We're working at quite a number of things in parallel here. For example, we need to get the infrastructure designed and orders put in for acquisition and construction. In some larger implementations, we may have an Infrastructure Manager working part time on the project. Certainly, we need to look at the technical architecture in terms of what tasks are done where, what computers are dedicated to what tasks, and so on. The picture on the left was used as a discussion centrepiece for this purpose.
Within the core of the project itself, we're building and testing all of the software components, establishing aggregation needs, creating work flow processes, and so on.
At some point, we will have done an initial load of the base star schema, then we will create an End User Layer (appropriate to the reporting tool we're using), and that will initiate End-User Application Development. With some prototype reports in place, we are ready to demonstrate a fully functioning pilot system.
At this point, we are at week eight of our crash project, but there's still a long way to go between pilot and production.
The picture on the left identifies a few of those issues. We need to add Slowly Changing Dimension functionality. We need to clean some source data before we can go live. We need to ensure that we have proper audit and tracking mechanisms. We must have error-handling routines. And so on.
Again, detailed, pre-prepared checklists come to the rescue. We remember training and communications. We remember involving the business in report requirements and design. We test our backup and recovery procedures. We don't leave anything to intuition or chance.
These two pages have described what can happen in a successful crash project, where the timeframe from start to finish is about fourteen weeks. More realistically, from greenfield to managed production should take about twenty weeks.