Automatic load balancing provides improved I/O resource management by reacting dynamically to load changes over time and automatically adjusting volume controller ownership to correct any load imbalance issues when workloads shift across the controllers.

The workload of each controller is continually monitored and, with cooperation from the multipath drivers installed on the hosts, can be automatically brought into balance whenever necessary. When workload is automatically re-balanced across the controllers, the storage administrator is relieved of the burden of manually adjusting volume controller ownership to accommodate load changes on the storage system.

When Automatic Load Balancing is enabled, it performs the following functions:

Automatically monitors and balances controller resource utilization.
Automatically adjusts volume controller ownership when needed, thereby optimizing I/O bandwidth between the hosts and the storage system.

Enabling and disabling Automatic Load Balancing

Automatic Load Balancing is enabled by default on all storage systems.

You might want to disable Automatic Load Balancing on your storage system for the following reasons:

You do not want to automatically change a particular volume’s controller ownership to balance workload.
You are operating in a highly tuned environment where load distribution is purposefully set up to achieve a specific distribution between the controllers.

Host types that support the Automatic Load Balancing feature

Even though Automatic Load Balancing is enabled at the storage system level, the host type you select for a host or host cluster has a direct influence on how the feature operates.

When balancing the storage system’s workload across controllers, the Automatic Load Balancing feature attempts to move volumes that are accessible by both controllers and that are mapped only to a host or host cluster capable of supporting the Automatic Load Balancing feature.

This behavior prevents a host from losing access to a volume due to the load balancing process; however, the presence of volumes mapped to hosts that do not support Automatic Load Balancing affects the storage system’s ability to balance workload. For Automatic Load Balancing to balance the workload, the multipath driver must support TPGS and the host type must be included in the following table.

For a host cluster to be considered capable of Automatic Load Balancing, all hosts in that group must be capable of supporting Automatic Load Balancing.

Host type supporting Automatic Load Balancing With this multipath driver

Host type supporting Automatic Load Balancing	With this multipath driver
Windows or Windows Clustered	MPIO with Fujitsu ETERNUS AB/HB Series DSM
Linux DM-MP (Kernel 3.10 or later)	DM-MP with `scsi_dh_alua` device handler
VMware	Native Multipathing Plugin (NMP) with `VMW_SATP_ALUA Storage System Type` plug-in

Windows or Windows Clustered

MPIO with Fujitsu ETERNUS AB/HB Series DSM

Linux DM-MP (Kernel 3.10 or later)

DM-MP with scsi_dh_alua device handler

VMware

Native Multipathing Plugin (NMP) with VMW_SATP_ALUA Storage System Type plug-in

With minor exceptions, host types that do not support Automatic Load Balancing continue to operate normally whether or not the feature is enabled. One exception is that if a system has a failover, storage systems move unmapped or unassigned volumes back to the owning controller when the data path returns. Any volumes that are mapped or assigned to non-Automatic Load Balancing hosts are not moved.

Contact our support for compatibility information for specific multipath driver, OS level, and controller-drive tray support.

Verifying OS compatibility with the Automatic Load Balancing feature

Verify OS compatibility with the Automatic Load Balancing feature before setting up a new (or migrating an existing) system.

Contact our support to find your solution and verify support.

If your system is running Red Hat Enterprise Linux 6 or SUSE Linux Enterprise Server 11, contact our support.
Update and configure the /etc/multipath.conf file.
Ensure that both retain_attached_device_handler and detect_prio are set to yes for the applicable vendor and product, or use default settings.