Skip to main content

skip to main content

developerWorks  >  Power Architecture technology  >

About: Power.org

An interview with Bill Dykas and Mark Ireland

developerWorks
Document options

Document options requiring JavaScript are not displayed


Rate this page

Help us improve this content


Level: Introductory

developerWorks Power Architecture editors, Editors, IBM

02 Dec 2004

This interview focuses on the structure and aims of Power.org, a major new standards consortium in the marketplace. Discover what this new body has to do with open standards, customizable processors, consumer devices -- and doughnuts.

IBM®, along with other leaders from the electronics industry, today announced the creation of a new organization devoted to defining open standards around the customizable Power Architecture™ processor family. The announcement was made at a Power Everywhere™ event in China (see Resources for a link to the full press release, which includes information on other announcements made at the same event today).

Standards and specs columnist Peter Seebach and the developerWorks editors spoke with William A. Dykas, manager of Power Ecosystem Development; and Mark Ireland, who manages the PowerPC® Business Line in the weeks leading up to the announcement. Bill and Mark both have been instrumental in working with the founders of Power.org, and in making Power.org a reality.

developerWorks: Can you tell us more about the announcement that is being made today?

William A. Dykas: I would say it's an extension of IBM's Power Everywhere initiative. It's one of the first instantiations of our plans to create an open environment, an open ecosystem for the Power Architecture, and all of those who are members of that ecosystem, to take advantage of the architecture.

dW: So this is a lot broader than the Apple®/IBM/Motorola® alliance?

Dykas: It is a lot broader than the Apple-IBM-Motorola [Alliance]... We're trying to create an environment, a broad ecosystem of many interested community members that are inclusive of not just IBM and Apple, in terms of owning the Power Architecture, but also inclusive of people who build systems on it; also inclusive of people that might be in market spaces as broad as -- from servers to consumer devices to embedded devices that might include industrial controllers. And as we like to promote at IBM, the Power Architecture works from embedded devices, all the way to servers, to supercomputers, to planetary rovers...

One of the primary benefits of the Power Architecture is its ability to scale, and when I mean scale, I don't just mean in terms of one, two, three, four, five, but to scale from very small applications, such as a toaster [laughs], all the way up to large supercomputers that we see today. Blue Gene®, for instance, is based upon the Power Architecture.

So what we don't want people to think about with respect to this initiative is that the Power ecosystem is solely focused on existing applications or the most popular or the most widely known ones, which you would know of as servers.

dW: So Power.org is sort of an umbrella organization for what could be a fairly large number of committees, marketing committees, technical committees -- you name it.

Dykas: Yes. This is a challenge to people's imagination. If you think about it, many standards organizations focus on a single problem that they're trying to solve, whether it be the VoiceXML organization writing an XML-like language that you can write voice application for speech recognition, for contact centers and things like that. It's a fairly highly focused initiative on a particular technical problem.

Power.org is a little bit broader because we are creating a place where people could come to work together on very broad problem sets. Or, it's analogous -- I don't know if you've ever heard of the Consumer Electronics Linux Forum? They have an organization and a structure; such that -- while it is focused on Consumer Electronics Linux -- that is the focal point -- but there are specific technical aspects that they've been working on with respect to creating a Linux® for a consumer product.

They were working on a number of problem areas all related to supporting their singular mission. They had a couple of focal areas that were really saying, "Look, while this is all going towards a Consumer Electronics Linux, we're working on these areas separately, and we'll work to combine them as time goes on" (see Resources for more on the Consumer Electronics Linux Forum).

I view this sort of the same way, except it's slightly different in that the Power Architectures can be broadly used; therefore, the work initiatives can be broadly defined. But if we were to think about horizontal and vertical aspects, the horizontal being very specific technical components that need to be worked on, some of those might be more related to the industrial market segments or might be related to servers -- or within a server architecture you could have servers everywhere from super high-level servers down to what we call high-volume servers, which will probably be equated to your low-end Intel® server. They come together and figure out whatever the problem is you're going to solve.

So there might be this intersection of horizontal technical or product sets intersected with markets -- and actually servers might be viewed as a market by themselves -- but there might be things that might relate to industrial controllers that have specific technical problems that need to be solved. So if you look at a matrix, there are specific technical and market opportunities that need to intersect, and the specific subcommittee might work on more than one aspect of that intersection. It just depends on how you prioritize.

Participating in the doughnut space

dW: This is a time of great changes in the processor industry, from processes to business models. What are some of the implications of the Power.org organization?

Dykas: Well, I think the way we look at it, from a business perspective, people's abilities to modify or build around different processor architectures today can be highly limited. There are varying degrees, but in many of the traditional approaches used by vendors, the architectural elements are very tightly controlled. Not only are they tightly controlled, the value chain, in terms of how people can build upon them, becomes highly limited.

In the approach that we're presenting here with this Power ecosystem, IBM is proposing that while IBM manages the instruction set to prevent fragmentation, we are creating a community and an approach that will enable people to build a lot of value around the Power Architecture and capture that value themselves. So if you think about a doughnut with a hole in the middle, the little doughnut in the middle is the part that IBM controls and captures value from. Whereas, the outer perimeter is a large number of other companies that are capturing or designing and coming together to create specifications that will then enable them to create designs and implementations where they could then, therefore, capture the value from it.

Doughnuts defined

In marketing, the doughnut diagram illustrates a "whole product" -- what started life as a generic product (as the doughnut hole), but has since been augmented by the thing or things needed for the customer to have a compelling reason to buy.

The concept was first formally introduced by Theodore Levitt in The Marketing Imagination (Simon & Schuster, 1983). Find the other classic doughnut definition on gurunet.com.

-eds.

It's a classic doughnut definition (see the sidebar, "Doughnuts defined" -- eds.). You've seen that, except we're trying to have a smaller doughnut in the middle and the larger circle around the outside with large white space, which represents the value, the opportunity, and, therefore, the value capture that can accrue to all those who participate in that space.

dW: It seems like you're trying to make sure all the players are talking to each other so that they may still be competing at some levels, but they aren't going to compete in ways that end up screwing things up for everybody.

Dykas: Right. Think of -- well, we see this phrase frequently, and this how you implement it. Collaborative Innovation is what we're trying to do here. There's a learning curve that goes along with this. That is, each company has to step outside and figure out how the structure can solve their problems by collaborative innovation. They have to say, "By coming together, I save money or make money by doing this."



Back to top


On AMCC and Power Everywhere

dW: The sale of some of the 4xx IP to AMCC earlier this year fueled speculation that IBM was getting out of the chip business -- that isn't the case, though, is it?

Dykas: All indications to me are that is not the case.

Mark Ireland: The AMCC deal was really the last step in completing that transformation to a licensing business model and a custom product business model through our ASIC business [Application Specific Integrated Circuit --eds.]. So we continue to sell custom products and we're pushing very, very hard on the licensing front, with AMCC being one of IBM's largest licensees for the PowerPC 4xx-series cores.

To talk about the AMCC deal, you have to look at the business models for embedded. If you're trying to expand the architectural penetration in the embedded space, the most successful path is proven to be through licensing. If you could imagine myself owning a licensing business unit, trying to at the same time compete with potential licensees with very similar standard products based on the PowerPC 4xx-series cores, that's very difficult to do.

So it was a conscious decision to move from PowerPC 4xx standard products in the embedded System on a Chip market to a licensing and custom products business model.

If you look at the commitment to the Power Architecture, the commitment remains very strong. In the System on a Chip business you need to have many different flavors of chips to penetrate the market. It could be a dozen different flavors for networking, and for consumers it could be hundreds of different chips. You just can't make all those different varieties as a single company. That's why the licensing business works, because you get maximum penetration and get other people bringing their IP in and doing chip integration to get those products to the market.

And then on the higher end, for the stand-alone MPU business, we continue to sell those microprocessors. We design those microprocessors for Apple, clearly for IBM Systems Group, and we plan to move those microprocessors into the embedded space, where appropriate.

This [Power.org announcement] is really just more validation that our licensing program is on the upswing, it's very successful and it's really demonstrating transformation from being a 4xx standard product provider for embedded solutions, [to now, after the] AMCC deal where we're actually really getting great licensing traction in the marketplace.

dW: So IBM will continue making chips, but IBM wants other people making chips, too, because that makes for a healthier market.

Ireland: Absolutely. The truth of the embedded space in the market is huge, and it's highly fractured. It involves much more IP than just the processor. So if you're going to get maximum penetration, you can't go at it alone in the embedded space. You need partnerships and other people designing chips around the processor.



Back to top


Organizational structure

dW: Are there any other existing models for open hardware and collaboration that we can compare this to?

Dykas: I think what's unique about this is that we are enabling a broader community. As I was sort of explaining before, compared to the other models, whether they be software or -- This is more closely aligned to what happens in the software world than is what's traditionally found in a hardware world. What we are trying to create here, with this approach compared to others, is to be more open than them, and to enable more companies to come together and take a greater level of ownership with the complementary aspects of the work they want to do.

dW: Okay. So it wouldn't be associated with something like ISO®, but it's going to do the same kind of thing that creates a standard.

Dykas: Right. I mean, if you think of something analogous. My analogies run more along the lines of a W3C®. While that's focused on Internet-type infrastructure components, our initiative is focused on the Power Architecture and the things that might go around to improve that for the community at large.

dW: How will that be structured? Are you going to have regular meetings for the whole organization, or will meetings only be held per-project, or per-committee?

Dykas: Well, how IBM envisions this, is similar to how other organizations work. You have a Board of Directors who are managing the initiatives of the organization, trying to promote others to come together to work on these things. There could be, for instance, a technical committee, a marketing committee, an operations committee, something that really manages the whole organization.

And then underneath some of these individual committees, for example, the most obvious is the Technical Committee. The Technical Committee would have individual subcommittees focused on, say, creating specifications for high-volume servers, specifications for digital rights management, specifications for trusted computing, specifications for nomenclature. You know, having anywhere from one to n possible subcommittees. Then there might be marketing; there might be compliance subcommittees. So they would all come together and you manage through these individual ...

Ireland: There are some initial working groups that IBM has envisioned and has actually been talking to potential founders about. For instance, there is a group of potential founders who really want to pursue Linux on Power-based servers in more of a mass market mode, not just from IBM. However these groups really need to be voted on by the board. So it would be presumptuous to say these are the working groups in advance of the organization being formed and voting on them. It's really a founding member board decision to form those groups.

So we are really giving the power to Power.org to make these decisions through collaboration; it's not IBM saying, "It must be this way."



Back to top


Commitments and dues

dW: How long has this been in the works?

Ireland: Really, since the August time frame. So it hasn't been that long. We've made a lot of progress.

dW: What has been the reaction of the companies we've approached to be among the founders?

Ireland: The founders have been very positive. This is clearly on a fast time line.

I'd say they were welcoming of the idea. They understand the value proposition for them. That's really where most of the focus has been working with the founders.

Because it's much more than just spending money on joining the organization. It's as commitment to contribute to the organization, which is much more than a membership fee, it's putting resources on it. That's the key, the value proposition to them, and their commitment to actually put resources into the organization to contribute.

dW: You mention membership fees; what kind of dues are involved?

Dykas: Yes. Some of the dues thinking is being finalized, but as we are right now there is a Participant level ... all the way up to the Board level, of being a member of the Board of Directors, which is where the power of managing the organization [rests].

At the highest level of the organization, where the Board of Directors occur, where the Founding Members will be on the Board of Directors managing the budget, the activities of the organization, certifying output or ratifying output of the individual technical and marketing and compliance committees, that's where the [masters] or the running of the organization come from. And those dues are currently being finalized.

dW: So it's not as if anyone has to be a multimillion dollar company just to play?

Dykas: Not at all.

dW: That's pretty open, as standards go.

Dykas: Yes, it is. And there's an intermediate level there called a Sponsor, which will be for those companies that, while at a Participant level you can participate in committees, you can't really run any of it. You don't have the right to run the committee, and then that's the next level of sponsorship.

As well, there is an Advisory Council, which is made up of those companies that have the greatest investment and licensing of the Power Architecture. You know, there are architectural licenses that enable people to dive down into the deepest levels of what Power Architecture is and do things and modify it for their needs.

For example, you would not have, say, an Application Vendor or a Linux vendor care very much about the Advisory Council at that level, but you might have somebody who is a chip manufacturer or a chip designer say, "Look, I have these needs for designing a Power Architecture-based chip, but I need these modifications." So those sorts of things would be run through the Advisory Council. That Advisory Council is an advisory council for IBM.

dW: How about individuals? Let's say I'm really curious about Power Architecture and I want to start going to meetings. Can I just do that?

Dykas: There is a certain level of participation that we're calling the Developer level, and your ability to participate in meetings, and that's more the individual level. The ability to participate in those meetings is more limited.

There will be what we're calling a Developer Community that will focus on supporting the individual in terms of what's going on with the Power Architecture, and then, also, there will probably be a Web site where other information about the work that's actually going on inside the Power organization is available for people to take advantage of and see what's actually being worked on once it reaches a certain level of maturity. But the organization, as a whole, is open to any company that wants to do work inside the organization with respect to the Power Architecture and come and work in committee.

dW: So as far as the structure goes, at some point down the road there might be, for instance, a sub-Technical Committee focused on something like, I don't know, things common to all the Linux on Power vendors, for instance.

Dykas: Yes! Exactly. Maybe it's a compiler. Maybe it's something related to tools.

dW: So if there was a compiler meeting, you would end up expecting to have, say, some Metrowerks™ people and some GNU people, and whoever else is doing compilers targeting Power.

Dykas: Yes.

dW: So would that be something where basically each of those subcommittees would set the rules for attending its own meetings?

Dykas: Yes.

dW: Will there be some information about each of those committees up on the Power.org Web site?

Dykas: At varying degrees there will be public access, and there will be private access ... but there are specific things where there's work in progress, that you don't expose to the broad audience until you're ready to kind of get it to that 0.9 level, and then when it gets to that 0.9 level you put it out for comment.

dW: Yes. That makes sense. So, basically, the idea is that any such committee that would be related to the Power Architecture should in theory be somewhere around Power.org?

Dykas: Yes.

dW: How would the intellectual property for that then be licensed?

Dykas: Well, the Power Architecture by itself is still licensed from IBM. But if you're designing, say, some complementary work around a bus structure -- say, in an SoC, a specific bus structure other than CoreConnect™ -- the goal is to have people be able to come together and work on that particular bus structure, and in the end part of the prerequisite for joining that sort of technical subcommittee is that subcommittees will choose from a limited number of IP rights policies before they begin. It could be a royalty-free basis -- it could be a RAND basis [a RAND meaning reasonable and nondiscriminatory terms --eds.], or it could be some other basis. But the point is that you agree to participate in some of the developing and technical activities of that group, and agree to whatever IP policy that group is formed under from the beginning.

dW: Are we going to be opening up documentation to the extent maybe approaching the same level of openness as, say, the Purple Book?

Dykas: Yesss..... I think right now we are working through the specific level of what IBM exposes and what IBM donates to the specific subcommittees. Some of these things will be dependent upon -- you know, it would be a balance of things that we know we need to expose and/or donate to see these new areas, and some things that we need to not donate.



Back to top


Enlarging the doughnut: Linux G5 servers and more

dW: The difference in the marketplaces for X86 and PowerPC embedded boards and single board computers is incredible, just because there are so many X86-related chips out there now. For X86, I can find five or six different boards I can buy for US$200 or US$300. For PowerPC, I found a couple vendors who will come down as low as US$1500. So I see some benefit to getting more licensing and people in the field.

Ireland: Absolutely. And for the single board computers, for the higher end of those board computers, are still using stand-alone processors, for the lower end, maybe more system on a chip solutions. On both fronts, we need more partners. We need more people designing those computers, we need more software available for those companies to succeed. This is really a great example of the whole premise around the Power organization we're launching.

With Power.org, we expect many more of these ODMs or OEMs to be providing board solutions and it can be higher-end solutions, server class machines or embedded solutions, more use for control. More companies who will be offering board-level computers using a variety of different software. We expect this is going to be a key focus, especially around high-volume servers.

One idea is to create a high-volume server platform based on the G5 or 970FX class chips, which are really derivatives of the POWER™ line. So we're trying to get a lower price point, high-volume server platform around the Power Architecture. You are starting to see solutions today from Apple and IBM. You see Virginia Tech stacking many of these 970FX processors together with Apple machines and also with IBM Blades to create very powerful supercomputers.

It is the plan to get as many direct sources available for member and non-member companies, to have more solutions available in the marketplace.

dW: Thank you so much for taking the time to talk with us today.

Dykas: No problem.

Ireland: Thank you.



Resources



About the author

The developerWorks Power Architecture editors welcome your comments on this article. E-mail them at dwpower@us.ibm.com.




Rate this page


Please take a moment to complete this form to help us better serve you.



YesNoDon't know
 


 


12345
Not
useful
Extremely
useful
 


Back to top