The following is an interview with Arvind Krishna, Ph.D., GM, development & manufacturing, IBM Systems & Technology Group on last Monday’s announcement by IBM of the OpenPOWER Consortium, which initially includes high-end GPU manufacturer Nvidia, high-performance IO builder Mellanox Technologies, Taiwanese white-ox server manufacturer Tyan, and Google. In the interview Mr. Krishna all but says that the initial aim of the consortium is to create a new generation of white box servers using high-performance computing (HPC) components built specifically to Google’s needs at a price that fits Google’s hyperscale architecture strategy. He goes on to imply that IBM and its partners may also challenge the x86/Microsoft Windows dominance of the server market and hints strongly that the IBM-Google alliance may expand to other, unspecified areas. He says specifically that sometimes alliances form because “very smart technical people on both sides share a common vision of how the industry ought to evolve… which is not necessarily the way that most others in the industry are behaving. That…forms a bedrock on which you can build the other parts of the relationship.”’ Because Mr. Krishna’s comments were very precise and very important concerning IBM’s intentions, the article is in the form of questions and answers and is a lengthy excerpt of the full transcription of the interview.
Dr. Krishna: Hi, good evening Bert.
CIOangle: Good evening, it’s so nice of you to find time to talk to me. I was at Edge this spring and I heard Ambuj Goyal talk about how IBM had to be an open systems company, because he said, “Today no one company can do everything.” Now this is the third major announcement, after Open Daylight and the alliance with Pivotal over Cloud Foundry, of a very significant association through the industry. I’m wondering is this is a formal strategy to develop alliances with other vendors across the industry?
Dr. Krishna: Well, yes.
CIOangle: IBM is moving and shaking this industry with these partnerships, and in particular an association with Google.
Dr. Krishna: I’ll come back to that one. But you’re first point is about Open Systems, right? I can’t help, but relive a little bit of our history here. I think it was 1999 when we embraced Linux.
CIOangle: Yes and of course IBM offers Linux on just about every hardware platform it sells.
Dr. Krishna: I think on everything. Actually a lot of our supercomputers are Linux.
I’ll also point to putting our weight behind KVM and the OVA or Open Virtualization, which you could say, “Well, wait, is KVM really successful?” Then I say, well, “95% of Open Stack implementations are on KVM.” That may be a telling statistic.
Then OpenStack last year, and people forget many other things that IBM contributed to the industry. SQL, in database standards, was a standard written by IBM and contributed to the industry.
So, Open is by itself a strategy. You create a big ecosystem, you got to put enough in that has value and then every individual in it will try to make money by, I would call it, quality of execution, quality of engineering, and providing real value beyond just what’s in the Open community.
CIOangle: IBM certainly has worked in the Open community for a long while. Then does IBM have a formal strategy now for developing strategic alliances through the industry or are the three that we’ve seen announced just in the last few months just coincidental?
Dr. Krishna: No, it’s not coincidental, but these things take blood, sweat, and tears to get there right. It takes a lot of hard work and a long period of time to get, I’ll call it things aligned. It might look like they are coincidental, but I think that they’re both strategic and in the business interests of all the parties to get there. It’s not opportunistic. It’s not grabbing something for like one quarter or one month or six month and be done. It is something which is sustainable and which should have lasting value to our clients for years or decades.
CIOangle: As the release said, is the first time IBM has offered its RISC (Reduced Instruction Set Computing) Power processors to third parties. It really looks to me like an entire change of strategy because PowerPC was always the top-end of IBM’s product lines. The most powerful thing and the premium. Now it looks like you’re changing that entirely. Why is IBM changing that strategy now?
Dr. Krishna: I think it really comes down to … I’ll call it the life cycle of any industry. You can always look at history. I think in every industry you kind of go from, “I need to get volume manufacturing down, so it’s mass standardization.” Then you begin to switch towards mass customization. And by opening up the RISC Power processor, it’s the Linux operating system on it, it’s the motherboard designs. You see, you’re allowed to optimize it to your intended purpose as opposed or having to buy only the models I provide. You’re allowed innovate around the processor. Today, when you buy a Linux box from, let say Intel, it’s their memory controllers, their I/O, it’s their designs, right? We say, “You can choose to innovate,” and that should open up optimization and hence competitive advantage.
CIOangle: Is the plan here that IBM will supply the other consortium members with Power chips, or is it actually going to provide the specs and allow Nvidia and these other manufactures to make their own Power chips?
Dr. Krishna: Well, the answer is very clear: Both. If they want to purchase them from us, they are welcome to do so. We’re no longer going to insist that they purchase a whole computer. They can choose to purchase just the chips. Or if they want to license the design and go fabricate it by themselves, that is permitted as well.
CIOangle: So if Nvidia wants, it could create a chip combining the Power CPU and Nvidia GPU, which would be a very high performance processor.
Dr. Krishna: Well, you can now begin to see the logic behind the selection of the founding members.
CIOangle: Why is Google in this consortium? Google of course doesn’t build hardware. It’s a gigantic hardware consumer.
Dr. Krishna: Google is a hardware consumer, at that scale it always worries about how they could optimize every component they use for their intended purpose. Giving them the ability to do that optimization beyond what they could they get from the industry today, it just doesn’t usually happen.
CIOangle: I see, so what you’re really saying is that one of the things is going to happen here is there’s going to be a custom designed system for Google’s needs?
Dr. Krishna: I didn’t quite say that, but I said it was possible.
CIOangle: That means a white box type server priced to go into a hyperscale architecture, because that’s what Google does.
Dr. Krishna: One can certainly imagine that. They can put together the dots that are there, right? Tyan, You look at what they do. Nvidia which I pointed out what they do. And what does Mellanox build? They produced highly optimized I/O, let’s think of it that way. I suppose so far as I/O being agnostic to the processors, maybe its no longer agnostic. Then Google who wants to consume perhaps some of these things.
CIOangle: If Google starts considering them, probably the other huge Internet companies that do this hyperscale computing model will also take a look at them, which means that you will get a huge increase in the volume of production of Power chips.
Dr. Krishna: Well, that will certainly be attractive.
CIOangle: Wikibon CTO David Floyer wrote a piece a few months ago about how Intel’s dominance is slipping. What he was talking about was that ARM, which of course also makes a RISC chip, has grabbed most of the mobile market.
Dr. Krishna: It’s hard to debate given that Android as well as Apple use ARM.
CIOangle: He said that this makes Intel vulnerable, so let me just ask this: Does IBM smell blood in the water? Is IBM going to take this and go after Intel/Microsoft with Power Linux servers?
Dr. Krishna: I think just it’s not polite to necessarily go after your competitors that directly. We must be a bit of — I’ll use the word bit of a sporting player. Do we believe that many customers who today use only Lintel are now going to be hoping for an alternative that gives them an advantage? Yes, I absolutely believe that. Perhaps like what ARM did to the mobile market, we could do for many people who run large data centers with OpenPower. That answered your question?
CIOangle: Yes, thank you very much. The idea of an alliance between IBM and Google is utterly breathtaking. I don’t suppose you would care to comment on any other possible areas where IBM and Google might work together?
Dr. Krishna: I’m not going to comment or speculate in that direction, but I will take you to a different place. Alliances between companies are formed often for business rationale. I would also tell you that they are sometimes founded because really smart technical people on both sides share a common vision. They share a common vision about how the industry ought to evolve. I have no hesitation in saying that in the steps leading up to this point I think we actually do share — and I can say that on a personal level — a common vision about how things ought to evolve, which is not necessarily the way that most others in the industry are behaving. That helps because it forms a bedrock on which you can build the other parts of the relationship. That answers a lot of your questions, but not all of your questions.
CIOangle: That is an intriguing thought. Thank you very much. This is been a very interesting conversation.
Dr. Krishna: Good talking to you Bert.
Latest posts by Bert Latamore (see all)
- Researchers explain why most IoT action will happen at the edge - September 27, 2016
- How big data and DBMS markets fundamentally differ - September 27, 2016
- Wikibon offers 12-point evaluation framework for big data - September 27, 2016