Scaling Out Agility in B.I. Infrastructures

One of the challenges in applying agile methodologies to the Business Intelligence space is that traditional architectures layer tiers by technical function, introducing tight coupling of technology that makes changes expensive.

This tightly coupled dependency makes an over all B.I. solution more susceptible to change and support risks. Changes are hard to implement, without disturbing related functionality that wasn't originally intended to change. And because changes are often "under the hood", end users will have little understanding about delivery delays ("technical debt") or having to go through another Q/A step that doesn't seem to provide any additional functionality to them.

Along the concepts established by Service Oriented Architectures, if Business Intelligence infrastructure, tools and interfaces become more service oriented, than platform-driven, the agility of B.I. (the content, the results) could be increased.

The linking arrows at each generic level indicate different ways to integrate across functions.
For example, the User Interface level, hyper links facilitate cross functional relationships, such as drill-downs and drill-throughs across disciplines (as appropriate, driven by security rules).

On the Process level, each subject matter has its own functionally independent component that can be referred to by other components, or call out to other components, which do not have to be aware, let alone dependent, on the other components. The key point being loosely coupled.

Lastly, the Data leve would integrate by remote database linking, replication, or a on-demand loading.
Data models should be more modular, and again loosely coupled, for example horizontally partitioned across hosts.


Now, let's take note that layered tiers by functionality still make sense where the system performance is more important the the agility of the business; a classical trade-off between speed and flexibility. New technologies and system architectures, such as Apache Hadoop can help reconciling the both.

In the spirit of focusing on providing intelligence to the business, these technical/architectural concerns ultimately should be delegated to a cloud operated by service providers who focus on optimizing along those lines, so your company can focus on the pertinent business questions.