Dell POWERVAULT MD3620I - Manual (Page 21)

Figure 10. Configuration Showing Double-Point Failure
While this method works well in the case of hardware failure, MRU only maintains one active path for each HBA group.
Therefore, if you have an ESX/ESXi host with four HBAs, only one HBA is active at a time.
Grouping HBAs And Creating Virtual Hosts
To achieve higher I/O throughput from the host to the storage array, group the HBAs in pairs and create virtual hosts for
each pair of HBAs. This allows for a fully redundant configuration, but also allows for two of the HBAs to be active at the
same time. From the storage array, the second pair of HBAs is defined as a separate host. Then, virtual disks can then
be mapped directly to the new host or host group. This same methodology can be used to group additional HBAs in the
same manner.
Managing Bandwidth
Grouping HBAs in pairs and using virtual disks to create a fully redundant configuration does require additional
management to balance the LUNs between the hosts/host groups and fully use available bandwidth between all HBA
groups. When this method is used in a fibre channel configuration, the Automatic host configuration utility cannot
determine which HBAs are configured to each fabric zone. Instead, the administrator must verify that a single HBA is
connected to both fabric zones for each HBA pair group.
21