- Sun Microsystems has a knack for being the life of the party.
It charged into the LinuxWorld conference here last week, arriving at the show several years late and with a lukewarm invitation. Still, Scott McNealy (pictured right), chairman, CEO and president at Sun, managed to charm the open source crowd with a witty keynote speech full of jokes long told in the Unix world but new to the Linux zealots.
McNealy's comic flair helped draw attention away from the fact that Sun just released its first general-purpose Linux server this week -- a two-processor system running on Intel chips and the new Sun Linux operating system. IBM, Hewlett-Packard and Dell all pointed to their laundry list of Linux products and called Sun a laggard. McNealy made jokes and stole the show.
Sun may have had the biggest news for this year's LinuxWorld conference, but questions abound as to whether its support will be fleeting or it has started a lasting relationship with the open source OS. In a recent interview with the IDG News Service, McNealy addressed some of these issues, saying Linux fills a very real gap in Sun's arsenal and adds choice to its line of servers that use Solaris on UltraSPARC processors.
McNealy said he is ready to live the "Linux lifestyle," even if it means following a feisty crowd.
IDGNS: What strikes you about the Linux community?
Scott McNealy: It’s a tough community, because if you have any copyright, ownership, revenue or profits, you are not a good person. At least with some of the crew, right? I think other people understand there are many different models, and the models can co-exist. I’m a person who can play all lifestyles, and I am not ashamed to be open source and a capitalist. I get invited to all the conferences.
Could you lay out how you go to your customers with these different operating system offerings you have now, with Solaris on UltraSPARC chips, Solaris on Intel chips, and Sun Linux?
We are giving them a choice. We are giving them 64-bit hardware or 32-bit hardware. We are giving them horizontal as well as vertical scaling capabilities. Then we are giving them Solaris 9 or Linux. "Two for the price of none" is what I am calling it on the x86 (Intel instruction set) box. You can run your Linux apps on Solaris with Linux extensions or you can run your Linux apps on Linux and live the lifestyle on x86. And we don’t charge you for either. It’s basically just bundled.
Plus, we give you a whole suite of applications, middleware and tools -- some from the open source community and some from the Sun ONE (Open Net Environment) stack on both platforms. We also give you the Control Station environment, so you can manage large numbers of these things in big Web site and enterprise types of deployments. It’s, I think, a pretty attractive thing. We are absolutely, totally price-performance-competitive on the hardware in the 32-bit space now.
Why go with the Pentium III chips for your first foray into this space? IBM released a beefier two-processor box this week, and some people call yours trailing edge technology.
We’ll be there. Right now we think the sweet spot in the pricing is with the Pentium III. The Pentium 4 makes more noise, but it has not necessarily proven its price-performance advantages yet. So we are pretty excited about the performance value per dollar that we are getting with this box.
Is that value mostly coming from the software side, then?
We think we can do a lot and have done a lot of tuning with Solaris and Linux on the x86 box. Customers prefer to buy metal-wrapped software, certainly metal-wrapped operating systems and metal-wrapped middleware. My definition of middleware is third-party software that should have been bundled last year for free.
Along those lines, I am wondering how you see companies like Dell, IBM and HP playing out with bundling software on their systems.
My reaction is, why wouldn’t you buy your 2-way Linux servers from Sun, as opposed to, say, Dell? We both get the hardware from the same Asian supplier, laundered through Intel at the same price. So there is no advantage there. We are absolutely at parity. We have two for the price of none in operating systems. Dell has to go out and buy Red Hat or go out and buy Windows or go out and find some Unix. So, we have a better choice there at a lower cost.
The third point is the Sun ONE stack. We have that whole suite of stuff. Dell doesn’t have any of that. They don’t do any R&D.
Fourth is my service and support capabilities. Every one of my engineers is a Solaris and a Linux engineer automatically because they are Unix engineers. I’ve got a worldwide service, support and distribution capability where you can call direct, locally, to get someone to help you. You are not getting someone at Dell that switches you off to Delhi or some other place where somebody is trying to answer your Linux or system problem.
The next thing we’ve got is an upgrade path to 64-bit. We have a compatible systems management suite, operating system and our application environment -- Sun ONE and Java environment -- that migrates to 64-bit if you do want to vertically scale or just performance scale. Nobody has the story we have.
IBM does have some advantages in that its application server and other middleware has more market share.
I don’t think they have any advantages on the app server. If you are writing to J2EE (Java 2 Enterprise Edition) and you are writing clear, crisp, compatible J2EE code, it runs on our stuff, and we have better price performance by far because of our pricing model.
It just seems that IBM can mimic your strategy of bundling software better than can, say, Dell or HP.
I would say that is fair, because IBM is number two in Unix servers. But falling like a rock.
I said it last quarter, but nobody talked about it. We crushed IBM. This most recent quarter we crushed them again.
Can you give me an idea of how some of the trends in the server world will play out then, particularly with HP coming on with servers based on Intel's 64-bit Itanium chip? If Itanium pays off, it would seem you have the most to lose, HP has the most to gain, and IBM sits comfortably in the middle.
We think Itanium is done, and we are not the only ones. Dell is not doing it, NEC has had plenty to say about it.
Have you ever seen an Itanium machine?
You are like the third person I have ever met that has seen one. You were in third grade when it was announced. Itanium is not going to happen. It’s going to be x86 64-bit, not IA-64 (Itanium architecture). That’s going to be their strategy. They have their secret little project going on up in Portland, Oregon, to try and compete, and that is going to be their long-term strategy. Itanium is done.
In the 64-bit space we will be on UltraSPARC IV or V before Itanium even has a chance in the marketplace. The folks I talk to on the architecture side say that dog won’t hunt.
I am trying to get a handle on whether your new Linux server will be a niche product or if you will expand your Linux server line.
I think in the 32-bit space, it is going to be by definition a low-end kind of blade and edge server. The reason is because an 8-way, 32-bit machine does not make a lot of sense. At that point, you are starting to talk about lots of processing, and lots of processing typically means large data sets, which means 64-bit address spaces, so you get the large memory set. Two-way is the sweet spot there.
Some of your big partners, like Oracle, have come out in support of Linux. It does seem like there is some demand for a 4-processor system running Oracle's database, especially with this trend toward database clusters.
Talk to Larry Ellison [Oracle's chairman and CEO] and ask him what the sweet spot is for x86. Is it 1-way, 2-way, 4-way, 8-way? He’ll tell you it's 2-way. We match them for sure on price performance on 4-way systems, and they can't even get to 8-way.
When you take a 99-way UltraSPARC III machine and add a 100th processor, you get 94 percent linear scalability. You can’t get 94 percent linear scalability on your first Intel chip. It’s very, very hard to do, and they have not done it.
In that context, how big of a deal is it to have you as the only major OEM with its own distribution of Linux, if it’s only going to run on this two-processor system?
I think we will do whatever makes sense from a price-performance perspective. People don’t write to Linux. They write to the app server, and then you don’t care if it's running on Linux. Then, if you need something bigger, you can decide if a 4-way Linux box or a 4-way Solaris x86 box or 4-way SPARC Solaris box is your best price-performance machine. Right now, a 4-way SPARC Solaris is the best choice.
So you are saying there is no intention to evolve Sun Linux the way you have evolved Solaris to the point where most of the ISVs (independent software vendors) write to Solaris first.
We did, it’s called Solaris. We put 64-bit, internationalisation, localisation, clustering, RAS (reliability, availability and scalability), dynamic reconfiguration, multipathing, all the rest of it into Linux. And it's called Solaris 9. We put it all in there.
What's your strategy the next couple of years out? You have obviously heated up the competition against partners like Veritas Software and BEA Systems by championing your own software products. You have worked very closely with them in the past but now seem ready to try and go it alone where it makes sense.
No, we have always been and always will be integratable. So we will give you a full and complete integrated stack that will be integratable. We believe in modularity.
That could pose some serious problems for Veritas and BEA. They could potentially lose a large chunk of revenue from their products that sold on Solaris.
Hey, you know, I am in charge of the Sun shareholders. I am not in charge of Microsoft’s, BEA’s or Novell’s or Veritas’s or any of the other folks. They have to figure out what their business model is. I am not about to critique them. We have committed to them and work very hard with them to make sure their stuff is integratable.
But it seems like your actions prompted them to form tighter relationships with IBM and HP.
You want to go through the market share numbers one more time? You do not want to have really strong tethers to people who are losing share at these kinds of rates.
(This interview was edited for clarity and content.)