During the course of last 2 years, I’ve conducted SAP HANA performance workshops and reviewed information models at various customers. I encounter similar modeling issues or mistakes almost at all customer projects. The good news is that most of the performance problems can be resolved by making with minor changes to the model.
However it is critical to understand the reasons behind the performance problems consistently observed in the HANA implementations. Here are my observations.
- HANA modelers have pretty good idea on how to build attribute views, analytic views, and calculation views which are part of HANA modeling. However, there is significant gap in the knowledge of using these HANA information views effectively to address the business requirements.
- Some of the HANA modelers do not have data warehousing background and hence lack the knowledge of basic multi-dimensional analytical reporting, which leads to poor solution design.
- Modelers and developers seems to have the knowledge of “what works and what does not work”, but they seem to lack the knowledge “Why it is behaving the way it is behaving”, hence the struggle to apply the knowledge effectively while building the models.
- Finally there is a fundamental flaw in the understanding of Software Development Life Cycle (SDLC). Is performance testing a phase before Go-Live or is it part of unit testing? HANA’s query response is extremely fast for the smaller data set regardless of the modeling approach. This gives a false positive of efficient modeling. The realization of performance issues become visible only in Staging/QA system where models are tested against large data volumes.
As I mentioned earlier most of these problems can be resolved with very little changes. So it is essential to build the model right the first time.
So let’s learn to model the right way the very first time.
Please stay tuned. We'll have a webinar on this topic soon.