Performance improvements in SPM 2.1 over SPM 2.0
I frequently get this question on improvements in SPM 2.1 over SPM 2.0 related to performance, so I thought it would be easier to compile this list and share it with you.
Here are all of the performance improvements in 2.1:
- SPM Simple Cache:
SAP simple cache, as explained in the performance article, is an application specific cache. Pre-calculated result sets are persisted in the in DB on the ABAP side, when the report gets executed the first time. For every subsequent executions of the report (until the underlying data changes with a new upload), this stored result set is used and a call to BW is avoided. Other users executing this report also avail the benefit of this cache as long as their locale and formatting settings are same as the saved report. The table where the results get persisted is OPMDM_PRECALC_A.
- Memory Monitor settings:
The flex memory consumption can be controlled by setting the memory thresholds from within the application. Follow the instructions in the note 1477454.
- Data streaming format:
For the releases 1.0 and 2.0, the data streaming format was plain XML. Starting 2.1, the switch has been made from XML to JSON objects which have a high degree of compression and help reduce network latency related performance losses. And on the Java stack the processing for those JSON objects is much faster compared to XML processing.
- Data Volume:
Rendering really large volumes of data (20,000 records and 4 columns – as an example) is a significant challenge in FLEX. Not only is it slow to load but often times I have experienced that the flash player simply quits. To handle this case, SPM 2.1 works around the flash player limitations and boosts the volume, by upto 2 to 4X. This enhancement not only helps improve handling of large reports but also speeds up the rendering process for smaller reports. Additionally, starting 2.1, we have started using the new (as of Flex 3) feature of OLAP grids to represent reporting data. This has an inbuilt understanding of multidimensional analysis and hence is faster in rendering large number of records.
- Cap on charting for really large values:
When there are too many rows (granular dimensions, like item level reports without filters), then charting of such data does not provide any value. In this case SPM caps the charting mechanism for really large values.
For general performance improvement advice, refer to this article.
And as has been mentioned in other places, we have seen a tremendous improvement in Flash player performance with the 10.1 version. So always get the latest flash player (regardless of the SPM release).