IBM i: A Brilliant Idea That Has Returned Again In PureSystems

June 18, 2012 Timothy Prickett Morgan

In last week’s issue, we presented the first part of an interview that Dan Burger and I did with Colin Parris, general manager of the Power Systems division at IBM, and Ian Jarman, the long-time IBM i product manager who in February was named analytics marketing manager for Power Systems and System z at IBM. In part two of the interview, we go a little deeper into synergies between the Power Systems-IBM i platform we have today and the new PureSystems platform, which inherits some of the attributes of the AS/400 and its progeny.

TPM: Let’s shift gears to the PureSystem machines. I like the design, and I think it is interesting, although I need to go through it and figure out the pricing. I think it meets a need for a lot of shops that have some Power and a lot of X86. It gives you a chance to start planning for their next upgrade cycle. No one is going to go in and rip and replace everything, but the next time they go to upgrade their Windows, Linux, AIX, and IBM i servers, they will think about having a single, integrated management capability for all of those platforms. I totally get that. I know they don’t want to use Insight Manager and OpenManager and Systems Director and VMware vCloud director and a bunch of other things to cobble it all together. There are too many tools and too many platforms.

I would say that the current PureSystems are aimed at the high-end of the midrange and the enterprise customers. The Flex System manager can span four racks and 16 enclosures, and that is a lot of iron for an SMB. Where is the smaller box? I think that a smaller PureSystem is even more important than Power7+, that you get these customers engaged. Most of the guys out in that room have one or two AS/400s–forgive me Trevor–and maybe a dozen or two dozen X86 boxes. Is there a way to engineer a machine that is suitable for them?

Colin Parris: That’s one of the things that we are working on right now. But one of the things we need to do as we roll this out is that we have to figure out is that the full experience with PureSystems will be there. Because the experience is the important thing for us.

TPM: What is this, Disney?

Colin Parris: It is. And the experience is actually the most important piece for us. This is what we are thinking through, and then the product pieces, we can adjust in certain ways and do so fairly quickly.

The other thing we have worked on with PureSystems is that the procurement process is going to be easier, and I think we’ve got that one down, and the upgrade and the entire support process, we have to make sure that these are done right. This is a journey for us, and I am not going to be the one saying that we have got it all perfect on day one. Once we address all of these issues, then we can get smaller boxes out. And then the scale changes dramatically.

TPM: There are a lot more customers who can buy this theoretical small PureSystem box than who will buy the current machines. You need a feeder system, which is one of the things that the original AS/400 had. There were deskside machines, bigger midrange machines, and then very large rack machines. I am always looking for that feeder system because it gets customers on a path, and quite frankly, the mainframe lost it. I remember 9370s and 9221s feeding into 4381s and 9121s that fed into 3080s and 9021s. When IBM ceded the midrange to Unix machines and proprietary minis, I think that was a bad day for the mainframe. It is not good for IBM to put together a good “engineered system,” to use Oracle‘s term, and then not address the needs of customers like the people who come to COMMON–and those who don’t, too.

I’ll help you design it, I have plans. . . .

Colin Parris: [Laughter]

TPM: For a lot of customers, a microserver is enough, and you know it already. A Power 720 with four cores is enough oomph for most Power Systems-IBM i customers. You can put that into a single-socket microserver board, a skinny blade or Flex node that fits into a tower chassis, and maybe you can slide in a few X86 nodes and have a switch in the back and storage in the front shared by the nodes. The nodes are hot-pluggable, it has a midplane just like the Flex System chassis, and you have integrated switching. Maybe redundant power supplies. You have the Flex Manager running on a service processor, or maybe you run it remotely on behalf of customers as a service.

Taking a blade chassis and flipping it on its side and putting it on some wheels is not the right answer, which is what IBM did with its BladeCenter-S and what Hewlett-Packard did with its BladeSystem c3000. These boxes are too big and too loud and have too much capacity and were too expensive. And what I can tell you is this: None of the other server makers know what I am talking about. Sun-Oracle is allergic to the SMB base. Dell is trying to go up against your high-end to make money, HP is doing the same thing, and everyone is worried about trying to get Cisco Systems out of the data center and selling hundreds and hundreds of machines. HP, Dell, and Fujitsu are taking their eye off the ball of their core business, which is SMB.

I don’t know. Maybe I am wrong.

Colin Parris: No, no. I like what you are saying. Initially, when you introduce something, you have to do so in a certain spot. But as I say, we have to get that experience right. And once we get that right, I think we can address other customer segments.

TPM: I agree. I don’t know exactly what the experience needs to be, but what I do know is that SMB shops are wrestling with the same multiplatform management issues and they have far fewer skills with which to do it. So, for instance, if you can’t mask the Virtual I/O Server so these SMB customers never see it, then you are dead. And you have to do it for VMware ESXi and Red Hat KVM, and you have to make it absolutely idiot proof. And the good news is, I don’t think anyone is doing that. Your Tivoli guys have some work to do.

Colin Parris: We have exactly the same issues in the growth markets, too. We have to grow out there [gesturing to the COMMON trade show hall] and in the growth markets, these are the exact same issues that you run into.

TPM: Where’s the initial interest for PureSystems coming from, and where do you expect to see it take off first?

Colin Parris: We are actually seeing interest all around the world. There’s a lot of interest in North America, but I was surprised by the interest coming out of the growth countries.

TPM: And I suspect that greenfield applications are the best place to start, just like Cisco has done with its UCS blades when customers decide to build a virtual desktop infrastructure cloud or just a regular private server cloud.

Colin Parris: PureSystems is interesting to customers for different reasons. The labor cost reductions–customers in North America love that, and what they want to talk about is the expertise and the automation so they can significantly drop their labor costs. In the growth markets, they lack the skills, so they like the automation, too. In the midmarket, customers are actually interested in consolidating servers in a way that they understand inside of one box. This notion of simplification is huge with the large customers.

So every market and every group of customers have a different perspective. And we have seen a great pipeline of customers growing, which is the thing that I like. What I really care about is the number of proof of concepts we have going out, which is a lot.

TPM: What is the role of IBM Rochester in the manufacturing of the PureSystems? Where is the Flex System iron made?

Colin Parris: It depends on what part you are looking at. We have various types of ITEs [that’s IBMese for IT Element, which is a server node or a switch blade or other element of the Flex System] and the chassis itself. There are things that we do in Shenzhen, things in Guadalajara, and things we do in a bunch of different places.

Ian Jarman: All over the map is a good expression here because we have different manufacturing places. We do have some manufacturing in Rochester, and what we do is system integration.

Colin Parris: That’s the great thing about i. So much of what went into PureSystems was taken from i, so to speak.

Dan Burger: Where are future IBM i investments being made? I think everybody wants to know something in that regard for the health of the platform. What are the key investments that you are making in the IBM i platform?

Colin Parris: I would say that there are three areas. Clearly, the cloud investment part is a big one for us, again because having the ability to reduce labor costs and to automate is a big thing. The second thing that we are looking at, some of which that we have talked about such as the Virtual I/O Server, that we clearly have to work on is this: There is a consumability part of this that we need to put some money into because there are a few things that our big clients have told us that we need to get right. The third area is not completely i-specific, but relates to the broader PureSystems. We’re doing a variety of things in that platform itself that benefits i, and what we have committees in which i weighs in.

Let me just be transparent here. One of the things that I think is just great for us is that because there is so much of a flavor of i in PureSystems, then i gets to ask for things that not only benefit the entire whole of the platform, but also i significantly.

Dan Burger: It’s often brought up that PureSystems is patterned after i. That’s clear. What is not as clear is how i will benefit.

Colin Parris: I see it differently. If I can get i mainstream, then i benefits. To me, i is a brilliant idea that we had which has returned again. You see it in what Oracle is doing, you see it in what clients need. So what I am trying to do is find ways to expand the aperture for i, so that a lot of what I was is what PureSystems becomes. I am not seeing I as this thing where I am just trying to grow a small base, but rather to expose i in a different way and that can allow it to grow in a very different pattern.

There are two ways that you can focus on growing anything. You can grow from a pinpoint position, or you can take an idea and seed it in a variety of things and then grow those.

I go back to these IBM i shops. They have eight, nine, or maybe 10 people. And they are directly tied to delivering business value. That’s what we need right now. We need IT shops so directly tied to delivering business value that it becomes obvious that you want to use this PureSystem platform. I need to reconstruct that in a huge midmarket perspective. So rather than say that i is the way to go, I say that all of the i attributes are what you are going to get in PureSystems and so now I have i everywhere, and I naturally have the actual IBM i nodes that naturally slide into the box. I am trying to create a much bigger platform to grow, because it helps me find the funds to do further investment.

Dan Burger: Do you think that PureSystems will add more workloads to the IBM i platform?

Colin Parris: Yes, I think in the end that it will.

TPM: In certainly will help preserve them at the very least.

Colin Parris: It will preserve them and then add to them in certain cases. The ISVs are not foolish. If they have applications on IBM i and they are doing well, why wouldn’t they invest in those applications? Some of the ISVs are trying to find differentiation, and it is clear that some of the bigger players have optimized for X86. So if you, as a smaller ISV, optimize for X86, there’s nothing different. But if you take the patterns of expertise and other optimizations on PureSystems, you will have advantages over competitors who are just on plain X86 machines. This is exactly what I want. And there will be other things that we will add to PureSystems that will allow ISVs to optimize even more. I want ISV applications to run best on Power, and if I can do that within the context of PureSystems, then I win because I can control every piece of that stack, whereas on X86 you have to go through VMware and Microsoft and Oracle or whatever. So my focus is getting the ISV to realize that they can be much better optimized on Power.

At the end of the day, the line of business manager doesn’t care what the applications run on. I am going to push PureSystems, and then once we are doing that, we are going to push Power, which gives another set of differentiation and optimizations. We have done studies to compare these different platforms. For example, we have done studies and on Linux on X86 servers, you can have 120 hours of downtime a year, but for AIX or IBM i on Power, it is more like 1.8 hours. No one can afford 120 hours of unplanned downtime. The other issue you have is security, and this is where it realty kicks you in the teeth. The minute you find a vulnerability and you have to patch it, you have to patch it right away. You can’t plan for that, because you can get hit at any point in time. The way it works now, the CFO goes in and buys a ton of X86 boxes, and then when they go to handle all of these security issues, they then buy a ton of software, and this just explodes the price.

But if you let the ISVs lead the discussion about unplanned downtime, automation, security, and patterns of expertise that is all built into the PureSystems platform, and this thing runs even better on these ITEs that run IBM i, it is a different discussion. And that is the discussion that we want to get to.