AutoRaid Model 12 or 12 H

[Steve White_8]

How can I determine what physical areas a LUN occupies on an AutoRaid. Does
LUN0 reside at the front of the drives and then work its way progressively
deeper into the drives cylinders? Then does Raid 0/1 information occupy the
area that is defined as unallocated avail for LUNs?

P.S. This thread has been moved from General to Disk Array. - Hp Forum moderator

[Erwin Dillard]

Steve, The operating system of the AutoRAID treats each LUN as if it were an
individual disk. However, in the array there is no correlation between the LUN
and any specific disk modules that are in the array. Since LUNS have no
association with the physical characteristics of the array, there are no
constraints on where the data in a given LUN can be stored within the array.
This gives the controller freedom to make placement decisions as the current
system workload warrants. One of the key concepts of HP's AutoRAID is that all
stored information is distrubuted evenly across all of the disks in the array.
So when a LUN is created it encompasses all of the disks to improve performance
and to maintain data redundancy. The AutoRAID automatically balances user data
to all disks for optimal performance.
As more and more data is written to the array, the controller calculates an
upper limit for RAID 0/1 storage. When this limit is reached then data is
written to RAID 5. The controller will constantly monitor and keep most
frequently used data in RAID 0/1 and less used data in RAID 5 (faster access
with RAID 0/1). The controllers will ensure there will never be more than 90%
of the data in RAID 5 and never less than 10% in RAID 0/1.
Unallocated space will be used by the controller for RAID 0/1 space if needed.
Example: with write intensive data bases it is advantageous to leave some
unallocated space to increase performance.

[Howard Curtis_1]

Erwin, (I'm actually in the US, not URUGUAY - small profiling problem, I guess)

When trying to determine if you've got an I/O constraint, how meaningful is the
performance data coming out of the HP Measureware product as it relates to
device busy, I/O rates and queues - when the devices being reporting are
defined to an AutoRAID box ? Is there any data available which would gives us
an idea of what I/O rates we could expect to achieve before we start seeing
contention ? What would be the appropriate approach to monitor I/O performance
for AutoRAID disk subsystems ?

Thanks... -Howie