Oracle’s biggest announcement at its recent user conference was the release of its 10g database and Application Server, which will harness the power of several servers for large computing tasks. Shortly after his keynote presentation, Oracle chief executive Larry Ellison met with a select group of

international media, including Computing Canada, where he outlined the company’s plans in more detail.

Question: People have been clustering computers into grid-like formations since the VAX. Why is Oracle getting involved now?

Ellison: We started the clustering quest almost 14 years ago. We built Oracle 6, which almost bankrupted the company — in 1990, 1991, that’s the foundation of the system on which clustering was built, but we never got it completely working. Every release has gotten little bit better — 7 got better, 8 got better. Now, with 10g, we have something that can cluster lots and lots of machines, but we’ve included the management tools that will go with it.

Question: How long will it take for grid computing to make real inroads into the enterprise?

Ellison: When the relational database came out it was clear to me that it was superior technology to the databases that preceded it. But it took years and years, and gradual cycles and new applications. Now this should be easier, because you can use existing applications to the grid. So it shouldn’t take as long as the original relational database. But still, it’s not something that’s going to happen overnight, but it’s going to be a natural move from where we are today. The economics are compelling, the reliability is compelling.

Question: As grid computing spreads, what will happen with Sun’s Sparc/Solaris machines?

Ellison: I find it very competitive across the low-end machines. Sun is also selling Intel/Linux machines. I don’t know, exactly. I don’t think these large machines are going to disappear overnight. There’s still a big business selling large machines. I’m just saying that this begins a trend, where we’re going to see a gradual movement from the one big server approach to the grid approach. It’s going to happen over time, and there’s going to be plenty of time for Sun to react and HP and IBM to react.

Question: People at IBM would say they have these huge servers with 1,000 processors. The most a RAC server can do is maybe eight processors?

Ellison: We scale the application server, we scale the database server, we scale the storage grid — we virtualize all of it. We’re the only ones who can solve that problem; we think that is our problem. Let me very clear about this: SAP is Oracle’s No. 1 competitor. SAP certifies Oracle Real Application Clusters. SAP does not certify any IBM clusters . . . SAP, our No. 1 application competitor says, “”Oracle, your clusters work, we’ll certify them. IBM, your clusters don’t work, we will not certify IBM clusters. We will not certify IBM clusters.”” That’s because their grids don’t work.

Question: What will grid computing mean for your outsourcing business? Will you be setting up data centres elsewhere?

Ellison: The answer is yes, we’re doing it already. We are using primarily the Wintel boxes in way more than 90 per cent of our customers in the outsourcing business. The data centre thing is very interesting, where they tend to be located. Is there such a thing as the EuroNet? Why would we have a data centre in Europe? Texas, where our data centre is, is cheaper than Australia or England. Land’s cheaper, electricity is cheaper, computers are cheaper. I have this battle in my own company all the time. They’ll say, “”Well, if we’re going to sell to Europe . . . If we’re going to sell to the Hague, we need to have a Hague data centre.”” There’s no Haguenet, there’s no Euronet, there’s the Internet, and who cares where the computer is? Why does the computer need to be in Europe or Asia Pacific? To me it’s one of the fascinating holdovers of our old-style thinking for about how the world works. We can have DBAs in India running computers in Texas and it works just fine.

Question: Why is Oracle not a more visible presence in the grid standards organizations?

Ellison: Because the grids we’re trying to build are enterprise grids that are running SAP applications. We’re not trying to build scientific grids. People who are looking for life in other parts of the universe are on a fascinating adventure, but just not one that’s central to our mission. We’re worried about more mundane things, such as running Siebel more efficiently at a lower costs.

Question: Does grid require a complete change from traditional centralized computing strategies to a more distributed one?

Ellison: Well, no. We’re saying keep all your customers in the single database. You decide where you keep it. You shouldn’t chop it up into different pieces and put the Dutch customer database in Holland, the French customer database in France and the German customer database in (Germany). If you mean distributed data, then no, we don’t believe in that in all. We believe in having information agents made up of a small number of large databases with all your customer information in one place. But what we’d like you to do is apply tremendous computing power to that huge database. And there we want to assemble a mondo-server made up of lots of standard parts . . . we’re taking a bunch of separate components and marshalling them to work together.

Question: Ultimately, how will grid computing change the enterprise?

Ellison: We’ll go from all of these trees in the forest to these huge, information-age systems. Once we can build these highly reliable, very large grids, we now can take all of the eggs, the customer data, and put them in one basket. And the quality of information that businesses will have will change radically. The companies that take of advantage of this will be much more competitive . . . companies today have hundreds, even thousands of separate databases. It has their customer information, their marketing information, their sales information. They can’t even see the trees. All they can see are leaves.

Share on LinkedIn Share with Google+