HP Integrity VM Accelerated Virtual I/O Overview

The label can be retrieved using:
scsimgr get_devid -D /dev/rdisk/disk49
For example:
scsimgr set_devid “guest_vme02” –D /dev/rdisk/disk49 sets the device identifier for disk49
to indicate it is part of guest vme02.
Guest SCSI Queue Depth Tuning
SCSI queue depth defines the maximum number of concurrent I/O requests that could be outstanding
for a device. The default value is 8. To alter the value of scsi queue depth, change the
“max_q_depth” attribute using scsimgr(1M). Please refer to scsimgr(1M) for details of how to change
the attribute either on a global, vendor or device basis.
In general, setting the value of the attribute should be based on the capability of the actual device.
Please consult the device configuration guide to find the recommended scsi queue depth for the host
backing store. In general, the sum of the queue depths across all guests sharing the device should
match the queue depth of the device in the host.
For raw disk backing stores used by only one guest, the queue depth should be same as it is in the
host. If the disks are shared, the value could be tuned to different values for each guest, depending on
the performance requirement of individual guests.
For logical volume backing stores, the tuning depends on how many disks are used in the creation of
volume group (vg) and how many logical volumes of the volume group are in use by the guest. SCSI
queue depth may be tuned up or down to control the number of I/Os reaching the end device to a
optimal value.
For file backing stores, tuning of max_q_depth is not recommended. File backing store performance is
tuned to work optimally with the default (8) max_q_depth.
Changing the max_q_depth attribute on a system wide basis is not recommended as different backing
stores/devices have different I/O capabilities.
Timing and Retry Policies
The scsimgr utility enables the user to set several tunables for retry and timeout policies (see
scsimgr_esdisk(7)), for example, infinite_retries_enable, max_retries, path_fail_secs. These tunables
must not be altered in the host as it will interfere with Integrity VM 4.0 storage operations.
Multipathing Tunables
HP-UX 11iv3 supports multi-pathing even using legacy device names. Multi-pathing using legacy DSFs
is controlled by the scsimgr tunable “leg_mpath_enable” which is set by default. This tunable should
not be altered in the host.
Useful HP-UX 11iv3 commands
HP-UX 11iv3 has added several new commands and options to existing commands to enable easier
storage management. Some of the commands are mentioned below. Additionally the load balancing
algorithms for multi-path device, for example, round-robin, least command load, etc. can be set using
scsimgr. Refer to the HP-UX 11iv3 Storage Area Management whitepapers for further information.