Tech Industry

Greene trumpets a virtual future

VMware's leader discusses the hows and whys of the industry's move toward virtualization.

PALO ALTO, Calif.--For years, VMware President Diane Greene led a company in an obscure part of the industry called virtualization. But all of the sudden, virtualization is a hot subject, and she no longer has the market to herself.

But as Microsoft's virtualization strategy and its open-source competitor called Xen both mature, VMware, an EMC subsidiary, is responding. At its second user conference, held this week, the company announced new versions of its core products--ESX Server 3 and VirtualCenter 2--that show the company focus moving from making a single server more efficient to making a large group of them more efficient.

Another change at VMware is an effort to make its interfaces into standards that its rivals can use--potentially simplifying affairs in the software realm but also solidifying VMware's lead.

They charged because they could. No one else was even close.

VMware's software lets an x86 computer be divided into separate "virtual machines," each with its own copy of an operating system. The company started with workstations, enabling programmers to simulate the interaction of several machines or debug new software in a protected environment. But now, most of the company's revenue comes from servers, which can run multiple tasks more efficiently.

Changes are in the offing. "VMware is a premium-priced start-up. They charged because they could. No one else was even close," Jay Bretzmann, director of IBM's high-performance Intel server division, said in a recent interview. He believes that the pricing model is starting to weaken with the arrival of competition but adds, "VMware is the de facto standard, the thing we're shipping today, and our customers are very happy about it."

And the company continues to grow. EMC reported on Wednesday that for the third quarter of 2005, VMware generated $101 million in revenue, a 67 percent increase over the $61 million it generated a year ago.

CNET's Stephen Shankland discussed the changes with Greene at the company's headquarters here.

To start out, why don't you describe what your company does?
VMware produces virtualization software. What that means is we take a physical x86-based system and we provide the multiple isolated, movable partitions that you can run operating systems with their applications in. In terms of what the customer gets, they get a way to drive utilization from, say, 15 percent, on up to 85 percent. They get very cost-effective ways to do disaster recovery, high availability, provisioning--all sorts of system-level services.

Pick a typical customer. What's their life before and after VMware? What changes?
A typical customer has got widely proliferated x86 machines, and depending on the power of the server, they can get a 10-to-1, 4-to-1 reduction in the number of servers they need. Or they can stop that proliferation and contain it better. And beforehand, to

Wherever we can, we let the instructions go directly to the hardware.
bring a new service online you have to go order the machine, install it in the server room, get it network-connected, make sure the power is there--it can be a multi-month process. Post-VMware, all they do is keep pre-built images of different software services like SQL Server, and when someone needs that service, they just find some excess capacity somewhere and deploy it.

So what's the penalty? Why doesn't everybody do this?
Actually, what we were finding is that for people who use it, it's become the default way that they run their x86 workloads.

But not everybody is doing this. Is it fair to say the bulk of the customers at this point are the big guys: pharmaceuticals, oil and gas...
On the desktop, we have millions of customers at this point; on the server, probably 20,000-plus enterprise server customers.

I was talking to one person a little while back who has hundreds of x86 servers. I asked about VMware, and he said it's too expensive. It's cheaper to buy a new Intel server than a VMware server license and he wasn't worried about buying servers that might be underutilized.
That's a customer that only wants to use a product to do server consolidation. They don't want to use any of the other advantages around provisioning, disaster recovery, fault tolerance, load balancing, serviceability. This is exactly the customer we faced from the first launch of the server product almost four years ago. We priced our product actually to be slightly cheaper than buying more servers. That was how we came to our pricing, because that was the only value people initially saw. But when you add up the power savings, the space savings, the hardware costs, and some of the software licenses, you do come out (paying) less. Then when you add in all the other functionality I just listed, the ROI (return on investment) is immediate and large.

So using virtualization to run multiple independent operating systems on the same computer is not new. It's something that Unix servers have been able to do for awhile and that mainframes have been able to do for decades. Why is it that it took so long to arrive on computers using x86 chip on Intel?
Yes, IBM came out with this in the late 1960s, early '70s on the mainframe. People are doing it today for some of the same reasons. When (Intel) came out with the x86 architecture, they put no support for virtualization in it because everybody thought that was something for the mainframes. VMware looked and said CPUs have gotten so fast, memory and disks have gotten so cheap, and there's full distributed networking and file systems, so the time is right to revisit this. In fact, the first project we called Disco because it, too, was something from the early '70s, but we modernized it and invented a lot of new functionality and brought it to commodity systems.

With VMware's software, an operating system or higher-level software thinks it's talking to a processor, but VMware software intercepts those commands and juggles the demands of multiple partitions?
Yes, the operating system thinks it owns the whole machine, it thinks it's talking to an x86 processor. Wherever we can, we let the instructions go directly to the hardware. Where we have to, we intercept them and (share hardware) resources.

For a long time, x86 chips couldn't do this. Now, all of a sudden Intel and AMD are putting in virtualization features--Virtualization Technology in the case of Intel, Pacifica in the case of AMD. How does that change things?
I'll just say we went to those companies right when we started and said, "Would you please add this support?" We knew it would make (VMware software) run better. So this is something we're very excited about and very supportive of. They're adding support for virtualization so it will run even faster and be a little bit simpler to do.

But with VMware, there's a performance hit?
It varies on the workload. If it's a very CPU-intensive thing it can be as little as 1 percent to 2 percent. On average maybe it's 15 percent.

We now have some open-source competition for you guys through the project of Xen, and Microsoft has arrived on the scene through its acquisition of Connectix. What's changed for you?
It's a huge testament to the value of having this virtualization layer, and in that sense we find it really exciting. Microsoft, right now, is on a hosted product where you already have an operating system sitting there. They've said that they're going to move to an architecture, like our ESX Server, that's hypervisor-based. We hope to partner with Microsoft. We think we do great things for Windows and helping them bring Windows into the data center, and we think this is a big universe. Then there's the open-source (Xen). For customers, it's not something you can use.

With all this industry interest, you need to bring the value to the customer and move the industry forward together. You start needing standards. We'll be 8 years old in January and we have a lot of experience, and we see some of the directions we think are architecturally suboptimal. So we're taking what we believe are the interfaces, the standards that will allow the whole industry to work together on this in a productive way, and we're opening them up and sharing them with the industry. We don't see any reason why we won't be working with Xen and Microsoft.

I talked to Xen and I talked to Microsoft and they didn't seem to have any interest in adopting your standards.
Yeah, I don't understand why that wouldn't be, given the depth of experience we have, the years of shipping product, working with customers, the fact that we've built, over the years, three different architectures for this virtualization technology. I think we understand what works quite well, and we're being open and no-strings-attached about how we share this with the world.

Perhaps if you're as powerful as Microsoft and have the ability to set a different standard, that would undermine (VMware's) market power. Whereas, if Microsoft adopted your standard for the underlying technology, that would serve to cement VMware's leadership by making what has been a de facto standard into an industry standard. If Microsoft comes up with some different standard, ships it with every version of Longhorn Server, maybe that will become the de facto standard instead.
I don't think that's the most productive for the industry, and I don't think it's the ideal scenario for the customers. I believe that everybody is going to win if we all work together on these standards. Microsoft has shown in recent years that they're more interested in doing what the customers want. It would be great if Microsoft would work with us on that.

Why didn't you do the standards move earlier?
Nobody came to us and asked us to do these things; this was a proactive thing on our own part. I think IBM and some of our partners are very standards oriented.

I was talking to a server executive at Hewlett-Packard, the largest x86 server maker. He said the base virtualization layer that lets you run multiple operating systems should be free. Do you agree?
Whatever you sell, you want to get enough money to invest in R&D. In VMware we have just a phenomenal R&D group, and it's in the customers' interest and our partners' interest that those people stay productive and keep inventing. But, as you increase (sales) volume, you don't have to charge the same thing to pay those people. So there are some economies that are going to change as the volume goes up. We're not there yet.

But this is the software business. Once you pay X amount of dollars to develop a product, you can sell it to a market over and over. It doesn't cost ten times more to produce ten times as many units.
That's exactly right. So I do believe that it's not healthy for the software industry to keep charging the same thing if you have a radical change in volume.

So as VMware is used more widely, the price will go down. But will the fundamental technology price go down to zero?
You know, I don't think there's anything at zero; you look at Red Hat, it's not zero. You have to pay to maintain things, you have to pay to support things, and you have to pay to do new things.

HP, Intel, AMD, Sun Microsystems, IBM--they're all funding development of Xen in varying degrees. Do you see price pressure on that core virtualization product?
Today, we don't see the price pressure.

Years ago, back in that early 2000s, you had hoped for an IPO.
Planned for an IPO.

Right. Why did you end up being acquired by EMC instead?
A lot of things happened. We were growing incredibly rapidly. We'd been profitable for a couple of years, so we clearly could go public. We also seemed to suddenly have a lot of suitors. We had always had offers, from when we were a year old, but all of a sudden it got kind of intense. We felt like this could work--maybe we should get proactive about this. I have to say it has worked out exceptionally well. The company has just thrived, you know, as an independent subsidiary of EMC. I consider it quite a luxury that I am really focused on growing VMware as opposed to the quarterly (results).

There are a lot of Silicon Valley start-ups that have an exit plan, which is to get acquired by Cisco or some big company, but there are also a lot of Silicon Valley start-ups that want to be independent. What did you weigh when you were going through this decision?
Independence for independence's sake was never something we were after. What we were after was bringing this technology to market and letting it realize its full potential. It's just really exciting for us to see the customer adoption of this technology and the partner cooperation and collaboration around these technologies and it's fun and that has grown and expanded since the acquisition. Had we lost that, it would have been the wrong thing to have done, but we didn't.

And you probably wouldn't be in charge of VMware?
No, probably not.

I often gauge acquisitions by what happens to the founders.
Yeah, we're all here. I thought it was pretty amazing that six years into the company, when we got acquired, we were all still there. And now it's two years after the acquisition.