Part Three – Enable real-time analytics with BW on HANA
In 2012, we continued our journey to HANA by starting to migrate our internal BW systems on HANA. This was the first system where the classical database was completely replaced by HANA.
It is important to mention that SAP runs single global instances internally and BW is used by more than 7.000 users every day. The backbone system of SAP’s business processes consists of ERP, CRM and BW. With the migration of our internal BW system we started to migrate our first mission critical systems to HANA.
Until HANA, BW ran on a very mature IBM UDB database, approximately seven terabytes big. The migration to HANA was as simple as a migration to any other database. However, we found that using RDS packages made the migration effortless and we implemented an architecture to support our high availability and disaster recovery requirements. We actually run today BW on HANA in a clustered scale-out set up across two data centers. Compared to the classical infrastructure set up, this is less expensive and very scalable in its hardware structure.
However, migration of BW to HANA was not only a database replacement. SAP ensured that the standard BW coding was optimized to run on top of HANA. This being said a lot of business logic moved from the application layer into the database. This is the reason why we are able to enjoy performance improvements of factors even beyond x100.
Our SAP in-memory journey started long before HANA. In BW we used a BW accelerator (BWA) which was an in-memory product. BWA already delivered a huge benefit in boosting performance of selected BW contents. However, the downside was that BWA increased the complexity and cost of our infrastructure and it was only able to support dedicated content.
With the migration of the complete BW to HANA there were several benefits: First of all we no longer needed BWA. This led to a cost reduction in our infrastructure as well as simplifying our operations. Secondly, with HANA, now all contents in BW run in-memory and significantly faster. This leads to the fact that there is no need to develop and maintain aggregates anymore. This was a huge benefit for us since aggregates were once absolutely necessary to run complex BW queries with an acceptable response time. The downsides of aggregates are how complex they are to create and manage.
With HANA under BW, we no longer need to have aggregates anymore. This tremendously simplifies our BW operations and the data architecture overall.
This is part three in a series ‘SAP’s Road to HANA’. See part one (Sales pipeline on HANA) and part two (Shortening Quarter End with the COPA Accelerator) here!
Very nice article and series Martin. Thanks for the insights.
I got so many questions from all over the place how SAP is adopting HANA, therefore I thought if was a good idea to share my view on this.
Yes, it's very good idea.