Application decomposition may not sound like a desirable thing, but as cloud computing evolves, it will become a more common enterprise application exercise.
As Brian Ott, vice president for the worldwide cloud program at Unisys notes, most enterprise applications running today are pretty monolithic. In order to take advantage of the economic potential of the cloud, many of these applications are going to have to be decomposed into more modular sets of components.
In that model, for example, a compute-intensive analytics element of the application might run in the cloud, while more latency-sensitive transaction components would run on servers in the customer's data center.
Ott says that in order to run applications in the cloud, IT organizations first need to study their application workloads. Once they ascertain what their application workloads are, then they need to break those workloads down into sets of discrete functions to determine what elements of the application would run best where.
Unfortunately, most IT organizations don't possess the necessary self-awareness of their application workloads to determine what functions should run where in the cloud. So Ott expects it will take a number of years before IT organizations develop a truly dynamic approach to running applications in the cloud. But one thing he is certain of: Those applications will be running in a hybrid model spanning both private and public cloud computing resources.