CNET también está disponible en español.

Ir a español

Don't show this again

Christmas Gift Guide
Tech Industry

Oracle's 'Mr. Linux'

Ever since Larry Ellison gave him the green light, Wim Coekaerts has been working on putting together a Linux strategy for the big database maker.

When Larry Ellison asked him to design a Linux-based Internet appliance, Wim Coekaerts reported back with a plan a couple of weeks later--and won himself a new task.

Coekaerts' quick turnaround was enough to convince Oracle's duly impressed CEO to charge the soft-spoken Belgian native with helping to figure out a Linux strategy for the software company.

If that job seemed a big one a couple of years ago, it has since grown in importance with more large corporations starting to adopt Linux in their data centers.

Oracle has never before controlled its own operating system source code, and the collaborative development process allows the company to prototype applications on the operating system without waiting for an outside vendor to respond to specific feature requests.

But Linux also poses a challenge. While Oracle has a clear self-interest promoting its work with Linux, much as has IBM over the last couple of years, it needs to perform a diplomatic balancing act and stay in the good graces of the open-source community, as well as that of Linus Torvalds.

CNET News.com recently sat down to learn about Oracle's strategy from the man known inside the company as "Mr. Linux."

Q: Is your group set up separately so as to avoid any interaction with the app groups at Oracle?
A: There's lots of interaction but we don't want the group to be part of one product division. Oracle products need to work with Linux, and so we give guidance to everyone. But we don't want to be in the middle of developers working on Oracle products. All of these projects are close sourced, while everything we do on the Linux side is open. We don't want to have closed source binary drivers and stuff like that

Isn't that a contradiction?
It's a nice arrangement. We want SuSE, and Red Hat and United Linux to use all that stuff so that their next versions run better on Oracle.

Do you have any kernel engineering groups that deal with other operating systems?
Oracle has traditionally worked with the other companies because we didn't have the source code. The nice thing about Linux is that we can now prototype on the operating system side and on Oracle at the same time. In the past, you would go to whomever and say, "We want to work on this feature, and we need this from the OS." Then they'd go back and have meetings, discuss how long it would take and see if they had the resources, and so on. It turned into a time delay.

With Linux, someone from the database group can come over and see if something will work. Even if it's not something that's guaranteed to work well yet and may take a few weeks, we can play around with it and do tests. That's a big advantage. You can do what you want with it.

What happens if a Linux user who also is an Oracle customer runs into a problem? Who fields the call?
They can call Oracle support, and we'll help them if the operating system breaks down. If a big customer is running the nightly batch jobs and the kernel keeps crashing, my group carries pagers and they'll fix that bug or create a work-around.

Has that happened?
It hasn't happened yet--which is good.

Does that suggest operating systems expertise will become one of Oracle's core competencies? Do you really want to have kernel engineers becoming an essential part of your business?
If a customer has an issue with Linux today, they call into Oracle support, and if it's something simple, we can help them out. If there is an actual bug in the operating system, but not critical, then we

There are projects that will feature things that no other operating systems have had--and it will happen on Linux.
internally funnel it through to Red Hat. We're not doing a Linux distribution. We're just making sure the customer has this safety valve. If it's a really critical issue, there's no time to play pingpong between vendors. We'll fix it, and once that's done, the fix goes to Red Hat.

But when you sit down with big customer, how does that play out? Microsoft claims it has a better story than Linux, as well as organized support when something goes bump in the night. Your story sounds a little more complicated.
I don't think so. Customers need to have a single point of contact, and when I say we also talk to Red Hat, that happens within Oracle. Our support desk does not tell the customer, "This is not our problem. Go talk to someone else."

How many people do you have working on Linux at Oracle?
If you talk about Linux kernel stuff, there are about 1,000 people that actually do development work. It's been that way for a long time, but we just have not been very public about it. Linux is Unix. When you have a lot of Unix competency in your company, it's really very easy to switch and doesn't take too long.

A lot people say the development of Linux is taking place faster than the development of prior operating systems. Some say that's because of the nature of the open-development process, while others say that's because Linux is only turning over ground that's already been trod by others. Do you agree? Is Linux maturing faster than other operating systems?
Yes, it is.

By how much?
The nice thing is that you have multiple companies working on it. You have Red Hat's kernel team working on functionality; you have SuSE's team doing stuff; you have other people not associated with distributors working on features. So you have this huge development process that's taking place in parallel. Part of it is Linux growing up and working on things that other operating systems have had in the past. But there are projects that will feature things that no other operating systems have had--and it will happen on Linux.

One criticism leveled at Linux is that it's reproducing things already there. In other words, that the open-source community is simply good at cloning existing technology. Is that a fair criticism? Is the Linux community creating anything original.
They are. In fact, Alan Cox (Editor's Note: Cox is a Red Hat employee and one of Linus Torvalds' key deputies) has been trying to get people to write RFCs and come up with a standards proposals because of stuff that's in Linux that you don't find in other operating systems. Yes, there are areas where it's been lax but there are a lot of new things that are coming up.

Is the standards evolution process in need of rethinking? In order to get a new kernel out the door, you need Linus Torvalds' imprimatur. Isn't that an impediment?
Not really. If you look at how the distributors work, even if there's a little time delay, they can still ship patches in time to get into the main kernel. It's a good thing (Linus) is who he is. He's very much in control. The decisions are technically made, not political. It's not because he likes one vendor better than another. That means he's quite predictable.

It's a good thing (Linus) is who he is. He's very much in control.

But he's one person and this becomes a movement based on his yea or nay.
It mostly is but it's open. He's always been very clear about saying, "If you don't like the way I do it, things are open, you're more than welcome to change it and add stuff that will get into the kernel." I haven't seen any real issues and don't foresee any. It's good to have one person in control.

Is Linus keeping up with the learning curve? A few years ago, he said, "Multiprocessing? Why would anyone bother?" Now he obviously recognizes it's real.
He sure is. He's an incredibly intelligent person.

What are your biggest disagreements with him?
I don't have any actual disagreements with him.

Do you think Red Hat will become the de-facto choice for server Linux distributions?
Red Hat is big in the U.S.; SuSE is big in Europe. You can't choose one.

Red Hat and Dell Computer argue that anything bigger than an eight-way machine is doomed to niche status. Do you believe clustered databases are the way customers will do big databases on Linux?
Not just on Linux.

But particularly with Linux because it doesn't scale. You can buy a 64-processor server that runs Solaris. When Dell, for instance, says you're never going to need anything bigger than an eight-way server, maybe that's because, well, gee, that's because Dell can't sell anything bigger than an eight-way server. It's hard to separate the technology from the marketing.
But this is the important part: Intel CPUs are really fast. They have 2 to 3GHz CPUs out there today, so if anything, we're mainly bound by memory. You can run a lot on a two-CPU box with 12 gigs of memory. That means you don't need a 16-way Intel box to replace a 16-way Sun machine; you might be able to do it with an eight-way system.

Let me ask the question another way: If you have a really high workload, is a 9i RAC (Real Application Clusters) on a bunch of two- or four-way boxes just as good as buying a big iron box?
I think in many cases it will be. Of course, there are going to be exceptions.

What does Oracle want from Linux?
Customers want Linux. It's a neutral operating system, and everyone likes that. As an operating system, Linux also gives us a chance to provide stuff that our competitors do. It lets us prototype and it runs well on cheap hardware. Also, Linux is making for a lot of cost savings--up to 20 percent--and ultimately, it will help our margins go up.

Are there profit differences between selling on Linux compared with other operating systems?
Traditionally every OS is the same in that you can download whatever you want if you have a license.

Has it made Intel servers more interesting to Oracle?
I think the answer is yes. It's all about cost. We're finally at a point where cheaper hardware can do the same thing. And Linux is all over the place.