Wind River's Linux transformation

It's been a slow, hesitant embrace but now open source is top-of-mind for CEO Ken Klein.

In the past two years, Wind River Systems got over its scorn for Linux. Now it's up to Ken Klein, chief executive for 15 months, to do more than embrace the open-source operating system: He's got to find a way to profit from it.

Wind River specializes in embedded software used in computing devices such as DVD players and nerve gas detectors. The company has historically sold its VxWorks operating system for these tasks, but it now also pushes Linux.

"It's brain surgery, but we're doing it with a micro-scalpel," Klein said of the transformation. Linux revenue remains small so far, and there have been some slipups along the way: Klein had to revamp a partnership with Linux leader Red Hat, choosing instead to build its own version of Linux for most products.

But some things are clearly going right: The company turned an $8 million profit after four years in the red as revenue rose to $236 million. The company expects revenue for its next fiscal year somewhere between $265 million and $270 million.

Wind River quietly began embracing Linux in October 2003, when the company started releasing programming tools that could be used to build software for the open-source operating system as well as VxWorks. That same year, Klein took control of the company, replacing co-founder Jerry Fiddler as chairman and Tom St. Dennis as CEO. Since then, the company has released its first version of Linux, tailored for networking equipment, and another is due out soon.

Klein, 45, is relaxed and optimistic, saying he has banned cynicism and negativity at the company. But he's also willing to use an iron fist, replacing two-thirds of the managers in the two ranks below him. He talked with CNET earlier this month at the company's headquarters in Alameda, Calif.

Q: What has Wind River been, and what is it becoming?
Klein: We were closed and narrow in terms of our partnerships. We were taking a very adversarial approach toward Linux. We've turned 180 degrees. We're viewing Linux as incremental to our business. In set-top boxes, Linux is a great fit.

There's another area of change. Partnerships with semiconductor manufacturers before were relationships of convenience. Now customers are sick of semiconductor suppliers and software suppliers not playing very well in the sandbox together. They want a deep ecosystem of partnerships with the semiconductor manufacturers--MIPS, Freescale, ARM, Philips, Broadcom, Texas Instruments.

Embedded software wasn't viewed as being very important. But it's changing.

Why did you make the switch to Linux?
Klein: We responded to the demands of our customers. Customers, particularly in the telecommunications marketplace and to a lesser extent in consumer electronics, were making a beeline to Linux. Our customers have an aversion to putting Microsoft in any of their devices. They're not interested in seeing Microsoft usurp value from their devices. The availability of talent and applications and the fact that Linux is free from a runtime perspective (Editor's note: Device manufacturers don't have to pay per-unit fees for using Linux)--those factors created demand.

The question was: Do we want to continue to ride the VxWorks horse or expand the range of possibilities by embracing Linux? Our customers were asking for help. Many experimented or tinkered with Linux, but it's not free. There are time-to-market costs. It takes awhile to test it and maintain it and support it. It was a clear mandate to us that someone needed to help them with that.

So if the customers don't want to hand control to Microsoft, why would they want to do that with Wind River?
Klein: You want a supplier with a stable platform and product development environment, but (one) that is not going to compete with you, a la Microsoft's entry in the set-top box market. They're not just interested in being a software supplier; they're interested in taking over the business altogether.

Why are you trying to replace the term "embedded software" with "device software"?
Klein: Embedded software was incidental and invisible. It was hidden. When you call it device software, you realize it's becoming the brains of the product. You're talking about a much more strategic opportunity.

There's a shift in the marketplace. The time-to-market windows are shrinking, down to nine months from 12 recently. That's shifted customers toward buying software instead of building it. The average OEM (original equipment manufacturer)--Siemens, Sony, Nortel, Motorola--they're spending 62 percent of engineering money on software. Whether they like it or not, they are in the device software business. It's software that's becoming the key differentiator, not the form factor, not the color of the plastic.

We see Linux bordering on becoming half the business probably in three to five years.

So does the term show a power grab? A move to give Wind River more prominence?
Klein: People didn't care about the embedded industry. Embedded software wasn't viewed as being very important. But it's changing. In five years, there will be 14 billion intelligent, connected devices. Software is going to make these devices intelligent. It's the iceberg principle. What "device software optimization" does is it shines a light under the water.

And the finances are looking up?
Klein: We had our first profitable year in five years. We had the highest cash flow from operations in five years.

What are your ambitions?
Klein: I see no reason why Wind River can't get to the next stage, which is a $500 million (annual revenue) company with 20 percent operating margins. That's goal No 1: to get from a $250 million company to a half-billion-dollar company on a subscription-based business model.

What's the revenue split between Linux and VxWorks?
Klein: Linux is still less than 10 percent. We're still in the really early days. The Linux revenue is dampened because it's sold as subscription. There's no product license fee or royalty. It'll roll in slowly over time.

How fast is the cutover to Linux happening?
Klein: We see Linux bordering on becoming half the business probably in three to five years.

What are the complexities of Linux in the embedded market?
Klein: It's very fragmented. In servers, one processor architecture matters. In the embedded market, there are about 60 that matter. That complexity is what I call the matrix of pain.

Are there intellectual-property issues with Linux? Your predecessors briefly tried to promote the open-source BSD version of Unix, arguing that its advantage over Linux was that customers could make it proprietary.
Klein: It's always on your radar. We have a lot of expertise on the issue of GPL (the General Public License that governs Linux and prevents it from being made proprietary and prohibits it from being tightly intermingled with proprietary software). What we've done is taken a developer suite and wrapped that around an optimized distribution with middleware that's been selected and tuned to address specific vertical markets. There's a clear demarcation in where the operating system stops and the developer environment begins.

Is the GPL a problem sometimes?
Klein: Where Linux is a fit, customers really have been embracing it. There is a certain set of customers in Asia concerned about GPL leakage. For them we have VxWorks.

By "leakage," do you mean the so-called viral issue around the GPL (which requires software that's tightly integrated GPL also be released under the GPL)?
Klein: Yes. They're afraid their proprietary middleware will seep back and be governed by GPL.

Is that a risk?
Klein: I think that's a perceived risk. Depending on how deep they are into the Linux kernel (the part governed by the GPL), it could be an actual risk. Most customers aren't operating at the kernel level. There are some borderline cases.

Do you use your own version of Linux, or is it Red Hat-based?
We start with the kernel at This is a parallel strategy to what we're doing with Red Hat. When the market requires a sealed, branded (Linux) distribution, then we're recommending Red Hat. But in the device software space, we expect these platforms will be modified by our customers.

Is it fair to say you initially planned to use a Red Hat foundation, but now expect to use that version only for some sealed systems whose operating system can't be customized?
Klein: That's one way you could frame it. As these markets mature, as we may see in (the) ATCA (Advanced Telecom Computing Architecture) space, we're going to be working with Red Hat. We reserve the right to change our minds at Wind River based on customer feedback.

Who are your primary competitors?
Klein: The biggest is roll-your-own. There are lots of midgets and one dwarf. Fortunately, we're the dwarf. But now, you can't get there from here building it all from scratch.

MontaVista Software is the only pure-play Linux vendor. They've been the unfortunate red-eyed mouse here and have made a lot of mistakes in terms of their business model. Fundamentally, it's because they're not adding proprietary value around Linux. They're trying to sell free software. It's just not a model that has legs.

Another competitor is Green Hills, which is primarily in aerospace and defense. They have been the stalwart for staying very niche, very proprietary. At the end of the day, customers want the ability to go open source or proprietary or both. Enter Wind River with a $250 million (revenue) run rate, a $1 billion market cap and 1,100 employees, we're extremely well positioned to serve this marketplace.

Microsoft has better than a $250 million run rate.
Klein: Sure. But a big part of this market is going to Linux. That's a barrier to entry to Microsoft. And our customers have an aversion to putting Microsoft in their devices. Third, this market is about reliability. You can't do control-alt-delete on a pacemaker. And it's a very fragmented marketplace. You can't just do it on one processor. Microsoft's business doesn't support that.  

Autoplay: ON Autoplay: OFF