HP StorageWorks 7110 - Virtual Array Service Manual page 51

continuous high demand from the host will preempt all data migration
activities.
AutoRAID manages the data placement to the individual 256 K-block. Each
LUN is divided into 256 K-blocks call clusters. A cluster can be stored in either
RAID 1+0 or RAID 5DP format. The virtualization data structures manage the
translation of the logical address (LUN) and the physical location.
The controller is programmed to manage cluster placement. It uses well-known
logic, or rules, about RAID level performance characteristics and storage
efficiency. This logic directs data that is frequently modified by small
transactions to RAID 1+0 storage. Data that is infrequently written, or data
that is written sequentially, is directed to RAID 5DP storage.
The behavior is similar to other hierarchical memory systems, such as data
caches or Hierarchical Storage Mangers. AutoRAID, like these other systems,
provide the performance approaching the highest level of the memory
hierarchy, at the cost of the lowest level in the hierarchy.
The controller provides information about data placement and data migration
through the Command View performance log. These logs provide details
about the storage level for each LUN, and any active migration the array has
performed.
End-to-End Data Protection
End-to-end data protection is a process within the array controller to validate
the integrity of the data stored on the array. This process is in addition to the
normal data checking provided by the disk drives. During a write operation,
as data enters the array controller from the host, the controller appends 8 bytes
of additional information to each 512-sector. This additional information
includes both a checksum and the logical address of the data. To
accommodate this additional information, the disks have been reformatted to
520-byte sectors.
During a read operation, as the data is returned to the host, the check
information is verified for correctness. An error in the check information will
cause the controller to recover the data using the RAID redundancy
information. If the recovery is unsuccessful, the transaction is marked
unrecoverable, and the array continues to process other host request.
Product Overview 51