HP Surestore Disk Array 12h - And FC60, Руководство для продвинутых пользователей, Страница: 66 / 466

66
Capacity Management Features
•
Hot spare group – All disks assigned the role of global hot spare become members of
this group. Up to six disks (one for each channel) can be assigned as global hot spares.
•
Unassigned group – Any disk that is neither part of a LUN nor a global hot spare is
considered unassigned and becomes a member of this group. Unassigned disks can be
used to create a LUN or can be used as global hot spares. Unassigned disks do not
contribute to the capacity of the disk array.
Disk Array Caching
Disk caching is the technique of storing data temporarily in RAM while performing I/Os to
the disk array. Using RAM as a temporary storage medium can significantly improve the
response time for many types of I/O operations. From the host’s perspective the data
transfer is complete, even if the disk media was not involved in the transaction. Both write
caching and read caching are always enabled.
Caching enhances disk array I/O performance in two ways:
Read I/O If a read I/O requests data that is already in read cache, the disk array
services the request from cache RAM, thus avoiding the much slower
process of accessing a disk for the data. A pre-fetch capability
enables the disk array to anticipate needed data (for example, on a
file transfer) and read it from disk into the read cache, which helps
significantly with sequential read I/Os.
Write I/O During a write I/O, the disk array writes the requested data into write
cache. Rather than writing the modified data back to the disk
immediately, the disk array keeps it in cache and informs the host
that the write is complete. If another I/O affects the same data, the
disk array can update it directly in cache, avoiding another disk
write. Data is flushed to disk at regular intervals (10 seconds) ro
when the cache flush threshold is reached.
Write cache is mirrored between the two disk array controllers. Each controller maintains
an exact image of the write cache on the other controller. If a controller fails, its write
cache content is flushed to the disk by the remaining controller. Because write cache is
mirrored, the operational controller automatically disables write caching until the failed
controller is replaced. After it is replaced, the operational controller automatically re-
enables write caching. Mirroring effectively reduces the size of available cache by half. A