In part one of this Q&A, Greene talked with SearchServerVirtualization.com about the far-reaching implications...
of inserting a virtualization layer into the computing stack. Here, in part two, Greene comes back to earth and talks about company's current products, how customers are using them and what they are asking for.
SearchServerVirtualization.com: You recently announced that a third of your customers are running VMware Infrastructure 3 already. What do you attribute that to?
Diane Greene: Isn't that remarkable? In just four months… [Ed. Note: VI3 was launched in June 2006]. We surveyed 1,800 customers and of the people that haven't upgraded, three-quarters of those said they'd do it by the end of the year.
[Customers] overwhelmingly said that server consolidation is still a valuable thing to do, but they realize it was about the VMware Virtual Infrastructure. This is brand new stuff, but it just works and it has this phenomenal functionality….like more interoperability, better performance, the Consolidated Backup feature, and VMware HA (High Availability) and DRS (Distributed Resource Scheduler). We're so proud of our products.
What are some of your next steps?
Greene: There's so much potential there; we can't execute on our roadmap fast enough. We announced [VMware] Lab Manager that brings in functionality to group all your virtual machines in a natural way that maps into a distributed development environment. The customers just love that.
Is VMware Lab Manager the first of a series? Would you then have, say, VMware Business-Critical Application Manager?
Greene: Test and dev is a market that we want to play in so Lab Manager is going to take on its own roadmap. Virtual Desktop Infrastructure [VDI] is another example of a solution that we've come out with; so is ACE. VMware Infrastructure will have better things around disaster recovery high availability and security. Those are the things that are foremost on our minds.
Speaking of VDI, why do you think it will be successful when previous attempts at server-hosted desktops haven't panned out?
Greene: There are a lot of things in place today that make [VDI] quite doable. One, it really is a full desktop. There are no issues. There are no caveats. It's a full desktop. It can play multimedia, it can play to any kind of device, and there are all these really cool thin clients coming out. And because of things like [VMware] ACE [the company's client-side provisioning tool] and the fact that a virtual machine can run on a desktop, it can have mobility. Then there's this whole virtual infrastructure sitting there to make it highly available, load balanced across the servers and backed up. There's also an added push; for instance, in 2005, Japan passed a law that said you have to get the data off the desktop for security reasons.
But the mobility issue seems to be holding people back…
Greene: We're working on making that totally seamless. Early deployments are for remote call centers remote workers, remote transient workers, contract workers, etc.
How is virtualization changing companies' internal accounting processes?
Greene: All our customers want chargeback. What we do is provide all the hooks for the resource usage -- you can tell how much disk CPU and memory is being used. We are also partnering with people that want to export this [information] in a nice user interface to the customers so they can internally charge to their users. In the same way, I hope the ISVs will map [their software] to the resources. The hooks will be the same hooks as for internal chargeback but they'll allow the software vendors to say 'You can use this much resources, but if you use more, you're going to have to pay us more.' Just like today, if you use two CPUs you have to pay more. But it will be less arbitrary.
What are you working on in terms of upgrading and patching VM images?
Greene: We see a lot of room for improvement there. It's not actually a lot worse than the in the physical world, but there are some opportunities, once you virtualize, to do it better. We're addressing a virtual machine that is dormant, making sure it gets patched. Today, a virtual machine is like a real machine – if a real machine is dead, you can't patch it, because it's turned off. It's the same problem, but it's going to be addressed.
A lot of customers complain about capacity planning. Do you hear the same thing?
Greene: We actually have this software called VMware Capacity Planner that can go out and look at your network, see how much everything is using and come back and tell people about all the software they have deployed, how much resources they are using. Then they can very easily use that information to virtualize. The number of engagements of that is pulling up very, very quickly.
But with the resource pooling that we have -- the ability to dynamically add and shrink resources -- a lot of those issues are going away because if you get it wrong you can just dynamically add resources. It's not a make-or-break decision anymore. You're not penalized if you got it wrong. If you overestimated you can take some of those resources away, and if you underestimated, you can add more resources.
Maybe capacity planning is difficult for people who are trying to use their legacy servers that aren't optimized for virtualization…
Greene: Actually, I think that's one of the powers of virtualization: You can take a legacy server and P2V it, and you don't have to change a line of software, you don't have to recompile anything, you don't have to port anything; it just works. You get it in that virtual machine, put it in a resource pool and see what happens. Then, you can dynamically take away or add more resources.
I think part of why we've grown so quickly is because we were so non-disruptive to legacy applications.
What are your plans for systems management?
Greene: Well, we have VirtualCenter that manages and monitors VMs, and we offer that as a software development kit to integrated into bigger frameworks -- for example, IBM Director. And we're going to keep on moving that thing forward. We're never going to try and do physical -- it's just virtual -- and we're going to make it manage what we do and exploit what we do in a very focused way.
We have no ambition of being this big uber super-system management player. We have our work cut out for us just making something that lets people take advantage of all the functionality already in our virtualization.