SAP NetWeaver BW Powered by SAP HANA Scale Out – Best Practices
Update with: SAP NetWeaver BW 7.30 SP5 / BW 7.40 SP5 on HANA and more current versions
Learn more about how to configure Configure SAP HANA Table-Distribution for BW-on-HANA Scale-Out Systems in an how to guide:
Scale Out Sizing
If memory requirements exceed the available memory of a single server node, a scale out solution consisting of multiple server nodes can be deployed. The architecture of a scale-out solution consists of a master node which stores the row-store data, and several slave nodes which hold partitions of business data tables (InfoCubes, DSOs, PSA tables). By increasing the number of slave nodes, additional storage capacity must be added to the system as needed.
For more detailed information about scale out in general please refer to “SAP Note 1637145 – SAP BW on HANA: Sizing SAP In-Memory Database”. You’ll find there also a report which helps you to estimate the necessary amount of main memory for your BW on HANA scale out landscape.
- It is strongly recommended to have a minimum of three nodes for a HANA scale out database available. In two nodes landscapes the table distribution and the parallel process degree would not be benefitial. “See SAP Note 1702409 Optimal number of scale out nodes for BW on HANA”
- In a distributed SAP HANA database you’ll have a dedicated server (called “master”), which handles the work load of the NetWeaver Stack, the statistics and the tables located in the row store. In case of 1 TB nodes the master node allocation should not exceed 512 GB. The biggest tables within the row store are usually statistic and basis tables, therefore a strong housekeeping concept is absolutely important to keep the allocation below 50 % of the node size. Please see also the following SAP Note 1855041 – Sizing Recommendation for Master Node in BW-on-HANA
Check latest hardware appliance updates and related SAP Notes
- Use only certified hardware and server configurations: http://service.sap.com/pam
- Contact your HW partner for latest patch sets and necessary improvements according to your hardware environment
- Establish SAP Support Connectivity and Partner connection, to ensure proper access for SAP in case of questions or open messages.
- Please consider for that the latest SAP Notes and some of interest below:
- SAP Note 1635304 – Central note for HANA support connections
- SAP Note 1634848 – SAP HANA database service connections (collective note)
- SAP Note 1058533 – TREX/BIA/HANA service connection to customer systems
- SAP Note 1592925 – SAP HANA Studio service connection
- SAP Note 1610702 – HANA Support processed and procedures
- SAP Note 1067044 – How to use the TREX/BIA connection
- SAP Note 1651638- How to establish WTS remote support connections – Data Services
- SAP Note 605795 – Windows Terminal Server connection in remote support
- SAP Note 35010 – Service connections: Composite note (overview)
Check appliance and host name setup:
Since scale-out landscapes are distributed systems, the communication between the different slaves is very important. You’ll find below some points to ensure that the setup is working correctly:
- For Standard scale out deployments:
- Since various internal and external host names are used for an SAP HANA system, problems occur during the setup of the connection of clients to the database, please see and check the following SAP Note to ensure that the host names are set correct in your scale out landscape: “SAP Note 1780950 – Connection problems due to host name resolution”
- In case of storage mirroring Disaster Tolerance solution please consider the following SAP Note: 1743225 – HANA: Potential failure of connections with scale out nodes
- In case of error messages:
- Note 1771873 – Index server crashes with traces displaying message ‘Too many open files’
Configure failover for BW on HANA scale out
- The detailed steps how to create and configure a failover setup for BW on HANA scale out can be found in the SAP HANA Administration Guide: http://help.sap.com/hana/hana_admin_en.pdf – Chapter 15.3.1 Configuring Clients for Failover
Upgrade of SAP NetWeaver BW before migration
It is very important to upgrade your starting SAP NetWeaver BW system to latest support packages before the technical export to SAP HANA. The latest
Support Packages always includes a lot of optimizations, especially for scale out landscapes, which has to be available in front of the export to SAP HANA.
Please lease find below the current recommendable releases, in general choose the latest one:
- SAP NetWeaver BW 7.30 SP8 or 7.31 SP5
- SAP NetWeaver BW 7.30 SP9 or 7.31 SP6
After the migration to HANA the following collection notes, according to the relevant support package should be considered or as recommended the latest support package implemented:
- SAP Note 1712334 – Important Notes for SAP BW powered by HANA on SP7
- SAP Note 1769670 – Important Notes for SAP BW powered by HANA on SP8
- SAP Note 1846493 – Important Notes for SAP BW powered by HANA on SP9
Important application- checks before migration
Use the automated check tool to check your system prior of the migration if certain requirements for migration are fulfilled
- SAP Note 1729988 – SAP NetWeaver BW powered by SAP HANA – Checklist Tool: Automate the check of best practice guidelines for operations and prerequisites for migration of an existing SAP NetWeaver BW deployment to the SAP HANA platform. Attachment to the note contains an ABAP program to perform checks in a matter of minutes that would otherwise take hours or days to execute manually. The output of the program is self-descriptive.
- This tool includes also a lot of the mentioned checks and perquisites of this blog here, so if you are considering to migrate to a scale out landscape, the tool is highly recommendable.
- Inventory InfoCubes (including non-cumulative key-figures) should be compressed on a regular basis, which is a common recommendation to keep reading performance on a stable level. Especially before the migration to a scale out landscape this should be done. If you have uncompressed request in a scale out landscape please consider the following “SAP Note 1773398 – SAP HANA: No compression for scale-out & non-optimized cube”
SAP HANA Support Package Stack:
- Please use latest SAP HANA support package stack and revision before the technical migration
- Please use and install latest SAP HANA Client according to the database version
In general the known steps and processes for SAP NetWeaver BW migrations / installation have to be considered:
- Please find the latest information about installing and implementing BW on HANA on the SAP service market place, please see attached link: https://websmp206.sap-ag.de/instguidesnw73 – Consider especially latest guides and SAP Notes for the migration process.
- SAP HANA related master and upgrade guides can be found on the SAP Help portal, please see attached link: http://help.sap.com/hana_appliance/
- Best practices and more information are available here: http://www.saphana.com/docs/DOC-2102
In addition to that, there are specific scale-out topics to consider:
- SAP Note 1775293 – Migration/system copy to SAP HANA using latest SWPM 1.0
- SAP Note 1776186 – SAP HANA BW – Scale out: routing to right indexserver inthe BW on HANA Scale-Out System
- SAP Note 1783238 – SMIGR_CREATE_DDL: Estimated Row Count für Faktentabellen
- SAP Note 1741632 – Basic corrections for the heterogeneous system migration
- Use SWPM 1.0 + Latest Kernel, DBSL etc. which is mentioned in the Notes above and fitting to the SAP HANA Database
- Check if table_placement is set in accordance with “Note 1715048 – BW 7.30 new features for installation or migration” or SWPM 1.0 is used as mentioned in the Note
- Implement and check “SAP Note 1815547 – Row/ColumnStore Check ohne rowstorelist.txt” before the migration in order to generate the latest rowstorelist.txt
- Please follow the instructions of SAP Note 1819123 – SAP HANA Landscape Reorg to ensure an optimized and correct table distribution across the nodes.
Make sure that all post migration steps via RS_BW_POST_MIGRATION according to the BW Upgrade guides are executed and finished successfully.
After migration & for ongoing projects
Use HANA-optimized objects
- Convert your BW InfoCubes & DataStore Objects to HANA-optimized, according to guidelines from: http://www.saphana.com/docs/DOC-3129
- InfoCubes: Since there are no limitations or drawbacks comparing the classic InfoCube with the HANA-optimized InfoCube, we recommend converting all InfoCubes to the HANA-optimized version to benefit from the better load performance and the more flexible design. Additionally the InfoCube-conversion runs as a purely HANA-internal procedure and is therefore in most cases very fast. But the conversion can be and should be done step-by-step or project-by-project. There is no need to run this as part of the complete DB-migration project, but it can be done in appropriate time-windows later. Remark: For InfoCubes with inventory key figures, please consider note 1780575.
- DataStore Objects: Please consider the limitations for when it is possible to convert a DSO to the HANA-optimized version (see the SAP online help). We do NOT recommend converting all DSOs to the HANA-optimized version, but only those DSOs should be converted where the DSO activation is time consuming and part of a performance critical load process. Additionally we recommend converting these DSO w/o Change log whenever possible. All other DSOs should stay the classic DSO type. Significant performance improvements for the activation process for the classic DSOs are also in the pipeline and will be made available as part of future HANA and BW Support Packages.
Table-redistribution in scale-out landscape
Some BW objects, particularly DataStore objects (DSOs) and Cubes are currently as a standard created in a scale-out solution on an SAP HANA database with as many partitions as the number of slave servers that are available in the system. To ensure an optimal distribution in terms of memory allocation it is recommended, starting with HANA Revision 52, to consider the splitting parameters of the following SAP Note
- SAP Note: 1819123 – SAP HANA Landscape Reorg
These parameters can be set during migration via SWPM or for running HANA systems by SQL statements in HANA Studio. Especially new installations / migration should directly leverage this new distribution.
This partition criteria are moreover important for very large tables in BW (> 2 billion records). In such cases a partionning will take place during mirgration into a multi node landscape according to the object type. Get more information here: http://www.saphana.com/community/blogs/blog/2013/04/15/bw-on-hana-and-very-large-tables
More information on HANA landscape redistribution with SAP HANA SP6 can be found in the excellent blog here.
With SAP HANA SP6 more over the following SAP Notes should be considered: SAP Note 1908073, SAP Note 1908075, SAP Note 1908082, SAP Note 1908133.
For Business Planning and Consolidation on NetWeaver 7.30
Scale Out optimizations / corrections
- SAP is constantly checking and improving the table distribution for HANA landscapes therefore the following SAP Note should be checked from time to time in order implement the latest improvements regarding column-store / row-store based tables: “Note 1815547 – Row/ColumnStore Check ohne rowstorelist.txt”
- High-cardinality InfoObjects are used in a BW query based on DataStore Objects see Note 1835020 – Long BW query runtimes due to joins of big masterdata
- To perform zero-elimination during InfoCubes compression, whereby the InfoCube is distributed across several notes use report included in the following SAP Notes: 1811659 – SAP HANA InfoCube: Cleaning up duplicates in compressed data
This collection of BW on HANA scale out considerations and best practices will be constantly updated!