Today I'm at the IBM Impact show in Vegas. I've made the trek to this SOA event ever since the first one in 2007. This time, IBM is making a concerted effort to link SOA and cloud computing. No surprise there: When it comes to trends, the cloud is the only game in town, and everyone wants in.
Still, I could not suppress a chuckle when I saw the announcement last Thursday for a new IBM WebSphere CloudBurst Appliance. I knew that IBM was the foremost promoter of the "private cloud" -- that is, architecture and technology for your own datacenter to build an infrastructure resembling that of a cloud service provider (such as Amazon or Google). But a cloud in a box? That's a good one.
[ For weekly insight about cloud computing, turn to whurley's razor-sharp blog on the subject. | Also, keep up on the latest in cloud developments with InfoWorld's Cloud Computing newsletter and Cloud Computing channel. ]
As it turns out, the appliance is a turnkey server running the new IBM WebSphere Application Server Hypervisor Edition. Like BEA/Oracle, IBM has developed a version of its app server that doesn't require an OS -- it runs right on top of the VMware hypervisor, boosting performance and saving the customer OS licensing costs. Installing an app server on the virtual version of bare metal sounds tricky, so IBM has preinstalled WebSphere and VMware on a box and and called it an appliance.
What does this have to do with the private cloud? Well, basically, Java shops that adopt SOA now have a simpler way to spin up services and scale them. So if you believe as I do that SOA and virtualization are essential parts of a "cloudlike" infrastructure -- agile, scalable, and centrally managed -- an app server appliance could have a role to play. IBM also claims that CloudBurst meters the use of resources, so those who use its Java-based services can be charged back as necessary. For centrally controlled deployment and scaling, CloudBurst integrates with Tivoli Service Manager.
Fair enough -- but here's the reality of any "private cloud" effort. First, no self-respecting IT person would call it that. Second, no one is going to rip and replace their entire infrastructure with appliances or anything else. CIOs may drool over the agility and scalability of Google's infrastructure, but it was built from the ground up for such a purpose. Corporate IT infrastructures were not.
The private cloud is just a new way of describing an idealized end-state for IT, with optimized resources, low operating costs, and the ability to scale capacity or spin up new applications on a dime. Any discussion of the private cloud must acknowledge how far we are away from that. If you're not Google, you're talking about a long, incremental slog -- parsing legacy systems and applications into managed services, figuring out where best to deploy virtualization and how to manage it, and adding new services from scratch where you can justify them.
No private cloud could envelop an entire, existing IT operation in all its diversity. But SOA, virtualization, and datacenter automation are all strategically important, and any solution that moves us closer to adoption merits consideration. I question any insinuation, though, that a huge, top-down architectural initiative like the private cloud could ever be plug and play.
Later today I'll be interviewing Senior Vice President and Group Executive Steve Mills, who heads up the IBM Software Group. I'll be asking him questions related to this post. If you have other questions, please send them along, and I'll try and get them answered.