Larry Ellison unabridged interview with GCN
Larry Ellison unabridged interview with GCN
Larry Ellison unabridged: National databases, middleware, and whether programmers wear blue pajamas.
During the recent E-gov conference in Washington, Larry Ellison, chairman and chief executive officer of Oracle Corp., sat down with GCN editorial director Thomas R. Temin. An edited transcript of the interview appeared in the July 30, 2001 issue of GCN. Here is the unedited transcript in which Ellison expounds on everything from national databases to whether programmers have dates. Also in the room were Kevin Fitzgerald, head of Oracle's federal operation, members of Oracle's public relations staff, and reporter Joab Jackson from Washington Technology.TEMIN: A lot federal research is going into how to tie disparate databases with middleware into larger virtual databases. Have they got it all wrong?ELLISON:
I think the ideal solution is to have these large central databases with lots of applications publishing information into and subscribing information from them. But that's not what we have today. So how do we get from each application with its own private database to lots of applications sharing information from one big database?
What we can do is, our approach is to-you might think of it almost as an active data warehouse.
When, to take a medical records example, the blood lab goes ahead and analyzes your blood that they not only put the results into the laboratory's database, they also take your results from you and store in the national database as well. So they store that information in a local database that's application-specific and store it in a global database, so that everyone knows where to find that information, so everyone who is authorized to see that information can in fact see that information.
So we gradually evolve all these separate systems-they just can't shut down, we can't magically turn them off-and we start collecting information from these systems. Not only do we collect information from these systems, but then as those applications get passed you're not only putting information into this large national database you're also looking for information. So it's working both ways. Gradually the application becomes less and less dependent on its local database and more and more reliant on the central database, until the local database just disappears.
Then you get what we think is an idealized architecture, which is you get all the [data] in one place with all the applications surrounding it. The great thing about that is it's based on SQL standards, XML standards, build applications in HTML and Java, and it's all standards-based.
But lots of people are building applications on top of that database. It's not any one company that's doing all the work. We tackle the whole issue of medical record-keeping, whether it's inside a hospital or for the entire country.
We've gone to enlist partners to build this system. GE, which builds magnetic resonance imaging systems, they have to store the data not only on this local diskette, the MRI machine, but they would also have to store this data into the national health records database. The same thing for the automated blood laboratories. Every time a prescription is filled, they store the information not just in the Rite Aid database, they have to store it in the national database.TEMIN: If such a database were to exist literally, it would contain trillions of data elements.ELLISON:
Let's take a real example. TRW provides a service for something that works exactly like this. But this is something that must be much more important and health care-it's credit. [Laughter] So we can't store your health records, but we do know if you can pay. So there we have a national system. And TRW has one that keeps track of all your credit records and they do it as a service for Visa and all sorts of other people.
So hypothetically Oracle could provide this as an online service. It doesn't have to be a government-run agency that does this. They can be privatized like these centralized credit services, except that we're keeping track of health records. And we think we can do a better job of managing these large central databases than anyone else in the world.TEMIN: IBM recently got a contract through the IRS's Prime program to start migrating its Tape Master File System to a relational database. Are you going to try and get a piece of that?FITZGERALD:
They are looking to have databases at various levels within the agency and we are doing a lot of the work inside the IRS now in terms of databases.TEMIN: But are you going to try to get a piece of this Customer Account Data Engine contract? Presumably they would use DB/2, but maybe not.FITZGERALD:
We're going to try, but there are a lot of issues inside the agency. Clearly we're very active and, by the way, the agency wants us to participate.ELLISON:
To give you a corollary, IBM just picked up a huge contract to manage Shell Oil's data centers all over the world. But at the same time Shell Oil announced that their worldwide database standard was Oracle. It is not uncommon. I mean, IBM works much more often with Oracle than it does with DB/2. In fact, when IBM wanted to show off their fastest hardware, they used Oracle software, not their own software.JACKSON: Service is becoming an increasingly large share of Oracle's business'ELLISON:
Wrong. If you look at our consulting business for the last three years, it's been going down. We have two services. We have a subscription service where we give you new versions of the database, but that isn't really a service but more like a license to renew with new versions of Oracle. That business is growing.
But labor, human labor as a business at Oracle has been declining for the last three years and will continue to decline. We believe that our whole job is to make application systems and our database less labor-intensive.
For example, one of the classic rules of thumb, say if you're buying a system from Siebel-IBM is a reseller of Siebel, and for every dollar of Siebel software they sell they like to collect $10 for putting it in. So you've got a 10-to-one ratio of labor to software. We think that ratio should be one-to-one. So we are putting tremendous pressure on our own consulting group that if they're doing an installation for a million dollars worth of software, it shouldn't cost $10 million like IBM, it should cost $1 million.
So we're trying to change the nature of the software business and make it much less labor-intensive. That's a huge counterpoint to us and IBM, a whole different view of the world. Lou Gerstner has taken IBM to be must more of a service business. We don't want to be in the human labor business. We want to be in the software business. FITZGERALD:
It's intuitively obvious to put integration in during the engineering phase, not during the implementation phase.JACKSON: So do you see yourself competing with the systems integration market?ELLISON:
No. In fact, our most important partners are the integrators. Our emphasis with our own service organization is very rapid selections with very low cost. That's the innovation we're doing in service, which is part of our war on complexity. Most problems occur through human error. Whether you're a pilot or a programmer, it doesn't make any difference. Most of the problems are human error. The less labor you have the lower the cost. The less labor you have the lower the cost. So it's all about getting the labor out.TEMIN: The federal government has been touting FirstGov, the portal, as a seamless way to get to government information. But portals are mainly collections of links, not truly transformational, transactional systems. How do agencies make the leap?ELLISON:
I think the portal is important because it at least helps you find things. I think integration of information from a presentation standpoint via a portal is very important. We sell special tools for doing just that.
I think in the end what you want is information integration not just in terms of its presentation in a portal but actually in its storage. That's where you get back to the theme of information shouldn't just come together when you see it but that information should be stored together because then you can relate one piece of information to another.TEMIN: How do you sell agencies on the migration costs from multiple databases to one? Middleware seems like a cheaper option.ELLISON:
Well, no, actually just the opposite is true. It's cheaper, even in the near term. Throwing in more middleware A) is very expensive-just look at the IT budget and how much is spent on labor. Look at the size of these contracts and they're all labor. If it were just a matter of writing a check for middleware and it's working, that would be one thing. But it's a little bitty check for middleware and a hundred to one on labor to use the middleware to connect everything up.
The middleware doesn't really do the job. It's a little like saying, 'Gee, I bought the lumber for my house and it's done.'TEMIN: Well, what do you do if you've got all these databases with different schema if you want to put them in Oracle or any database?ELLISON:
We have a variety of gateways and tools to allow you to take information out your database and put it in an Oracle database. So you've got all these systems, all these satellite systems and we're saying it's not terribly unlike a data warehousing strategy. You've got all these satellite systems, all these separate databases. Just starting writing the programs to move that data out of those IMS databases, off of those mag tapes at IRS and start populating an Oracle database.
Once you have the database built, once you have current information, you start writing the new applications against that database. And that's a much simpler approach than the lumber, this middleware. I've been in this business for a long time, and I think we're the world's largest supplier of middleware. Trust me, this is not solving any problems any more. I suppose, taking the equivalent of if you're working on a car, the middleware is a little bit like a wrench. It makes it easier to attach things. Let's way we greatly improve the wrench and make it a socket wrench. Put a ratchet on it-a better wrench. But it does not fix the car. It helps the labor a bit.TEMIN: Discuss thin client computing as it's evolved over the last year or so. Now it seems to be cell phones, handhelds and so on.ELLISON:
In terms of everything we do, the PC is a thin client. All thin client means is the only application you have on your PC is a browser.
That's also true of a tank, a Palm, a cell phone. Every Oracle application is visible via a browser and only visible via a browser. So we only support thin clients. We treat your PC as if it were a thin client. The only way you can access an Oracle application, the only way you can access Oracle data is through a browser.TEMIN: Have you declared client-server officially dead?ELLISON:
We declared it dead for Oracle in the mid-'90s but now everyone has joined us. I don't know of anyone who doesn't say client-server is dead. SAP says client-server is dead. Siebel says client-server is dead. PeopleSoft says client-server is dead. No one is fighting the client-server war. Even Microsoft has done that with .Net. Microsoft is the last to abandon client-server. When they abandon client-server, it's dead.
It wasn't me who declared it dead. Since the mid-1990s I thought it was a very bad idea, a horrible idea as a distributed complexity model. The idea we could put applications on millions of desktop computers is a form of madness.
It's not unlike the form of madness we have now, which is taking all our data and putting it in lots of little databases everywhere. It's the same lunacy. You don't want to distribute your software. You want your software on servers. You don't want a lot of little databases, you want a few large databases. Fewer databases are easier to manage and you get better information. Costs less, better data.TEMIN: Another government topic. Government is supposed to buy COTS, commercial off-the-shelf. So they either spend 50 percent more modifying it, or vendors like Oracle create government editions of applications. Are those really COTS?ELLISON:
Well, is a French version still COTS? The government version and the commercial version probably have 90 percent in common. Just like we adopt our accounting software for Italy, we adopt our accounting software for the government. [FITZGERALD:
The military Sealift Command took our software and instead of customizing to it their unique work processes, they said, 'We gonna change our processes to do what the software does.' First government agency to do that, and in half the time, quarter of the cost.]
Commercial companies make the same mistake. One company spend four and a half years to do an installation, making heavy modification to the software. Madness. So the simple notion is who should finish our software? Should we finish our software or should our customers? We think it's a bad idea for our customers to finish our software. It's hard enough for us to finish our sofware, and it's our business-software.
So should government finish our software or let us finish our software? Government shouldn't ask the question, 'How have I done business for the last 50 years and say, Oracle change your software to the way I've done business for the last 50 years. That's a very bad approach.
The question is how do I want to do business for the next 50 years with this new technology called the Internet. And can I do a better job. I need to look at every process, every part of my organization, simplify it, standardize it-and that's what our software does. And you can do that without modifying our software.
Eighty five percent of our implementation right now are done without modifications.TEMIN: Government takes a lot of hits for large-scale systems deployment failures. Is government that much worse than the private sector, or are its mistakes simply more public?ELLISON:
That's an interesting question. Private industry is pretty bad. I think it's very competitive. [Pauses to think.] The government has had some incredible failures. I mean, the air traffic control is one of the more amazing ones, the IRS. But there are some pretty astonishing systems commercial systems going in, where people spend $2 billion installing SAP. Two billion to install an accounting system-unbelievable.
So, no, I think government has to share the [spotlight]. If they handed out medals, I don't think it would be all government winners.TEMIN: If you were the federal chief information officer, what would you do?ELLISON:
I would have several initiatives. I would talk about having an information architecture for the government. And talking about, we need to stop buying applications and start managing information.
A national health records database would be one initiative. A national public safety database would be another one, where we would be providing service to state and local police agencies.
I would put tremendous pressure on the intelligence community for collecting and story intelligence information. I might increase the charter of one of those agencies and give it responsibility for managing this very large database, gathering the data and data analysis. I would have these super-agencies'gather information on behalf of several agencies.
The size of the savings targets would be staggering. Intelligence would be a very high priority of mine, national defense, public safety, health records.TEMIN: Everyone has bet on the Internet, Java, SQL and the other standards. Is there anything you worry about at night that could blow it all up?ELLISON:
No, I really believe this is kind of it. The reason I believe this is the last architecture is for the first time the computer network looks like every other essential network in the world. All networks look exactly the same, whether it's the water network, the telephone network, the television networks. They're all built the same way with simple end user devices.
We gotten all the applications off the PC. The only application on your PC is the browser. Everything's on servers, where it's professionally managed, so all the complexity is hidden from the end user.
The network is based on simple standards; all the networks interoperate. You can have all sorts of non-PC devices; you can add any sort of device to the network you want to-televisions, tanks, cars, On-Star. All this stuff is being added.TEMIN: On-Star, that's true luxury compared to on-dash systems with the maps where you fiddle with all the buttons. It's easier to push a button and ask Sally.ELLISON:
Well, that's the television commercial. It's a very strange commercial'I'm going to call someone on the phone, my American Express special agent, and ask, "Where am I?" [Laughter]
But the worst commercials are IBM. You hear these Turkish guys talking, and then this girl is talking back to them. Except, there is only one problem: They can't do that. [Laughter] Maybe the Codenauts brought that over from Mars. They brought the software that converts Turkish to English and back and forth. I mean, they're very good commercials. IBM is advertising things they can't do. It's "coming"-they say it's "coming" from IBM. When? In the year 5000? It doesn't say "coming soon." [Laughter]
Then they show guys from other planets. Is there something I'm missing? Are there people from other planets here now? I realize they have to try and sell DB/2 here on earth. Actually it's very clever-"Let's sell DB/2 to guys from Mars." They can't even sell DB/2 to their hardware guys across the hall.
But these guys with the funny suits-the think I find extremely annoying, actually, is that's what IBM thinks programmers look like. I'm a programmer. IBM thinks programmers dress up in blue pajamas and plastic bubble hats [laughter] and walk around calling each other "meatball." You know? They don't think we go out on dates. That's very annoying. It's not true. Programmers date!