This chapter describes how to customize your IBM Spectrum Scale for HANA, including the file system and VDisk layout.

This chapter includes the following topics:

The following levels of changes are needed to configure an optimized environment for SAP HANA workloads:

Because of the different IO workload behaviors of SAP HANA, we create one file system for the log workload and one file system for data workload.

During the tests (which are eligible for any customer environment), you can share data and log file system among many security identifiers (SIDs).

Optionally, you can consider creating file sets to separate the SIDs from each other.

A third file system for the SAP named shared includes no special recommendations and can be created without any special recommendations.

For more information about creating a shared file system, see Appendix E, “Side aware configuration examples” on page 43.

The minimum normally required settings are not described further here because they are set by the default deployment procedure. This document focuses only on the parameters that you need to change for an optimized SAP HANA environment.

During the installation and setup of your HANA environment, you can customize your file system and IBM Spectrum Scale device names to your own naming conventions (see Figure 5-1). for more information about how to set up and configure your IBM Spectrum Scale file systems, see the Recommended File System Layout page of the SAP website.

Now adjust a pre-configured ESS (default deployment) for an SAP HANA workload. Depending on the chosen model, you can proceed with the default settings, which are deployed by using the default deployment procedure. With the default settings, you cannot scale enough to exceed more than four HANA databases in parallel. If you want to exceed this limit, you must increase the size of the VDisk loghome to 100 GB (the default is 20 GB).

Increasing the size of VDisk loghome is not possible and the loghome VDisk cannot be re-created if any other customer VDisks are available. A prepared sample file is found in the common /usr/lpp/mmfs/samples/vdisk directory.

Rename the sample and edit the file according to the example that is shown in Example 5-1.

A larger loghome is always good to use for better performance. The more loghome that is available, the more pagepool is need. If your loghome is configured to “large” in terms of existing pagepool and buffer sizes, it is not fully used.

Create your VDisks for the ESS Building block by using the mmcrvdisk -F yourfilename command. The Spectrum Scale daemons must be running to create the VDisk.

The block size in the IBM Spectrum Scale file system is an unchangeable value. Therefore, select the block size from the beginning. SAP HANA has three core file systems with different needs:

As described in 5.2.1, “Internal VDisks” on page 27, create a VDisk stanza file and edit it according to your naming conventions (see Example 5-2). If you are creating a highly available or site aware configuration, add a failureGroup (fg) identifier for each VDisk in the file. The fg must be the same within a building block when you want to use IBM Spectrum Scale replication later on and mirror your data between two ESS building blocks. If you have more than one building block per site, the failure group identifier must be the same over all of the ESSs in the same site. For more information about a highly available scenario configuration example, see Appendix E, “Side aware configuration examples” on page 43.

Create the VDisk by using the mmcrvdisk command. Create the NSDs by using the mmcrnsd command.

Next, create the file system with the parameters that are shown in Example 5-3 (in this example, without IBM Spectrum Scale replication enabled by default).

For more information about a replicated scenario, see Appendix E, “Side aware configuration examples” on page 43.

As described in 5.2.2, “Log file system” on page 28, create the data file system (see Example 5-4).

Create the VDisk by using the mmcrvdisk command. Create the databases by using the mmcrnsd command.

Create the file system with the parameters that are shown in Example 5-5 (in this example, without IBM Spectrum Scale replication enabled by default).

For more information about a replicated scenario, see Appendix E, “Side aware configuration examples” on page 43.

As described in 5.2.3, “Data file system” on page 28, create a VDisk stanza file (see Example 5-6).

Next, create the file system with the parameters that are shown in Example 5-7 (without IBM Spectrum Scale replication enabled by default).

For more information about a replicated scenario, see Appendix E, “Side aware configuration examples” on page 43.

You can create an own set of file systems (log, home, and shared) for each SAP instance (SID). You also can operate in an environment with multiple SIDs sharing systems. If you plan to use multiple SIDs, create file sets for each SID.

During the SAP verification and certification, 12 productive SIDs within the same IBM Spectrum Scale file system in parallel were tested successfully in our example.