Nearly all enterprises say cloud computing's primary mission in their business will be a combination of offloading...
peak demand and backing up critical applications. This explicitly demands a hybrid cloud in which applications, data and users form a public/private ecosystem. Making that happen without a fork-lift upgrade of IT can prove difficult, but there are ways to reduce the pain and assure an orderly cloud transition.
There are two paths to a hybrid cloud in a technology sense; the virtualization path and the service-oriented architecture (SOA) path. If your company has made a significant commitment to either of these technologies in your data centers, you can build on that commitment to create a hybrid cloud. In fact, the more you have invested in virtualization and/or SOA, the less you'll likely have to spend to exploit cloud computing. If you're not decisively in either camp, then you made need to take a stand in your data center before stepping into the cloud.
All virtualization tools, including those from market-leader VMware, can be extended into the cloud and hybridized with public cloud resources. The key component in this, sometimes called a "director," assigns virtual machines (VMs) to either your own data center or the cloud based on policies. You can use cloud offerings like Infrastructure as a Service (IaaS) or Platform as a Service (PaaS) as long as they look like your own virtualized pool of computers, but you will have to deal with migrating data into and out of the cloud.
At the same time, all of the popular SOA-based middleware frameworks -- including those from IBM, Microsoft and Oracle -- will support extension into public cloud resources. Here, you can either choose from a vendor-specific cloud with PaaS functionality that's compatible with your internal SOA middleware or host your middleware on an IaaS cloud. The same issues with data migration, of course, will also exist here.
Choosing between virtualization and SOA
But what if you haven't made a commitment to either SOA or virtualization? That's where the real risk of a "cloud revolution" with high startup costs comes in. Decide whether you're a better fit for one of these technologies or a totally public cloud application model. The best choice will depend on your adoption costs and your benefit case.
If you get most of your software from a single vendor, there's a good chance that vendor will either offer SOA-enabling middleware or support a third-party offering. Your cloud project would have to first reorient your applications to fit your vendor's SOA model and then extend that model into the public cloud. The "revolution risk" here will come primarily from any applications you developed on your own or obtained from specialty vendors. If these aren't SOA-ready (meaning that they're componentized and can be orchestrated with a workflow engine) then you'll need to revise or replace them. Software is the "revolution risk factor" in the SOA-centric cloud migration.
If SOA evolution won't work for you, the next option is to adopt virtualization. Nearly any application can be made to run in a set of hosted VMs, so the real risk here is that your current servers may not be good candidates for virtualization. If that's the case, you may have to upgrade your data center to obtain efficient hosts for your VMs. In most cases, enterprises with relatively new servers won't have any problem getting virtualization to work, but older servers come with increased risk. Your current hardware thus becomes your revolution risk for a virtualization-centric path to the cloud.
Other options for a cloud transition
If neither of these models work at reasonable cost points, you have two options. First, you can move your entire IT operation into the cloud. This will normally create a high transition cost, but it also displaces future capital costs, power, facilities, etc., and the benefit case may cover your costs. If it doesn't, the second option is cloud integration at the GUI level.
Many enterprises today have Web front-end software that lets them access applications from a browser and also compose the workers' GUI using these front-end tools. Because these tools will normally let you build a screen from data obtained in multiple applications, you can combine cloud-hosted applications with nearly any applications you run in your current data centers. This GUI-integration approach can serve as the basis for a long-term strategy, but it can also be used to help manage the transition costs to hybrid clouds for users with a few hard-to-migrate applications.
Cloud computing offers capital and operations cost reductions that have to be traded against the cost of transitioning to cloud services. By taking a path that manages your transition cost and risk, you'll end up with the best cloud outcome.
About the author:
Tom Nolle is president of CIMI Corporation, a strategic consulting firm specializing in telecommunications and data communications since 1982.