Sunday, June 29, 2014

BW ON HANA Performance



BW ON HANA Performance

These figures were reported by customers, which already have BW on HANA operating in their IT landscape.

1. Red Bull
Red Bull started to run its SAP NetWeaver 7.3. Business Warehouse on HANA in 2011.

The migration.
Non-disruptive fast migration.
Project completed in less than 2 weeks.
Database size before migration: 1.5 TB; reduced by 80% after migration.
The result:
Boost in load performance.
Simplified architecture: InfoCubes no longer required as an aggregation layer
Data can be read directly out of the DataStore Object (DSO).
Replication of data to InfoCubes and Business Warehouse Accelerator (BWA) Indexes no longer necessary.

2. Utilities industry representative
This utilities customer also upgraded to and operated its SAP Netweaver 7.3. BW on the latest HANA database release.

Data Load Performance
For DSO activation of 5.2 million records, performance improved by factor 32x. Before: 21 hours 40 minutes - After: 40 minutes.
Data load into write-optimized DSO write optimized and InfoCube Upload with 50K invoice header + 350K items. Load acceleration by factor 2.7x. Before: 1hour 30 minutes -  After: 30 minutes.
 Query Performance
Query with aggregated result set: Performance dramatically improved by factor 471x for queries on aggregated data. Before: 471 seconds - After: 1 second.
Performance acceleration by factor 5x on granular data sets. Before: 308 seconds - After: 64 seconds.
Massive Data Volume Reduction
From xRDBMS space of 4.3 TB reduced to 0.73 TB RAM on HANA
Average compression factor (column tables) 5.8x

3. Automotive industry representative:
This automotive customer upgraded and migrated their data on BW 7.3 on HANA.

Data Load Performance:
InfoCube loading speeds up about 1.39x.
DSO activation accelerated 4.15x
Query Performance
Compared to BW on a traditional relational Database Management System (DBMS)
BW queries were 8.5x faster
BO WebI execution, 11.6x faster
Response times with 40 millions data records in HANA-optimized InfoCube: 45 seconds; 26 times faster compared to the legacy DBMS.