For some time, VMware Inc. has been synonymous with virtualization. But now the company has worked hard to move beyond this categorization and into cloud computing. News Director Alex Barrett sat down with VMware CTO Steve Herrod to discuss the company's new strategic direction, problems with cloud computing, and what happens if cloud computing turns out to be a massive flop.
You joined VMware in 2001; how has VMware's strategic direction changed?
Steve Herrod: One of the biggest changes to come in the past nine months has been our push into application development. We acquired SpringSource, RabbitMQ, and GemStone, for distributed data management. We also announced joint initiatives with Salesforce.com for VMforce, and with Google to run Spring applications on Google App Engine.
We suck the brains out of a server and put it in a [virtual machine].
Steve Herrod, VMware CTO
At the core of virtualization -- and the reason we've been successful -- is because we take nasty applications and put them in a VM [virtual machine]. My colorful way of saying it is that we suck the brains out of a server and put it in a VM. But something like 60% of applications that are written are custom apps. We also want to have a story for when you're writing new applications.
How is the Google initaitive different from the VMforce announcement you made with Salesforce.com?
S.H.: Salesforce is a trusted company with a strong customer database and a big customer base that is looking to extend [this data with custom applications]. Google is a completely different story -- it's an extremely developer-focused cloud, and it integrates well with Google Docs and identities.
Another component of the announcement is mobile devices, with integration to Google Web Toolkit. We want to run modern Web apps that look good and have a great UI [user interface] on a variety of devices.
What are VMware's broad goals for cloud computing?
S.H.: We're moving from straight virtualization to virtualization as an enabler for cloud computing. We need to help IT build the private cloud, to create differentiated but compatible public cloud offerings and -- to bridge those two -- by letting IT managers view IT assets that reside on on-premise and off-premise resources.
We also need to address security. One thing you hear CIOs say is "My data needs to stay in one state" for compliance reasons. I think we'll see the emergence of different kinds of public clouds: the New Jersey cloud, the hydroelectric cloud, the cheap cloud, the secure cloud, etc. One way we do that is by working with partners. We currently have 2,000 cloud partners -- that's up from zero one and a half years ago -- and each has a different angle about why they matter, and who they appeal to.
VMware has put so much effort into cloud computing. If cloud fails -- as other hyped outsourcing trends have -- is it a risk for VMware?
S.H.: If you think about cloud the way it was perceived previously -- as one giant übercloud -- then yes. But people are starting to recognize that there are core components of cloud computing that [apply to computing as a whole].
If you ask the developers ... they've skipped past IT [on cloud adoption] and gone around them.
For one, cloud computing implies levels of efficiency that are different than what you have today in terms of hardware and people. And there's no argument that we have to get better there.
Also, the notion of self-service is key. All of us in our consumer life recognize that we shouldn't need 12 signatures for something to happen. We're starting to see that with developers, who are voting with their feet and leaving [the confines of internal IT]. If you ask a CIO if they're using cloud computing, they'll say no. But if you ask the developers, you'll find that they've skipped past IT and gone around them.
But if cloud computing is going to be viable, it's going to have to have a solution for government regulation, which we're only going to get more of. If cloud computing hasn't been viable so far, it's because of the "I have no idea where my data is" issue.
How do you solve cloud's geographic problem?
One way is by building private clouds, which gives companies complete control, as they they have today. Another is to create groups of public clouds, where each has a trait assigned to it. For example, in the education sector, we're starting to see universities and state-education departments join together to provide shared cloud resources -- what we're calling "educlouds." It's sort of a cloud coop; with each contributing a piece, you're able to expand more than you would have on your own.