Using SAP HANA as a Logical Data Warehouse Platform
The Enterprise Architecture function at BAT created a technology agnostic vision for their BI roadmap during the early part of 2014 by researching industry trends, key supplier roadmaps, peer companies and internal challenges.
The key requirement that emerged was the need to create an underlying platform for:
- Speed to Market – A system that can be developed and deployed at a rate which matches the speed at which people need to work
- Data Sources – the business needs access to multiple sources of data in one place.
- Self Sufficiency – The ability to use, edit, refine and adapt information to a specific need without relying on IT support
- Accessible Information – Available internally, externally, in highly visual format with information that is to be available instantly and as close to real time as possible to deliver the greatest utility
- Seamless User Experience – An intuitive interface, simple navigation and straightforward to use, a single way of working, in a single location, providing everything required
- Integration with Microsoft Products – As with many businesses Microsoft products form the core of a lot of business processes, good BI should work with, not against that
The architecture foundation required was what Gartner refer to as the “Logical Data Warehouse”. This allows the data to appear in one place even though it may be physically in different places. BAT found that SAP HANA had features such as Smart Data Access to allow the linking of data and HANA Live to significantly simplify the architecture in order to deliver the requirements stated above.
The project was kicked off with Bluefin Solutions in September 2014 and was ready for use from November 2014.
Whilst this implementation is a PoC, it is acknowledged that if it did go live the human impact would be significant because it would accelerate the deployment and performance of existing solutions and also open up other opportunities. Additionally, it is widely anticipated that a full-blown implementation of HANA – and the adoption of cutting edge technology that it would represent – would aid the retention of expert technical resource.
Under the overall objective of “speed to market”, the ease and speed of development was a significant gain. HANA modeller is very simple to use. Reports can be quickly prototyped using SQL and then moved into HANA Modeller Views to enhance performance.
Under the overall objective of “Accessible Information”, three particular items were noted:
• Real-time data – SLT was used to load data into HANA. This is a replication tool which enables changes in ECC6 to be copied into HANA in near real-time. SLT has again exceeded expectation with its performance and ease-of-use.
• Connectivity to tools – HANA was connected to BI tools from Microsoft, Tableau and SAP. Although it was possible to connect to all the tools simply, there were differences in how they behaved. Investigation into front-end tool requirements will be the subject of the next phase of this project.
• Performance – Although the HANA Server was a very low specification the HANA Calculation views delivered stunning performance, and the simplicity of the architecture significantly improved the development time and potential trust in the data by removing data latency and the potential for error in multiple transformations.
The advantages of HANA Live Browser & Documentation were also noted. HANA is self-documenting, both through document creation and through a portal. When new views are created the portal updates and provides full lineage and definitions.
The table below shows the measured performance improvement for a selection of key reports:
In short, even using a development specification HANA server the results were still very impressive, exceeding expectation.