On the hot seat at Sun

Sun Microsystems has an ambitious plan to revive its fortunes. Now it's up to newly appointed CTO Greg Papadopoulos to design the technology that will make it happen.

Mike Ricciuti Staff writer, CNET News
Mike Ricciuti joined CNET in 1996. He is now CNET News' Boston-based executive editor and east coast bureau chief, serving as department editor for business technology and software covered by CNET News, Reviews, and Download.com. E-mail Mike.
Mike Ricciuti
11 min read
Greg Papadopoulos is on the hot seat. The chief technology officer of Sun Microsystems is tasked with a monumental job: Design and build the underlying technology needed to turn around Sun's revenue slide and restore the company?s standing as a technological and thought leader in the marketplace.

And now, Papadopoulos also has some big shoes to fill. He has just taken over the responsibilities of departing chief scientist Bill Joy, who pioneered many of Sun's key technologies.

Sun's comeback rests on a plan to deliver to customers prebuilt packages, once every quarter, of software components designed to work together without a hitch. The company will roll out more details of that plan at its SunNetwork customers conference this week.

The two cornerstones of the plan are Project Orion, a collection of server software for handling e-mail, hosting Web sites, running Java programs and performing many other tasks; and N1, Sun's forthcoming software designed to let programs run efficiently on collections of different hardware. Underlying those initiatives are budding industry trends like utility computing and Web services.

CNET News.com asked Papadopoulos about Joy's legacy, what utility computing really means, and where Sun and Microsoft really clash.

Q: With Bill Joy's departure, you take over his responsibilities. How will that affect your day-to-day duties?
A: First of all, I can't replace Bill Joy. Nobody can replace Bill. He's a true individual. The responsibilities I take over for Bill are his research projects and a couple dozen people. I've been paying for them anyway, so there has been a dotted-line responsibility, and the person who has been running the projects for Bill reports to me. The structure is the same. It's maybe a 10 percent change in the number of people working for me. And I'll miss him. I don't want to minimize that. But for the past few years, he's been off worrying about other things, so that I don't even think that Sun would be the best place to express himself. He has a lot of really interesting ideas about where we are going in nanotech and biotechnology. I think he will still be involved in the industry.

Bill Joy has done work on Jini, among other projects. We haven't heard much about Jini lately. Where are those projects now, and what does the future hold for them?
Jini is taking the same path as Java. Java first had visibility on devices like set-top boxes. It really gained traction and ended up back on devices with J2ME and the like. Jini is doing the same thing. The sort of on-the-fly discovery and contract negotiations turn out to be terribly important to people who are building complex systems in the data center. Web services have a way too coarse-grained interface to think about using for some infrastructure applications. It's way too creaky for that from a management point of view. Jini has been in front of that. N1 is drawing heavily upon it.

Overall, what is Sun's commitment to R&D spending?
As far as R&D in the company, one of the things that has kept our stock price down and has depressed our earnings is that we refuse to cut our R&D spending because we think there is a lot of innovation that has to take place to make this new model happen. We feel like at this point we are so far out in having pushed people into having to try to digest things that are disruptive from a software model point of view, that we have deliberately slowed down to let people catch up a little bit. We have a lot of other stuff coming up, and people just want to get their stuff to work. So there has been a deliberate pragmatism about that.

There's so much going on in the industry right now. I think the biggest piece of crap going around right now is that this is a mature industry and innovation doesn't matter.

Well, that's Larry Ellison's argument, isn't it?
Larry has been busy trying to acquire more assets and ERP systems. Maybe there isn't a lot left in that area. You have to look at these frontiers where the demand is. And he is in the position of trying to turn demand into Oracle database sales. The whole software industry has to deal with this reset of what open source is going to mean to create that. That's the big story.

For me, Sun has never made money on software; I make money on systems. So I'm not going after my own children if I go in and try to make the software business more efficient. We know that has to happen. But to my mind, the way you stifle innovation is that you go to a utility computing model and you outsource all of the people.

The drug for the industry has been this binary software lock.
So what's your thinking on utility computing? How close are we to the vision? What does the vision even look like?
There are multiple ways to look at this. There is utility from a financial model. That is, who is going to own title to equipment and whether I do this as an on-demand model or something else. (It's) the difference between purchasing and a lease. That's not a technology model. That is really a financial model that is taking the old view, the current view, of computing and trying to make it more fluid and automated.

That's a very different thing than asking where is computing going to go. The fundamental change will be that we are not going to have these purpose-built systems (so that someone would need to say) I want to deploy that app server, so I'll get an Oracle database there, and this ERP system and run it here, and hand it over to the utility guys and say, please make more of these. The new model is that I just have computing stuff. There's a computing fabric, it's a set of resources that I can throw computation and storage at, and it responds efficiently, just like when I throw a job on my PC. It responds to that and it allocates resources to go run it.

So those are two very different views. It's the financial model of delivering what we have and pretending that you are automating that, versus really doing computing at network scale. Re-plumbing and virtualizing existing systems ain't doing computing at a network scale. You'll know it when you see it, because the systems themselves will be stripped of their identities.

Where does Sun fit into that picture?
Network computing stuff. If you believe in the growth of the network and the growth in demand for services on the network, we're squarely about how you provision those services and that is certainly in the context of how people think about data centers today. What's the kit--both hardware and software--that you need to provide the foundations of a service delivery platform? And that is network computing. From a bigger perspective, it's provisioning all of the way out through the network, because applications now encompass everything from handhelds to big storage systems. So why have we been so dogged about Java on cell phones and getting a few hundred million of those out in the world? I want to understand the consequence of having that point-of-service demand out there and what is the container for that service and how to provision it.

So, how do you differentiate in that world? A lot of the things we try to accomplish--and we have been very successful with the Java platform--is to reduce the artificial switching costs that have dominated IT. We have very high switching costs and that's what results in monopolistic rents from companies like Microsoft. We have gone through this.

That's interesting to hear from Sun. I always thought of Sun workstations as sort of the Macintosh idea, but only on Unix. You have Sun workstations, disk drives, storage, other hardware and software, and it all comes from Sun. So, the switching costs there were substantial.
The drug for the industry has been this binary software lock. That's really the root of it. I attract someone to my particular binding of an operating system and microprocessor. It happened in the past because people write at the source code level but deliver applications out to customers as a binary, so all of the verification was in the binary. And once I've got your binary, I've gotcha.

If you look at the new expression in software, you write to an application server or a Web server or a database. There's no binary lock. If I deliver directory services or JavaBeans or server pages, it doesn't say SPARC or x86, and it doesn't say Solaris or Linux. None of those APIs are exposed either. It's just the set of contracts that say this is what a server page looks like, this is what a directory entry look like. And this is huge. I don't think the industry has caught up with the fact that this fundamentally means that vendors don't get to lock people into your binary cycle the way that you used to. I think Sun is ahead of the curve in realizing that.

So where is the lock-in, then?
The switching costs are the things that make it hard for people to move around. That adds friction, and that's what keeps people stuck. But the cost of computing is not that. The cost of computing is, in fact, the management of this stuff. That's where 80 percent of the dollars are going. It is the fact that once you get just beyond the component level, the assembly (of applications) is an exercise left to the user. It's not engineered. So you can ask a basic question, like, we're going to build computers from networks. We have all of these server and storage components. What's the operating system we have? What's the thing that figures out how to connect and manage all of this stuff? That's N1, and I'm going to off and invent that stuff, and I can go off and run a business model on the development side like Orion, and I can do to software what Dell did to hardware. I can make lots of money doing that.

Sun has never made money on software; I make money on systems.
So that's your differentiation in the market?
Yes. The business has got to be driven by great product. You have to be competitive. You live and die by it. We see a lot of two-way Xeons and Linux out there. It's really been a sweet point in the market. A lot of people have moved to that because it's cost-competitive and the switching cost hasn't been there. You can run Apache on it, you can run Apache on Solaris, it doesn't matter. If you get to the next level out, not the developer trying to do it but the administrator trying to deploy it, I want them to care. The more stuff you have under N1 management, you are going to get a network effect out of those resources. You can incrementally add resources to your network, storage, processing, etc., and it won't add a linear increase in management costs. If you are spending 80 percent on your management, I can drive a lot of efficiencies out of that. That's where the services industries?-IBM Global Services, HP, etc.?-add to your management burden. I am more than happy to take dollars out of their pockets.

So you are saying you can deliver computing to the IS department for less money, and actually help to shrink the IS department. So that means their budgets will shrink. So how is that good for Sun and the industry? Won't they ultimately have less to spend? How does Sun generate continuing profits from that scenario?
The question is whether you believe that actually happens. If I could sell you a machine that is 10 times faster than current machines, I will take market share. Or, you can turn that around and say, "Well, it's 10 times faster, so people will buy a tenth as much computing and that there is a fixed amount of computing that needs to get done." That's where that view falls apart.

The question is whether you think that is true across the horizon of information technology. And that is patently false. One of the things that drove sales of Sun systems in the 1990s was that it was hard to create computing systems big enough for companies to go do something like a single ERP instance. That was really testing your mettle to do something like that. And now, that's standard practice. We know how to do it. You have this exponential improvement in computing, but ERP is not getting exponentially complex. So you would certainly expect it to get cheaper. If you went to a CFO and said, "I can give you a million times more computing power," they would say, "OK, I'm not going to close my books every millisecond, so you are going to give it to me at a millionth of the cost." But if you go to an R&D department or to Wall Street, they will say, "Please, bring it on." I think we can actually free-up dollars from other parts of the business. They would definitely see a competitive return doing better simulation of product development, for instance.

One thing that always confuses me is how Scott McNealy constantly criticizes Microsoft. But where do Sun and Microsoft really overlap? It's not 100 percent clear.
Software development, really. Once that decision has been made, then we don't clash. If you say you are going to do Microsoft development, then I don't get to be in that equation. We accept that we can't compete in that equation. The biggest issue we have to deal with is that there is a monopolistic business at the center of that--there is a billion dollars a month of cash associated with that. You want to be very careful about them using that monopoly to build a new one.

There are ways that that happens, taking interfaces on a desktop system and linking them to a given product, like in streaming media. You have to have Windows servers to drive Windows streaming media. The thing that is on the device is Media Player. Nobody else is allowed to drive that. That's taking the monopoly on the desktop and moving it out. I cannot compete with that. A far better state for the world is that it's on the wire, it's open, we have standards and you serve media, not just Windows media. And you can compete on that. That's where the clash is.

Does Web services change that equation?
I think Web services is probably the best level of agreed-upon interoperability we've had. There's only two stacks now--.Net and Java. The world should have two, for competitive reasons. To the extent that we coexist so I can't tell what's on the other side of the wire, that's a beautiful state to be in.