Datasheet

Note: For D5000, DS4000 and DS3000 storage subsystems with dual controllers, a connection

must be made to both of the controllers to allow an active and a failover path. When the volumes

are created on these systems, the host OS type should be set to DEFAULT or AIX (not AIX

ADT/AVT, or failover/failback oscillations might occur). For all storage systems, it is

recommended that the fabric configuration uses separate dedicated zones and FC cables for

each connection.

After the VSCSI client driver in IBM i detects a second path to the same set of LUNs through a

different VIOS LPAR, the disk names change from DDxxx to DMPxxx. This is identical to multi-

path I/O to a directly attached DS8000 subsystem. If a path is lost, the disk names do not change

back to DDxxx, but a path failure message is sent to the QSYSOPR message queue in IBM i. As

soon as the path is restored, an informational message is sent to the same queue. To monitor

the status of the paths in IBM i:

 Start Services Tools and sign in.

 Work with disk units.

 Display disk configuration.

 Display disk path status.

Note that while the HMC allows a single VSCSI client adapter to be tagged as an IPL device, it is

not required to change the tagged resource if that path is lost. As long as the alternate path

through the second VIOS is active, the IBM i LPAR will be able to IPL from the same load source

LUN. The alternate path must exist prior to the loss of the original IPL path and prior to powering

the IBM i LPAR off.

Another option to VSCSI storage, as just described, is NPIV attached storage facilitated through

both VIOS partitions. The sets of commands listed earlier in this section do not come into play

because the VIOS partitions become passthru virtual servers.

6.4 Best practices for Fibre Channel storage

When configuring LUNs for IBM i (virtualized by VIOS), follow the best practices outlined in

chapter 18 of the latest Performance Capabilities Reference manual, available here:

http://www.ibm.com/systems/i/solutions/perfmgmt/resource.html. Note that some of its

recommendations apply only to IBM i using virtual storage outside of the chassis environment.

In addition to the guidelines in the Performance Capabilities Reference manual, follow these

additional recommendations:

 Use Fibre Channel disk drives (and not SATA or FATA) to create the RAID ranks/arrays

for production IBM i workloads

 Use 15K RPM drives for medium and heavy I/O IBM i workloads, and 10K RPM drives for

low I/O workloads

 When creating a host connection to the WWPN of the Fibre Channel card on the node,

specify at most two specific host ports. Do not create the connection so that the Fibre

Channel adapter can connect to all host ports on the storage subsystem, which is the

default for some subsystems.

 Properly zone the switches between the adapter ports and the SAN host ports.

 When configuring vSCSI LUNs for IBM i within VIOS: as soon as the LUNs that will be

virtualized to IBM i have reported in VIOS, and prior to assigning them to the IBM I client,

change their queue depth to improve performance. Start a Telnet session to VIOS and

o lsdev - attr -dev hdisk#, where # is the hdisk number you want to display.

o chdev -perm -attr queue_depth=8 -dev hdisk1#