Rosamilia: That’s an important question. The Unix marketplace may vary quarter to quarter, but it’s under pressure. Over time, we’ve increasingly taken share from competition to the point not of saturation but to where success becomes a matter of increasing share in a decreasing marketplace. Industry analyst firms say the Unix marketplace is declining about eight percent per year, while Linux is growing. So we obviously wanted to take part in a growing marketplace. Now, we could have just said, “Okay, let’s just do Linux. We’ll just do it on POWER8 and go big. We’ll do the porting centers and do lots of other things.”
Pund-IT: But that wasn’t enough.
Rosamilia: No. The decision to actually go with the OpenPower ecosystem was an even more profound change than the option of Linux, in that it isn’t about going it alone.
Pund-IT: It’s about collaboration.
Rosamilia: Yes. The same principles we’ve applied in software, where I’ve spent a lot of my career, and around collaborative open standard platforms like Apache and Linux and Eclipse and Open Stack, where we’ve been major contributors, all apply here. Then we are able to build on top. We looked at all this and realized it is exactly the model we want for hardware. So how do we do this?
Pund-IT: Were there other examples you considered?
Rosamilia: Not accidentally we’d seen what ARM was able to do, more in the mobile space than anywhere else, obviously. But by being able to enlist a set of partners with other expertise, we really saw how collaborating with these companies could work. But then we pushed it a step further to allow derivative works where participants can contribute back to the IP or keep it. That was a profound decision.
Pund-IT: How is it progressing?
Rosamilia: Over the last year, in addition to forming and then figuring out how to govern the Foundation, we’ve gone from the first five to more than 25 members total. The original five provided a lot of collaboration value at the system level. Not at the chip level. We had done a lot of that and it takes longer than a year.
Pund-IT: What kind of system issues are we talking about?
Rosamilia: They really helped us understand points like what it’s going to take to connect NVIDIA’s GPUs onto Power via high speed interconnects. What kind of power or energy requirements are we going to have? How do we work with Melanox on their technologies or use RDMA to support remote access to memory with coherence? What could we do with tie-ins to make our stuff more easily manufactured by someone else by, say, putting Power on somebody else’s motherboard? Plus, obviously we wanted to know what we could do with Google to make sure that we were meeting its needs, so that it could invest in this and gain the benefits of this great innovation and collaboration.
Pund-IT: Is it progressing as you’d hoped and planned?
Rosamilia: I believe the potential is even greater now that we’ve gotten more wood behind the arrow. The people that are coming into OpenPower are not saying, “Okay, I’m in,” and then going away. They’re coming in with their pencils sharpened and saying, “Here’s what we need in order for this to be a great flash-attached combination. Here’s what we need to be a great provider of FPGAs or to ride on FPGAs. Here’s what we need to get the kind of performance advances we want.”
Pund-IT: Some might say this approach runs counter to traditional measurements of industry progress, like Moore’s Law.
Rosamilia: We believe this is the only way for the industry to stay on track with Moore’s Law. It’s that profound, since we’re not seeing enough innovation at the chip level to drive those kinds of returns. It isn’t enough to say we’ll just keep doing what we’ve been doing—lather-rinse-repeat, lather-rinse-repeat. We’re going to get outstanding performance with OpenPower, and it’s really happening at the system level, which was also the other big driver. To do this broadly at the system level, IBM doesn’t have all of the necessary pieces. We have some of them, but being able to incorporate the other providers really gave us a whole new system level of thinking around optimization.
Pund-IT: As you were preparing the strategy around OpenPower, how were those first partners chosen?
Rosamilia: I wish I could say that we had this great recipe and put together the mix accordingly, but it was really more based on a few of us having conversations with a few different people. Like, hey, maybe we could do something about this or that. Then all of a sudden it started to come together.
Pund-IT: How critical were the original five as catalysts?
Rosamilia: I think they were the ones who were ready and willing to commit. They said, “We see this. We understand that this is the way forward and we’re willing to invest.” Obviously it’s a time commitment for them, an intellectual and sweat equity commitment. Some financial commitment is required but it’s really more about their brain power and what they’re willing to bring to the table.
Pund-IT: How many people outside the process understood what IBM was doing?
Rosamilia: I think soon after we made the announcement, some people noticed. Some people got how profound it was back in early August. But most just shrugged and said, “That’s another foundation or consortium getting together. We see these things come and go.”
Pund-IT: While OpenPower represents the future you’re heading toward, you also need to ensure that you don’t leave anybody behind.
Rosamilia: Exactly, and people ask, “Well, how are you going to do that?” I think that z/OS and z/Linux are great examples. We made the decision to add Linux to the z platform in 2001, and we have seen good traction. Many people seemed worried that customers would move off of z/OS onto z/Linux, but what I saw was a lot of people attaching z/OS with lots of little applications and tools. I realized that if we could provide a way to do that via virtualization on a much bigger system, then we could enable customers to capture huge economies of scale. And so it wasn’t a matter of gathering up any random Linux workload. It was about supporting and consolidating stuff that was already connecting to the mainframe. In any case, we haven’t done anything to diminish the value of our flagship mainframe operating system.
Pund-IT: On another topic, IBM STG has changed pretty significantly over the past decade. There was the integration with the company’s Software Group (SWG) but there’s also been an ongoing divestiture of technologies and products that IBM considers commodities, like PCs, hard disk drives and, most recently, the System x (Intel-based) server division. From a broader perspective, why has IBM done that to what used to be its signature hardware business, and what is STG going to look like in the future?
Rosamilia: I get that question a lot and I think that it’s really up to us to define that as we go along. You know, the headlights in this (IT) industry are nowhere near as far-projecting as they used to be. But if I think about the kind of change that we’re going through, the divestiture of our System x business is really based on our looking for an innovator’s margin. If we’re going to pour innovation, in the form of IBM Fellows and distinguished engineers and IP and patentable material into something, we want to make certain that it’s providing differentiated, appreciable value. What we see happening in the x86 server space is that it’s going toward a volume model, and, quite honestly, we’re a much better value provider than we are a volume provider. So you have to know both your strengths your weaknesses.
Pund-IT: That being the case, the deal with Lenovo should also give System x a fresh start.
Rosamilia: I think this move is good for Lenovo, good for IBM and good for our clients. Quite frankly, it’s also good for our employees. About 7,000+ of them will be re-badged at Lenovo and they’re going from being a low-margin business to being a high-margin business. So it’s a much better place for them to be.
Pund-IT: Are any other STG businesses undergoing significant change?
Rosamilia: It’s really about moving up the food chain to find where we can get that higher value, where can we deliver innovation, and remaking our storage portfolio is another area to keep an eye on. We’ve already delivered some things around flash technology that we see as disruptive. We think storage virtualization and software-defined storage will profoundly change the storage industry.
Pund-IT: Analytics obviously continues to be a very large area for STG.
Rosamilia: The thing about analytics that a lot of people are finally realizing is you really want your data and your compute co-residing in the same place. The alternative is shipping data that’s in one place – sometimes that’s my mission-critical enterprise data – somewhere else to perform cross analytics. And many clients would prefer not to do that for security reasons and duplication reasons, and because it takes a lot of processor time and bandwidth to move the data. By housing all data on a common platform, I can keep mission-critical information where I want it to be and augment it with the social and mobile feeds that will provide even more insights. This doesn’t represent a change in mainframe but just augments what we’ve been doing. That’s how I see us going forward.
Pund-IT: Related to analytics, I wanted to ask you about IBM’s broader CAMS (cloud, analytics, mobile, social) strategy around new and emerging workloads, and STG’s place in those efforts.
Rosamilia: In a way, we get the best of both worlds. I would argue that we’ve been providing what we now call private cloud capabilities on the mainframe for a long time, and we’ve been doing analytics for a long time. So while it’s new for many people, it’s part of what computer science has been working on for thirty years. Plus, IBM is steeped in analytics, and we’re not just sticking with our knitting on structured data. We’re making a major investment in HADOOP and combining the best of what we’ve done in the past with what we’re doing now.
Pund-IT: How about mobility and social?
Rosamilia: In that space, we’re in the business of securing the device, connecting the device, provisioning the device, user-managing the device and updating the device. All of those processes are really important parts of people’s mobile experience, but they’re absolutely critical to mobile service providers and their partners. We now see financial services companies where more than half their transactions originate not from an ATM or from a teller in a bank, but from a mobile device. In emerging geographies it’s even greater, because they don’t have and may never have branches. They’re skipping a generation of technology. There are parts of IBM that are innovating in social and collaboration capabilities, but for me and for STG, it’s more about making sure we are providing a good home for that data within the server and storage areas, and that we can effectively capture, analyze and manage that data. There’s now an extra S in CAMS for security because it pervades all of these things—the stakes have gone up in every way.
Pund-IT: Excellent. Thanks for providing such a great look into IBM and STG.
Rosamilia: Fantastic talking to you. It was great to catch up.
© 2014 Pund-IT, Inc. All rights reserved.