Inline deduplication, cross- volume inline deduplication, inline compression, and inline compaction are supported with SAP HANA in an SSD configuration.

Enabling the storage efficiency features in an HDD configuration is not supported.

The use of NetApp Volume Encryption (NVE) and NetApp Aggregate Encryption (NAE) are supported with SAP HANA.

QoS can be used to limit the storage throughput for specific SAP HANA systems. One use case would be to limit the throughput of development and test systems so that they cannot influence production systems in a mixed setup.

During the sizing process, the performance requirements of a nonproduction system must be determined. Development and test systems can be sized with lower performance values, typically in the range of 20% to 50% of a production system.

Starting with ONTAP 9, QoS is configured on the storage volume level and uses maximum values for throughput (MBps) and number of I/O (IOPS).

Large write I/O has the biggest performance effect on the storage system. Therefore, the QoS throughput limit should be set to a percentage of the corresponding write SAP HANA storage performance KPI values in the data and log volumes.

NetApp FabricPool technology must not be used for active primary file systems in SAP HANA systems. This includes the file systems for the data and log area as well as the /hana/shared file system. Doing so results in unpredictable performance, especially during the startup of an SAP HANA system.

Using the “snapshot-only” tiering policy is possible as well as using FabricPool in general at a backup target such as SnapVault or SnapMirror destination.

FabricPool is a good option for moving log backups to another storage tier. Moving backups affects the time needed to recover an SAP HANA database. Therefore, the option “tiering-minimum-cooling-days” should be set to a value that places log backups, which are routinely needed for recovery, on the local fast storage tier.

The following overview summarizes the required storage configuration steps. Each step is covered in more detail in the subsequent sections. Before initiating these steps, complete the storage hardware setup, the ONTAP software installation, and the connection of the storage FCP ports to the SAN fabric.

With HDDs, a maximum of two DS2246 disk shelves or four DS224C disk shelves can be connected to one SAS stack to provide the required performance for the SAP HANA hosts, as shown in the following figure. The disks within each shelf must be distributed equally to both controllers of the HA pair.

With SSDs, a maximum of one disk shelf can be connected to one SAS stack to provide the required performance for the SAP HANA hosts, as shown in the following figure. The disks within each shelf must be distributed equally to both controllers of the HA pair. With the DS224C disk shelf, quad-path SAS cabling can also be used but is not required.

In general, you must configure two aggregates per controller, independent of which disk shelf or disk technology (SSD or HDD) is used. This step is necessary so that you can use all available controller resources. For FAS 2000 series systems, one data aggregate is sufficient.

Multiple-host SAP landscapes with SAP HANA databases can use a single SVM. An SVM can also be assigned to each SAP landscape if necessary in case they are managed by different teams within a company. The screenshots and command outputs in this document use an SVM named hana.

Within the storage cluster configuration, one network interface (LIF) must be created and assigned to a dedicated FCP port. If, for example, four FCP ports are required for performance reasons, four LIFs must be created. The following figure shows a screenshot of the four LIFs (named fc_*_*) that were configured on the hana SVM.

During SVM creation with ONTAP 9.8 System Manager, all the required physical FCP ports can be selected, and one LIF per physical port is created automatically.

The following figure depicts the creation of SVM and LIFs with ONTAP 9.8 System Manager.

An FCP port set is used to define which LIFs are to be used by a specific igroup. Typically, all LIFs created for the HANA systems are placed in the same port set. The following figure shows the configuration of a port set named 32g, which includes the four LIFs that were already created.

An igroup can be configured for each server or for a group of servers that require access to a LUN. The igroup configuration requires the worldwide port names (WWPNs) of the servers.

Using the sanlun tool, run the following command to obtain the WWPNs of each SAP HANA host:

The following figure shows the list of initiators for SS3_HANA. The igroup contains all WWPNs of the servers and is assigned to the port set of the storage controller.

The following figure shows the volume configuration of four single-host SAP HANA systems. The data and log volumes of each SAP HANA system are distributed to different storage controllers. For example, volume SID1``data``mnt00001 `is configured on controller A and volume `SID1``log``mnt00001 is configured on controller B. Within each volume, a single LUN is configured.

For each SAP HANA host, a data volume, a log volume, and a volume for /hana/shared are configured. The following table shows an example configuration with four SAP HANA single-host systems.

The next table shows an example of the mount point configuration for a single-host system.

The Linux LVM can be used to increase performance and to address LUN size limitations. The different LUNs of an LVM volume group should be stored within a different aggregate and at a different controller. The following table shows an example for two LUNs per volume group.

At the SAP HANA host, volume groups and logical volumes must be created and mounted. The next table lists the mount points for single-host systems using LVM.

The following figure shows the volume configuration of a 4+1 multiple-host SAP HANA system. The data volumes and log volumes of each SAP HANA host are distributed to different storage controllers. For example, the volume SID``data``mnt00001 is configured on controller A and the volume SID``log``mnt00001 is configured on controller B. One LUN is configured within each volume.

The /hana/shared volume must be accessible by all HANA hosts and is therefore exported by using NFS. Even though there are no specific performance KPIs for the /hana/shared file system, NetApp recommends using a 10Gb Ethernet connection.

For each SAP HANA host, a data volume and a log volume are created. The /hana/shared volume is used by all hosts of the SAP HANA system. The following figure shows an example configuration for a 4+1 multiple-host SAP HANA system.

The next table shows the configuration and the mount points of a multiple-host system with four active SAP HANA hosts.

The Linux LVM can be used to increase performance and to address LUN size limitations. The different LUNs of an LVM volume group should be stored within a different aggregate and at a different controller. The following table shows an example for two LUNs per volume group for a 2+1 SAP HANA multiple host system.

At the SAP HANA host, volume groups and logical volumes need to be created and mounted:

The volume options listed in the following table must be verified and set on all SVMs.

You can use NetApp OnCommand System Manager to create storage volumes and LUNs and the map them to the igroups of the servers.

The following steps show the configuration of a 2+1 multiple-host HANA system with the SID SS3.

This section shows an example configuration using the command line with ONTAP 9.8 for a 2+1 SAP HANA multiple host system with SID FC5 using LVM and two LUNs per LVM volume group.