Home > Storage > PowerVault > Guides > Dell PowerVault ME5 Series: VMware vSphere Best Practices > Recommended multipathing (MPIO) settings
Block storage (iSCSI, FC, or SAS to vSphere hosts from the PowerVault ME5 array) has the native path selection policy (PSP) of most recently used (MRU) applied by default. If the vSphere hosts connect to the PowerVault ME5 array through SAN fabrics that follow the best practices described in the Connectivity considerations section, multiple paths are presented to each volume. Half of the paths go to the active controller that owns storage pool from which the volume is created, and the remaining paths go to the passive failover or alternative controller. Since the PowerVault ME5 array is ALUA compliant and is recognized as such by vSphere ESXi, I/O is correctly routed to the owning or active controller.
With MRU, only one of the two or four paths to the active controller transports I/O, and the remaining path or paths only transport I/O if current path fails. Changing the PSP to round robin (RR) enables the I/O workload to be distributed across all the available paths to the active controller, resulting in better bandwidth optimization. We recommend using round robin for SAN-attached volumes.
Besides changing the PSP to round robin, we also recommend changing the default number of I/Os between switching paths. The default setting waits until 1,000 I/Os are sent before switching to the next available path. This setting may not fully utilize the entire available bandwidth to the SAN when multiple paths are available. We recommend changing the default number of I/Os between switching paths from 1,000 to 1, as described in the following subsections.
Note: In direct-attached configurations, such as SAS and direct-attached Fibre Chanel, there are typically only two connections, one to each controller, and two paths. In such a configuration, with only one path to the active or owning controller, round robin has no benefit over MRU.
There are other ways that you can apply these setting to all the datastores mounted to all the ESXi hosts in a vSphere environment. Two examples are shown in the following subsections.
Modifying the SATP claim rule is advantageous because it applies to all current and future datastores that are added to the ESXi host, but it requires a reboot. After you create the rule and perform a reboot, all current, and future datastores have the recommended setting applied to them.
To automatically set multipathing to round robin, and set the IOPS path change condition for all current and future volumes mapped to an ESXi host, create a claim rule with the following command:
esxcli storage nmp satp rule add --vendor "DellEMC" --model "ME5" --satp "VMW_SATP_ALUA" --psp "VMW_PSP_RR" --psp-option "iops=1" --claim-option="tpgs_on"
SATP claim rules cannot be edited; they can only be added or removed. To change an SATP claim rule, you must remove and readd it. To remove the claim rule, issue the following command:
esxcli storage nmp satp rule remove --vendor "DellEMC" --model "ME5" --satp "VMW_SATP_ALUA" --psp "VMW_PSP_RR" --psp-option "iops=1" --claim-option="tpgs_on"
Perform a reboot for the claim rule changes to take effect.
You can change the PSP from MRU to round robin using the VMware vSphere Web client for each existing datastore. However, this process can become unwieldy if there are many existing new datastores that require changes. It also does not permit changing the number of I/Os between switching paths. It is often more convenient for an administrator to write a short script to change all the devices on a particular host. For more information about the commands required, see VMware KB article 1017760.