Ed Zander speaks out
Sun's president talks about his company's competitors, UltraSPARC,
In an exclusive Interview with SunWorld Online, Sun president Ed Zander covered a range of topics including when Sun will ship a 64-bit versio n of Solaris, the future of NT, HP, SCO, UnixWare, and the Pentium Pro, how Sun will make money with Java, and when users will see computers using the UltraSPARC-II chip. (4,200 words)
SunWorld Online Despite your stock going up, there have been reports in places like Computerworld that Sun has been losing marketshare to IBM and HP because the SuperSPARC processor wasn't fast enough to satisfy people. What will the UltraSPARC do for Sun's future and Sun's customers?
Ed Zander There are no short answers there but I'm looking at the same data that other people are looking at, and new IDC and Dataquest data in unit-volume shipments for the period ended in June actually indicated us going up in marketshare. We have a slide that shows that for five consecutive years, we have been between 39% and 37%. It goes up and down. So I take issue [with that] especially on unit volume. Even with what I would consider to be not the best position of performance, we have maintained market share. I think this first half of the year we grew, actually, so I disagree.
As far as revenue goes, the issue I have with that is the measurement criteria and the way they do it. HP doesn't give out their numbers, so analysts have to guess what HP does. Ours are audited numbers. We have every indication that we are growing this business and growing in unit volume. We ship today more workstations and servers than HP, IBM, and DEC combined.
SunWorld Online I think the issue was that your growth had declined slightly and theirs had increased. No one is saying that you're not still No. 1.
Zander We're still No. 1 by a fair amount.
The thing that people don't understand is that Compaq is No. 1 in the PC industry with 12% marketshare. We are No. 1 in the workstation industry with 35% market share in annual sales. To have one out of every three workstations and servers that ships worldwide is almost an incredible position to be in.
SunWorld Online With your processor performance lagging behind your competitors, why are you doing so well?
Zander We began to get into the commercial marketplace. Most of the sales in the '80s were to the technical marketplace. But with the boom in the commercial side and rightsizing, and our focus on servers, our server business grew 50% year to year. Our data warehousing, customer management, telecommunications, and financial services products all had growth. Plus, the offloading of VAX/VMS, HP proprietary systems, and IBM mainframes spread a lot of growth. In that case, it was applications and scalability and database performance as opposed to raw floating-point, CPU, graphics performance -- not to say that we haven't been under a lot of pressure in the technical space. I'm agreeing with you. In the technical markets, graphics, performance, and floating-point does count. We have had a lot of pressure there and we've been under severe competitive pressure to move our performance up.
On Tuesday we correct that and go after this market with a vengeance. We're back and they're going to regret it. Somebody said the other day that HP's going to regret the day they bought Apollo. And SGI users are going to ask, why pay more?
SunWorld Online What went wrong with SuperSPARC?
Zander I don't know, you'll have to to have to ask the engineers that, I'm not technical person. In this market, you have to scale and scale quickly. You've got to scale performance and you've got to have predictable increases in performance in a very timely fashion. We just didn't get it out of that generation. We did not get the scalability and we did not get the satisfactory timely release of the different clock rates. I don't know if it was the process or the design.
SunWorld Online So I understand you sold 400,000 boxes in 1995.
Zander No, SMCC sold close to 300,000 boxes. I think if you take a look at the Solaris and SPARC licenses, the overall Unix licenses, it may be closer to that number [400,000].
SunWorld Online How many are you projecting to sell in '96 and what percent of those will be UltraSPARC?
Zander I don't give those out, I really don't, I'm not trying to be cute. I feel our business is growing at a pretty good clip, at least Wall Street thinks it is. And the Ultra series we're announcing -- if you come on Tuesday the message is, and we believe this, this is comparable to the Sun 1, SPARC 1. This is the third major announcement of this company in 13 years. And just as the SPARC 1 in '89 was $12,000 at 10 MIPS or whatever it was, then in 18 months it was the entire product line. Stay tuned, you will see servers based on UltraSPARC. Everything in the product line will be replaced. If you take a look at what we did with SPARC, you can probably guess it will happen at that rapid a rate. By this time next year, a majority if not most, the bulk of our product line, will be UltraSPARC-based, Ultracomputing-based.
SunWorld Online That leads naturally to the next question. You have to have Solaris 2.x to run on these systems, but only half of your installed base is on 2.x, what do you expect users to do and why haven't they moved? Why is it taking so long?
Zander It's funny, but I don't hear that much anymore. All of our servers, the 1000s and the 2000s, all the successful products are Solaris-based. Any customer that wants to buy a server from us is buying Solaris 2. There are a number of customers who have desktop products which are written for Solaris or SunOS. They will move. They are moving. It's not a big issue. The nice thing about this version of Solaris is that we're not porting, there are no recompiles, no ports, it just runs. We have testimonies from tens of ISVs, 30, 50, a hundred, that will tell you they just plugged it in and it ran. It runs two to four times faster.
SunWorld Online You mean Solaris 2 on UltraSPARC.
Zander Yes, all the apps now -- Cadence, Mentor, Parametrix -- they're all there. I don't think this is an issue at all. This is a machine that's going to drive everyone who wanted it [Solaris 2.x]; the last bastion that was looking for an excuse [to move to Solaris 2.x]; this is it.
We've got some incredible power, incredible graphics, incredible networking, incredible balanced systems that are going to scale up and down our product line. We have also put in place a very ambitious, aggressive, funded Solaris 1 Solaris 2 team inside the company, and this team has been out there for the last nine months in our top 100 accounts basically getting them porting aids, porting tools, porting guides, for their custom applications, system-administration issues, and everything else.
I'm not trying to downplay it, but as a '96 issue, it isn't what it was a year or two ago. We have very few customers who come in and ask us for Solaris 1. There are some people who are just happy with what they do. The facts of life about operating systems are that they take a long time to do and a very, very long time to go away, whether it's us or Microsoft's distributing DOS.
I sometimes bump into DOS users and say why aren't you on Windows or Win95, and they say because DOS works and I've got my applications and I'm into it. It takes a while, because people do things with operating systems that they don't do with hardware. They write code, they get into the kernel, they do custom applications, they've got their trained employees, they've got their users.
I've said this a number of times: We underestimated two years ago how quickly you can move a very large installed base. That's why there's a code of "can't do" in the company that McNealy and team laid out. That's why every subsequent release of software, whether it's a compiler or an operating system, runs the applications. We don't break applications anymore. You can't even talk about it inside the company.
SunWorld Online What about when you come out with a 64-bit operating system?
Zander We're going to run the applications. We're not going to do it again. In life you get to make a mistake maybe once. You don't get to make it twice.
We have tons of customer quotes. They had the beta machines and they just couldn't believe it. They found that the application just ran, they didn't have to do anything.
SunWorld Online Does it run in optimal way? Do you need to tune it if you want to get the best performance?
Zander Nothing. Nothing. You can get more if you want.
What's really incredible about this release is -- you'll see when you talk to customers -- we're going to do a lot of benchmarks with competitors. This is going to be a very technical announcement in that we're going to do data, benchmarks, customers, applications, performance. What's really interesting about the customers is this is truly an applications machine as opposed to those SPECmarks and all that other good stuff we talk about.
What we found out is that if the SPARC 20 for example is 1, and these machines were a 2, then the applications are running 3x. And the reason is the convergence of what we did with the VIS instruction set on graphics, the network bandwidth, and the CPU power all on the chip and all on the board, and the bus architecture. It's more than just the chip. This is our major theme. This is not a chip announcement, this is a paradigm announcement from Sun. This is a new systems architecture that brings visualization, 3D animation, imaging and video together with high-speed networking in a very powerful 64-bit processor. We've eliminated all the latencies, all the extra boards, and all the extra chips. These customers, when you see their numbers, the SPECmarks could be 2x, and they're getting 4x performance here, or 3x or 5x. As I told the press, don't take our word for it, just talk to them. And talk to the ISVs.
SunWorld Online Just briefly going back to the 64-bit operating system, when do you plan to have something like that available?
Zander I think you have to talk to our software arm for dates.
SunWorld Online I just mean something general.
Zander It's a two-stage effort. First of all, many of our customers, especially in the commercial marketplace don't ask for that. They ask for stability, quality, scalability, performance, and application credibility. The second thing is that in some of the applications, what they're really looking for is file and address space, as opposed to a 64-bit operating system. And I think you're going to see us address that in Solaris. You've got 64-bit extensions, effectively, and that will happen over the next year, and then beyond that is a 64-bit operating system.
SunWorld Online Do you think you might having something out by '98?
Zander I think you gotta talk to them.
SunWorld Online You must have a vague moving target, because HP and Intel are moving pretty fast to do 64-bit operating systems and they're talking about it a lot.
Zander That's all they're talking about it. If you think SCO's going to develop a 64-bit operating system...
SunWorld Online Well, but HP...
Zander HP's getting rid of Unix, they're doing NT. Can't you see that? Doesn't the press -- not you -- but doesn't the press just see the cards? HP announcing that they're doing Intel chips, they link the Precision architecture with Intel, now the Intel chip is dividing into two chips, two P7s, they've got the Intel P7 and the HP P7.
SunWorld Online They swear they're doing one chip.
Zander No, they said it. I'm not saying it.
SunWorld Online When did they announce this?
Zander They said it this week. The press is reporting that there's two P7s. One is the Intel P7 and one is the HP P7 because they couldn't get the binary compatibly that we said they couldn't do. It's too hard. And now you're telling me, and I've been in the software business for four years, and I did the AT&T merge with Berkeley Unix with the best engineers in the industry and I did Solaris 1 and Solaris 2, and yet you tell me they're going to take UnixWare, SCO Unix, and some HP stuff and meld together 10 million lines of code and produce a 64-bit binary operating system in a year or two, and they are going to do it out of SCO which is right now laying people off?
SunWorld Online Well, this gives you a real good opportunity then. Because the HP Intel thing -- if they follow through on what they said they'd do -- would be a great competitor to you.
Zander I'm betting as much as I'm sitting here right now that the HP answer for that chipset in 1997 and 1998 is one and only one operating system: NT. HP is out of the Unix business. If you're an HP user buying Precision architecture or HP-UX, you've got nowhere to go. You're VAX/VMS. This is no different than DEC. DEC still sells a billion dollars of VAX/VMS, so that doesn't go away. But HP-UX and the Precision architecture is VAX/VMS and they're going to migrate their entire base to NT P7 -- which by the way, may be a very great competitor, but that's the target.
This SCO thing is just a diversion. HP has invested nothing in objects. There was the Taligent investment, which is ppplllffff. So they have no people on things like objects, no people on things like Java, no people on things like Internet, they have down-scaled that core operating system group. They are just not investing the people to do operating systems, especially Unix. This is just "let's give it to SCO." When has SCO ever built anything other than a small business Unix operating system?
SunWorld Online Of course, they bought UnixWare, didn't they?
Zander UnixWare -- you might as well take that and throw it in the bay. That's a waste of time too. We're talking about building a 64-bit Unix on which people will take Oracle, Sybase and run their enterprise, and SCO's going to do that? I mean, I don't know sometimes, the members of the press -- I read this stuff and I go, I could sell them anything, I could sell them the Brooklyn Bridge.
SunWorld Online So, tell me, what technical superiority does Solaris offer over NT?
Zander Today? Well, technical, that's pretty easy. That's not the problem with NT, by the way. If you're running a technical battle, it's the scalability, it's the security, it's the systems administration, it's the network management, it's the networking capability -- enterprise networking capability, as opposed to local networking capability. I just think it has so many technical features about it. That's not the problem with NT. The issue with NT is it runs Microsoft applications. The user, because of the lock in of the applications and the operating system, and their desire to track their productivity apps, buys NT.
SunWorld Online So what's Sun's response to that, Sun's strategy?
Zander Well, our strategy was to try to get the industry to see if Microsoft could uncouple their applications from their operating system, and that has so far has failed. So our strategy is threefold. One is to continue to invest and find technical solutions whether it be WABI or something else to run Microsoft productivity applications native on Solaris. Two is to continue to embrace Microsoft desktops with Solnet and Solstice, administration, and network management.
SunWorld Online So, to be a really good server for Microsoft?
Zander To be the server for Intel desktops. And third is to play out probably the most exciting thing this industry has written about or seen in the last couple years, and that's the Java thing. There's two stories developing here next week. Tuesday is Ultracomputing, and Wednesday we tacked on a Java day. We didn't even publish it, we just put it out in the Bay area here and in about four hours we had 1,100 people sign up. It's sold out. We could probably get 5,000 people in the room. These are developers.
SunWorld Online How do you guys plan to make money off Java?
Zander You don't. You make money off of Java by creating a new paradigm that enlarges the playing field, in which you get to play and sell your servers, sell your desktops, sell your software structure. If the paradigm of the year 2000 is only Microsoft/Intel, then I ought to go golfing somewhere. If you can create application developers that begin to write network-based applications on things like Java, that separate the operating system from the application, then you create a whole new paradigm that you enlarge, you just blow out. You do in the next ten years what the PC industry has done in the past ten years. You have tens of thousands of developers now writing new-age network-based applications, and if that's the case, we're going to put our Internet technology on security, on Java, development environments, operating environments, the servers we provide, the video servers, and high-performance desktops and eventually Java devices. So someday, if you believe current thinking -- not only from Sun but [Larry] Ellison or Netscape or some of the other visionaries and luminaries -- then your low-end PC gets replaced by a Java device. All you need is an intelligent browser, a smart microprocessor, a piece of glass, and a network connection.
SunWorld Online There we'll be on the beach, faxing.
Zander Well, you might have your cellular phone on the beach doing a lot of interesting stuff with Java applets coming across it. I don't know. This is going to take a while to play out as we sit here today, but as I caution people, the workstation industry started in '80, and in '85 it still wasn't a billion dollars. The PC industry started -- what? -- '77? -- and it wasn't a billion until God knows when -- '82, '83, '85. It takes a gestation period for a paradigm to emerge. So I think when you write the history book ten years from now and look back, you'll see that '94, '95, '96 was the convergence of bandwidth, 64-bit microprocessors, of video compression technology, of Java-type languages, the exposure of the Internet (both little i, meaning private, and big i, meaning public) -- and yet people didn't think how to apply it.
Java for the first time has excited the intellectual capacity of programmers. We haven't run a Java ad, we haven't run a Java press release, we don't have a Java brochure, we have nothing. All it is is the Internet and it's appealing to the masses, to the programmers, to the developers, not to the bosses, the CIOs, the CEOs. And the reason is the programmer now sees a dream ahead. He's dreaming again, he's creating again, he's not writing Microsoft Word. He's beginning to think "I can use the network to write applications." It's fascinating to see how this thing plays out. Over the long term if we can break the MS monopoly. Everywhere I go I ask people why they are buying NT. Everybody says so I can run my Microsoft applications. Nobody says it's better than Unix.
SunWorld Online Java over the long term will be how you break the Microsoft monopoly?
Zander By changing it to a network-centric model as opposed to a desktop model. If I told you two or three years from now what's important on your desktop isn't important. In other words, if you want to buy a Mac, you can go buy a Mac, if you want a Unix, you can go buy Unix. If you want Windows, you can go buy Windows. Then you get your applications over the network. Most of your applications over the network, not all of them.
SunWorld Online That sounds like the IBM microkernel strategy. They're porting their microkernel to the SPARC platform. Does Sun have any interest in making Solaris run on the IBM microkernel?
Zander [Looks incredulous, makes sound like he's spitting out his spinach.] Doesn't sound interesting to me. Sun runs on the PowerPC natively pretty well.
Maybe I'm just biased because I've done this for four years, but don't microkernels and merging source codes and developing new operating systems take years? It took us three years [to develop Solaris 2.x] and we knew what we were doing. This stuff is hard, it takes a long time. Look at NT. It's taken them a number of years. It just takes a while.
Solaris has won. It's the alternative to NT. There isn't any more volume Unix around, there isn't a better multitasking, scalable, secure enterprise-wide, client-server operating system around. It's got more volume and more applications than anything. It ain't gonna be HP-UX because they're getting rid of it. It ain't gonna be the Unisys and NCRs. It ain't gonna be UnixWare. We won. The sixty-four-dollar question is will there be an alternative to NT? If it is, Solaris. But Solaris with objects, Solaris with Java, Solaris with Solstice. We think we've got a shot at providing a fundamental alternative environment to an NT environment. That's the plan.
SunWorld Online How is Solaris on Intel doing?
Zander I don't know, you gotta talk to them.
SunWorld Online Oh, come on, you must talk to your old friends. Don't turn into an IBM on me where one hand doesn't know what the other is doing.
Zander I would not want to represent them, is what I'm trying to say. They were in San Francisco the other day, and they had a huge booth running the Pentium Pro for their announcement. I think they're doing better, they've got a really good version of this thing out. It took them a few years to do, but they've got applications running, Oracle and Notes. Solaris on Intel runs pretty well, and it runs pretty well on the PowerPC.
SunWorld Online Why are you running on PowerPC?
Zander Because we said that's going to be the strategy of our company. We're going to take our software to places and make them cross-platform. We have no choice. And they think the PowerPC is going to be the third volume platform: SPARC, Intel, PowerPC.
SunWorld Online I haven't really seen you since you left SunSoft. What have you done personally to change the direction of SMCC?
Zander Well, I think SMCC was in reasonably good shape when I took over. I think we had to do a couple of things. I had to establish that the company was not a Unix-box workstation company, that the two words "network computing" were ingrained in every SMCC person's brain. We got advertising and marketing going. We reorganized the the company into a series of business units to try to get time-to-market and get a good focus. We got focused on sales of applications and on industries, much more of an industry focus in there. And we got back time to market. We're pretty much focused on quality. I laid out eight or ten things we wanted to accomplish and we've done pretty well. Knock on wood.
SunWorld Online One other quick question. You previewed the UltraSPARC 2 at the Microprocessor Forum recently. When can we expect to see systems based on the chip?
Zander Chet [Silvestri] has announced he's taped out, spec'd the chip, and I think he expects to see it coming out in the second half of '96.
The big news about this round of UltraSPARC 2 -- which is very exciting -- is we already have a 143 MHz, a 167 MHz, and we'll be showing you Tuesday a 200 MHz UltraSPARC 1 chip. That's already scaling faster than I can inhale it. That's a surprise.
SunWorld Online So we might see UltraSPARC 2 machines sometime in '97.
Zander Well, it could be '96. I don't know. It's hard to predict.
SunWorld Online Anything else you'd like to talk about?
Zander We're a little harried, we've got a show to put on. This whole thing was supposed to be in a hotel in New York, but we really wanted to bring it back to our roots, bring it back to the Valley, bring it back to our employees, and more importantly bring it to our customers who have stayed with us in the technical market for the last ten or twelve years. We wanted to celebrate it here. We got lucky because the Cirque de Soleil had some extra tents.
SunWorld Online They're not sending down any jugglers?
Zander No, this is a not a circus.
SunWorld Online If I remember correctly, Scott McNealy juggles.
Zander Well, McNealy may do something, he might do something with a dog.
About the author
Cate T. Corcoran is a San Francisco-based freelance writer. Reach Cate at email@example.com.
If you have technical problems with this magazine, contact firstname.lastname@example.org