Kris wrote: "Wouldn't it be better to let the RAIDing be done by the HSZ instead of by software? In my experience the HSZs do a pretty good job, especially at RAID 5."
Although the original design intent of the HP (formerly StorageWorks) RAID Software for OpenVMS was to do RAID-5 in host software (and it accomplished that about a year before controllers which could do it in firmware/hardware became available), and it could provide RAID-5 protection for plain, ordinary local disks which weren't behind an expensive array controller, eventually array controllers with RAID became readily available and relatively inexpensive, and a host-based product incurs the overhead of extra I/Os required to update parity metadata so as to cover the "write hole" in RAID-5 writes (extra I/Os which weren't needed in controller [pair]s with mirrored non-volatile write-back cache which could be used to hold RAID-5 metadata during writes) and this hurt host-based RAID-5 performance, so use of the product for RAID-5 has tended to decrease over time.
But RAID-0 (disk striping) capabilities were added to the product along the way, so in fact today the most common use in practice for the product is to do disk striping (RAID-0) or RAID 0+1 (stripesets of shadowsets). This allows I/O workload to be spread across multiple controllers, so that I/O loads can be handled which are greater than any single controller can handle.
Some customers use the software to combine a number of relatively smaller disk volumes (which can all shadow-copy in parallel, each with their own copy thread, and thus complete faster) into larger volumes for use by VMS or applications.
It also has the capability of doing partitioning (either of an array of disks, or even of a single individual disk), for folks who need that capability on VMS.
