Data Protection for SAP uses the IBM Storage Protect™ archive function for all backup activities. When you are setting up the IBM Storage Protect server for use with IBM Storage Protect for Enterprise Resource Planning, the following considerations help you to optimize performance when you set up the IBM Storage Protect server.
- Dedicated backup server
- A dedicated backup server allows sharing of resources and provides an efficient resource usage.
- CPU power
- For a specific data throughput, the CPU load on the backup server is approximately 60% of the load on the database server. Therefore, backup server CPU power is not as critical as the CPU power of the database server. However, demands on the IBM Storage Protect server CPU do increase when several clients access a single IBM Storage Protect server.
- Storage hierarchy
-
Not following these requirements can lead to recovery issues and a deadlock situation.
The specific interaction of current SAP HANA versions with IBM Storage Protect for Enterprise Resource Planning: Data Protection for SAP HANA implies special requirements for the type of storage media that are used, and the rules for data movement in IBM Storage Protect storage. The requirements are different for data files and for log files: you must always use separate management classes and storage pools for BRBACKUPMGTCLASS and BRARCHIVEMGTCLASS.
-
- Data files
- For best restore performance it is important that files that were backed up simultaneously, are held ready for parallel access during restore. By internal data movement to physical or virtual tape in IBM Storage Protect storage after the backup, files that are bound to be restored in parallel can end up on the same volume.
The following suggestions help to avoid a situation that would lead to an increase of the restore duration by media wait. Use devclass disk, sequential file, directory-container storage, or physical tape, Virtual Tape Library (VTL) as the first storage pool for BRBACKUPMGTCLASS.
- Do not move any SAP HANA data files in IBM Storage Protect storage from the first stgpool to physical tape or VTL storage. This rule is the case for storage pools on random disk, sequential file, directory-container, physical tape, or VTL storage.
- Do not use nextstgpool to point to storage other than random disk or sequential file or directory-container to prevent migration to physical tape or VTL.
- When you use physical tape or VTL storage as the first
stgpool for HANA data files, do not run space reclamation on this storage pool.
- Do not use move data from the first
stgpool to physical tape or VTL storage, regardless of the devtype of the first storage pool devclass.
-
- Log files
- Always use
devclass disk, sequential file, or directory-container storage as the first storage pool for BRARCHIVEMGTCLASS. Storage pools that are receiving HANA log files do not require more space allowance beyond the real amount of data to be kept in IBM Storage Protect. The estimated size that is assumed by Data Protection for SAP HANA is expected to be sufficiently close to the real log file size.
- You can reduce the space consumption in used
devclass by using compression. For this specific purpose, the IBM Storage Protect API client parameter COMPRESSIon Yes in dsm.sys is expected to be more efficient than the parameter RL_COMPRESSION YES in initSID.utl.
- Do not move any HANA log files in IBM Storage Protect storage to physical tape or VTL storage. Further considerations for this rule are as follows.
- Only random disk or sequential files or directory-container are allowed as defined in
nextstgpool for migration.
- Run move data actions only to random disk or sequential file or directory-container.
- If HANA log files are stored on physical tape or VTL, you must move the data to sequential file before the recovery operation.
- Parallel sessions
- The IBM Storage Protect server allows the use of several tape drives in parallel to store data. This setup can increase overall data throughput. To fully use this feature, two conditions must exist. The corresponding IBM Storage Protect node must be allowed the appropriate number of mount points and the device class must be allowed the appropriate mount limits.