Why Microsoft will rule the cloud … someday

If there’s any company you associate with deskbound fatware, it’s Microsoft. Most of its revenue still comes from Windows and Office, and so far, Redmond’s ventures into Web-based applications have been pretty timid. For example, the browser-based versions of Word, Excel, and PowerPoint unveiled at the Professional Developers Conference (PDC) last fall still lack a firm pricing model or release date.

And then there’s Windows Azure, Microsoft’s new development platform in the cloud. When Ray Ozzie introduced Azure at PDC 2008, I was impressed. A real competitor to Amazon Web Services plus a big stack of .Net stuff? Cool. Yet just this week Steve Ballmer said that Microsoft will not have “the ability to go to market” with Azure until the end of 2009.

[ For a more detailed description of what Redmond’s platform-in-the-cloud contains, read David Strom’s “Making sense of Microsoft’s Azure.” ]

Yes, Microsoft is late to the cloud. And yet, years from now, an opportunity exists for Microsoft to dominate cloud services like no other player could.

My opinion derives from a recent conversation with Steven Martin, senior director of developer platform product management. He made no overtures about world dominance; that’s just irresponsible speculation on my part. Instead, Martin focused on the same stuff everyone does when talking about cloud computing: big, compute-intensive applications that you wouldn’t want to buy hardware for because you run them only occasionally. Or, in the same vein, new applications you want to pilot or test.

I’ve gotten the same bucket of cloud computing apps pitched at me by Amazon, IBM, Salesforce, and even Accenture. But when I forged ahead into future speculation, Martin gave back some interesting stuff. For example, when I asked about extending on-premises Windows Server virtualization into the cloud, Martin said that was something Microsoft was “actively exploring.” Hyper-V now has a shot at competing directly with VMware, thanks to the addition of live migration in Windows Server 2008 R2. So imagine an on-premises pool of Hyper-V virtual machines extending into Azure — with cloud capacity as “overdraft protection” — all managed as a whole.

What really got me going, though, is that Microsoft is actively cultivating .Net partners to take up residence on Azure. Epicor, an ERP software provider, is creating a software-as-a-service .Net version for Azure. MicroFocus, which specializes in tools to migrate Cobol apps off mainframes, will soon let you deploy on Azure. OpenText, a content management company, is also taking the Azure route. The list goes on. Plus, along with .Net Services, why not throw in the entire stack of Microsoft servers? Exchange, SharePoint, SQL Server, and Dynamics CRM are there now (although not necessarily ready for prime time). Already, says Martin, an “industry leader in oil and gas” is using BizTalk and Windows Workflow hosted on Azure for process integration.

Plus, remember those dozens of Web services protocols we’ve heard about for years that so few developers actually use? Those will be baked into Azure, too. So will Active Directory services. In other words, you can imagine a day when Azure will contain everything Microsoft makes, plus an ecosystem of third-party .Net services and applications. No customer datacenter would ever have that much Microsoft stuff.

The result would be an incredibly rich environment for developing composite applications in the cloud. And as Martin points out, “in the cloud, we have opportunities for a higher degree of componentization.” By stringing together components in this ultra-rich environment, VARs could create vertical, cloud-based applications tailored to all manner of small and medium-size businesses. Enterprise IT will stay away, of course, but it may not be able to stop certain departments or lines of business from giving it a whirl.

All this is years away and Microsoft has all sorts of opportunities to screw things up — with incredibly complicated pricing models, for example. And yes, Azure is a homogenous .Net world for Microsoft developers. But that’s the point: A walled garden can enforce standards and provide a reliable middle tier.

Meanwhile, the dream of “Internet-based applications” built from componentized services strewn across the Web has proved largely impractical, Google Maps mashups notwithstanding. Beyond one-off, point-to-point development, can Internet chaos ever be organized to move beyond mashups? Will we ever get any closer to, say, a federated identity system everyone can agree on? It seems more likely that there will be multiple clouds — more and bigger platforms still only marginally interconnected. One day, Azure seems certain to be one of the largest.