Cobalt Networks' CEO Steve DeWitt shares his thoughts on the future of network appliances, open source software, and Cobalt's new role as a division of Sun Microsystems.
|PHOTOS © GARY WAGNER|
On September 19th of last year, Sun Microsystems stunned the hi-tech world by announcing that they were purchasing network appliance manufacturer Cobalt Networks for $2 billion. The deal was a surprise for a number of reasons, not the least of which was the price — the highest Sun had ever paid for an acquisition. The fact that Cobalt’s appliances are based upon Intel’s x86 architecture and run Linux makes the news even more shocking. In its entire history, Sun has never sold anything other than SPARC and Solaris-based systems.
This past November, Cobalt CEO Stephen DeWitt took the time to sit with Linux Magazine’s Robert McMillan and Adam Goodman to talk about what caused the massive turn-around at Sun and what Sun’s plans are for Cobalt’s Linux based appliances.
Linux Magazine: I’ve been just a little confused about the acquisition. I heard that Sun’s plans are basically to drop Linux and adopt Solaris for your product line.
Stephen Dewitt:That couldn’t be further from the truth. Cobalt will become the appliance division of Sun. That’s what my job’s going to be; I’m going to run the appliance division of Sun. In the world of appliances, the value-add of operating systems is underneath the covers. You don’t buy an appliance because of its operating system. You buy an appliance because of what it does and how it does it. It’s economics. It’s return on investment. Those are the types of arguments that you use when buying an appliance. You don’t know what’s underneath the hood.
There are classes of appliances where certain combinations of hardware, OS, and application technology make sense. In the hosting world, for example, our appliance platform leverages Linux extremely well. In the future, there may be classes of appliances that are based on other operating systems. Solaris-based appliances may make sense for certain enterprise applications where you have a lot of vertical scale requirements. But we are in no way shape or form departing from our Linux path.
LM: So what do you get from Linux that you wouldn’t get from Solaris?
SD: Well, I don’t think it’s a question of what Linux gives vs. what Solaris gives. I think what we’re able to do with the Linux platform is unique. We’re able to control the end solution that we deliver. When you get one of our appliances and you’ve got a problem with it, you call us. When we look to deliver additional areas of value add, offering very high availability across a very distributed server file, that’s all technology that we’ve built internally, where we’ve leveraged the Linux environment to do it very quickly, very reliably, very cost effectively. We don’t pass on any incremental cost to the end customer because of any royalties or licenses that we have to pay on the operating system that we use.
So there are some things we’re able to do in the Linux world that give us tremendous horizontal scale. Solaris was built for tremendous vertical scale. I mean [Solaris is designed] to crunch and support the most demanding client-server applications out there. So for certain classes of applications that require that sort of capability, there’s an opportunity to appliantize those applications. Remember, our value proposition is not that we’re a Linux company or that we support blah, blah, blah operating system. We build appliances.
LM: So do you think of yourselves as a Linux company or an open source company in any way?
SD: No, we’re an appliance company. And, we are tremendous advocates for open source. Open source is a critical means to an end.
LM: Are you saying without open source there would be no Cobalt? If there hadn’t been a Linux or an open source operating system that you could use, how would that have changed things?
SD: Well, back in the fledgling days of Cobalt, we looked at a number of operating systems. We looked at BSD and all the ones that you would imagine. The reason we went Open Source was really three-fold.
First, we wanted to have control over the reliability issue. Having access to the source guarantees to us the ability to make the right sort of kernel modifications. We were able to deliver a solution where we knew what the reliability of that solution would be. We knew it going in. We knew what the failure rates were going to be. We knew how to control the end user’s usage in order to maintain reliability. That was the number one reason why we decided to go the open source route.
The second reason we chose open source was that it allowed us to get to market fast. We were able to go from literally a concept — no boards done, no layout, no software, nothing done — and take a product to market in a period of five months. It changed the hosting industry. And, we operate in six-month development cycles. The best systems companies in the world can’t spin in six-month cycles. We spin in six-month cycles for complete appliances — hardware, software, all the surrounding value-add.
The third reason was the economics. I don’t think we could have hit consumer price points without open source. We do not pass on incremental costs to our end customers. That now allows companies like AT&T to offer very rich dedicated hosting or value-added hosting solutions at consumer price points. And when you’re going after these mass markets, you have got to hit it. If cable costs you $200 a month, you wouldn’t buy it. But if it’s 39 bucks and I get 50 channels, that’s great. If for another $3.99 I can get a movie by going “click,” now I’ve hit consumer price points. Those are the three reasons why we went open source.
LM: Do you think Sun’s business model is going to migrate towards the Cobalt model? Or, do you think Cobalt is going to more closely follow Sun?
SD: I think we’ll be a bit disruptive inside of Sun. I think it would be kind of naïve to think that we wouldn’t be a bit disruptive inside of Sun, because we do things a little bit differently. Sun has had a strong history of owning its own intellectual property — almost to a fault — and we’re open source based. I think there will be a hybrid business model that comes as a result of the merger. I think Sun is very attracted to Cobalt’s model — the way that Cobalt works with developers. Sun has an awesome history of working with developers. And you know, in the small time that we’ve been in the face of the industry, we’ve attracted a huge number of developer partners — from big flagship companies like NTT DoCoMo to small Web companies that are building the next generation class of applications.
LM: Did you ever look at what happened with Star Division and Open Office? You know, Marco Boerries and what happened when his company was acquired by Sun?
SD: You always look at that. I’ve been involved in a lot of acquisitions. Between Cisco and Symantec, I’ve done a lot of deals. I’m a firm believer that there are a couple of key things that you look for in an acquisition. You look for great cultural fit. That’s very, very key, because high employee retention is obviously critical. Building better mousetraps is a piece of cake; building great companies is very hard. Great companies are built with great people. When you go into an acquisition you are acquiring the people, not the technology, not today’s products; it’s tomorrow’s road map. And I think we have an awesome cultural fit. I’ve known Ed for a long time. I think we have a shared vision in terms of the role of the network.
LM: Well, Sun has been heavily criticized by open source developers over the last few years. That’s kind of why I was asking about Star Office and what had gone on there. I guess I don’t know enough about your business model right now — how much you really depend on the Linux community to be a successful company. But, I would be apprehensive if I were in your shoes because of Sun’s relationship with open source software. Do you think they could take a core part of your product and try to turn it into something that –
SD: It’s not going to happen. This goes all the way up to Scott and Ed [Scott McNealy and Ed Zander, CEO and President of Sun Microsystems -Ed.].
LM: But would you have thought it was a good deal if Sun said, “Well, we’re basically interested in buying you because you have a good brand, but we really think you should be running on UltraSparc”?
SD: We wouldn’t have done the deal. It’s more likely that Sun would have tried to build appliances on their own. I think Sun showed something significant. I’m not saying this because they’re my new boss. I don’t have to say that and I wouldn’t. I think Sun showed the world something here.
This is about appliances, not operating systems. At the end of the day, Cobalt equals appliances; there are a lot of things that are powered by Java right now, and you have no idea that there’s Java underneath the covers. Sun has a long history of building enabling technology that ultimately gets hidden underneath the covers. We build appliances. We build a class of infrastructure that allows developers to get their applications to market in a unique way. Now we’re going to build appliances for the service provider market so that service providers can then differentiate themselves, vis-à-vis other service providers that are out there, by offering different types of applications, generating different monthly recurring revenue and different levels of profitability. We’re going to build a class of appliances for enterprise customers.
Enterprise customers have gotten very savvy to the fact that there are classes of applications that can be very easily appliantized. That drives costs out of the IT infrastructure. We are going to build a class of appliances for the broad markets, the consumer markets, the residential markets, the SMB markets, etc. At the end of the day, you’re going to have an appliance that serves those markets. I’m not going to be talking to you about the operating system. I’m not going to be talking about whether it’s running AMD, UltraSparc2, or UltraSparc3. It will not be part of the value proposition. If you go to our Web site right now, you’ll see we don’t talk MHz and we don’t talk operating systems. If you want to buy a microwave with the little spinning plate that pops your popcorn evenly, you go out and buy that. You don’t worry about the embedded technology that’s making it all happen. It’s application driven.
LM: That’s interesting. I understand what you mean by that, but at the same time you’re saying, “Well, if they had wanted to switch the operating system or the hardware platform, then for internal political reasons, we would have said ‘no.’”
SD: No, that’s a fair thing to say. I think if they’d come in…our OS developer’s and our general skill base is Linux. But Linux and Solaris are not that different. Unix is Unix. There are obviously differences, but it’s not anything in comparison to Linux and NT, or something along those lines. So we have a lot of developers and a lot of people that have Unix skills. It isn’t a big reach.
LM: Do you guys see any potential for a technology conflict between yourselves and Sun? I mean, Solaris and Linux –
SD: Oh man, I sure hope so [grins broadly].
LM: Can you just tell me a little bit about how the merger with Sun happened? Did Ed Zander just call you up one day?
SD: Oh, Ed and I knew each other for a long time. I knew Ed from my Cisco days, because Cisco and Sun have always been strong partners. So I knew Ed very well and didn’t have a problem picking up the phone.
They came over about a year or so ago and got an update in terms of what we were doing. My posture towards Sun was very consistent all along — I thought there was an opportunity for the two companies to articulate to developers a new way of delivering applications. That really is the end game here. It’s the end game for Cobalt, and it’s the end game for Sun to attract new developers to your value proposition for the delivery of their applications.
Sun was very keen to that. And Sun, like any company that builds general-purpose servers, was looking for a way to differentiate itself from Lintel or Wintel or any of those guys. So they were interested in the appliance space.
They also had seen some tremendous competitive success in companies like Network Appliance. I mean, think about this — Network Appliance is a company that popped on the scene with a class of infrastructure that was sold right into the core market of Sun — right into the breadbox of the enterprise. Their sales force was selling a general purpose hard-ware platform with software value add on top of it against a turnkey appliance. Guess who was winning? So all of a sudden, the lights are going on over there. They’re going, “Well geez, what’s the difference between an appliance and a general purpose server?”
Anyway, we had a good dialogue going. A lot of it was just casual conversation between Ed and myself. As quarters would go by and we’d have good results, and they’d have good results, we’d talk and ask each other, “How are things going? What are you seeing?” because we actually have a lot of common competitors.
LM: During this dialogue with Sun, did you have to educate them about any aspects of open source?
SD: Absolutely, but I think at first they were attracted to our brand. “We keep hearing about this Cobalt company. We know they’re doing appliances and at incredible price points. How the heck can they be doing that?” Then all of a sudden, when we really started rolling with our hosting appliances — around the time that we were going public and probably six months following our IPO — we started to win some pretty significant service providers.
If you’re a sales rep selling IT infrastructure, you walk into the data center and you see your Cisco routers and Sun boxes right there. Then all of a sudden you started seeing racks and racks of these blue appliances. And I think a lot of the interest bubbled up through their sales organization going, “This company Cobalt is starting to land some pretty substantial business here.”
LM: Were you going up against Sun for that business?
SD: We weren’t going up against anybody. That was the amazing thing. We unseated nobody. And that’s not all that much of a surprise, because service providers were looking at rolling out something new.
If I’m a big service provider that’s had a history of selling dial tone and want to get into the data services game, now it’s a question of, “What are my data services?” It’s like the apartment building sale; “Well, I can offer you shared hosting, and you can have two e-mail addresses and 5 MB of disk space for $19.95. Or, I can move you into your own home, and by being in your own home you can have unlimited e-mail, unlimited disk space, and I’m also going to offer you three kinds of value add. Do you want a shopping cart? A merchant account? An e-commerce thing? ” And boom, I can do it for $199 a month. Now you are starting to talk some very compelling economics.
That was really the sweet spot we hit — enabling the delivery of a class of hosted applications not only at consumer price points, but that were also profitable for a service provider. Because if I’m selling you bandwidth and services, and that infrastructure costs me $900 and I’m getting $200 a month, I’m cash-flow positive on that box in a few months. If I lease it, which is what most service providers do, I’m cash-flow positive on consumer applications from day one. That’s why our business model worked.
This is the biggest land grab ever in the infrastructure world, and we are able to offer service providers a highly profitable model because our price points are very attractive. And I think, like I said before, when we won accounts like AT&T, bells and whistles went off over at Sun.
LM: You really didn’t go up against any direct competitors for the new business you were winning?
SD: There were competitors that had a different model than we did — nobody competed on our model; we were competing against general-purpose servers.
LM: Right. Well what about today though? Doesn’t Dell sell appliances, and –
SD: Yeah, there are a lot of people that are starting to build appliances and they’re going to start to eat up some of the things we do as an appliance, which means that we have to continue to focus on building the stack. And that’s the road map that we’ve been on. Now we’re on our fourth, entering our fifth generation. Most of the world is now entering the first generation. What they call appliances is nothing more than Red Hat on a 1U box. Now we’re starting to see second-generation boxes, which means Red Hat on a box with a Web front end. It’s very easy to compete against.
LM: About moving up the stack as everybody climbs into this market –that leads me to thinking about next generation application development and something we’ve been talking a lot about lately, which is Microsoft’s .NET initiative. How do the kinds of things that Microsoft is doing factor into your plans or Sun’s plans?
SD: The bottom line is that it’s a network-centric world. If you’re a developer today, you’re not thinking about how you’re going to burn CD-ROMs. You’re thinking, “How do I go out and capture this huge opportunity that’s out there?”
Think about all the great buildup that happened in the networking world. I sit on the board of a few companies and I talk to a lot of small companies. I haven’t read a business plan lately that says, “Oh, we’re going to focus on the Fortune 50 blah-blah.” Nobody thinks that way anymore. Now everybody’s thinking, “How can I get it out to a billion people?” That’s changed the whole mindset.
So if you’re Microsoft, you have to articulate a vision where the Net is the foundation. The reason why people consume technology from Microsoft is that Microsoft delivers technology that speeds application development, makes it better, makes it more reasonable, makes it easier to use, and makes it more reliable in a distributed world. That’s the proving ground for the companies that want to have architectural influence.
Whether you’re Microsoft, Red Hat, or Sun, you have to compete for that value add. It’s not, “Oh geez, look at the installed base of Windows 2000.” Who cares? It’s irrelevant because it is so small in comparison to the market people are building towards. That’s the great leveler. That’s the beauty of the Internet — it changed the perspective of the developers.
So Microsoft’s initiatives, the stuff that Sun does, and the stuff that everybody else does has to cater to that. We’ll see. The gun went off a couple of years ago, and the race has begun. Everybody’s got to fight for it. I mean, Microsoft, Sun, and all the big market share leaders have to fight for it as much as anybody. Granted, they have huge developer followings. I mean, Microsoft has tens of millions of developers that are out there; they dominate the client world. Sun has millions of developers; they dominate the enterprise world. It’s going to be a hell of a battle.
LM: Other than yourselves, who do you think is leveraging open source most effectively from a commercial point of view?
SD: Application developers are really the only audience I can say –those that are looking at building a solution where the value proposition is not tied to the operating system. There’s a lot of innovation happening there, a lot of software developers that are building some very cool things on top of Linux. There’s a huge percentage of next-generation net apps that are being built on top of Linux. I don’t know what the percentage is, but it’s a big one.
Nobody’s actually explained a business model [for a Linux distribution] to me that I’ve understood. I may be a shy country boy trying to make it in the big city. But at the end of the day, you have to sell goods and services and you have to have a differentiated value add. I don’t believe you can hang a dollar value on an operating system. I don’t buy into it.
LM: It worked for Windows for a while.
SD: That’s why they’re in front of the justice department right now, and that was so 80s. That’s not going to make it in the future.
Adam Goodman is editor and publisher of Linux Magazine. He can be reached at firstname.lastname@example.org.