Today is another one of those days that I feel like a total rookie. I’m afraid this post is really much less of a “here’s how to go do something cool” and really about the future. Like many others out there, I’m being introduced to the big wild world of SAP products outside of the SAP Business Objects stack. I’m a small part of a great big project where I’ll finally have an opportunity to interface in some way to SAP BW. On the one hand, what a great opportunity to learn more and really go deep with this capability. On the other, it’s a little intimidating shifting gears on all those years of history of traditional data warehousing and data marts to report in a new way.
This project is actually very cool. I’d love to be able to share it some day…but you get confidentiality agreements and stuff. What I will say is that for a huge corporation, it is very progressive and shows they actually DO care. The project itself has just a few core tenants that stand atop all of the dteailed business requirements:
- Make cool dashboards
- Create an easy to use adhoc environment, because we are looking for trends and patterns we don’t already know about between the distinct subject areas
- Create a very navigable taxonomy flowing from the dashboards to the detail level data
That aside, a fair share of the data I’ll be dealing with is already in SAP BW. That makes it seem like a bit of a slam dunk. All those opportunities to use SAP BW Accelerator with Explorer and Xcelsius on top of that….sweeeeeet. But then it starts to get a little funky.
Coming from the other direction, it also incorporates a significant amount of external data. This includes other disparate data marts and even some Excel based data. That part isn’t a big deal. The SAP BW experts assure me that integrating data is definitely doable. But as this model starts to grow into three good sized subject areas, are we really scaling SAP BW correctly so that a single universe can consume all this data and make it usable in an ad hoc environment while performing well?
I think the reason I struggle with this today, and why SAP is doing what they are doing with the semantic layer, is that federation is really the key to future success here. Without federation, it gets hard to see clearly how big SAP BW sources can translate into a single universe without impacting performance. So what do you do today? With federation at the semantic layer a ways off, and even potentially longer for a large enterprise, what is the appropriate architecture? SAP BW or a traditional data warehouse (presumably loaded with DI or other ETL tool).
I’ve actually already heard some strong opinions both from SAP as well as consulting partners with experience here. Guess what…opinions vary based on who you ask.
Yep, it’s definitely too late to decide tonight.
I think federation is big, but it is only the first step. We also need aggregation to make things fast for the dashboards, and solid business rules to make sure the aggregations match the details. Then, we need sandwiches, because all of this work is making me hungry.