HP ProLiant DL380 G5 DPSS, Руководство администратора

Страница: 47 / 168

Arrays
See Figure 18 . With an array controller installed in the system, the capacity of several physical drives
(P1–P3) can be logically combined into one or more logical units (L1) called arrays. When this is
done, the read/write heads of all the constituent physical drives are active simultaneously, dramatically
reducing the overall time required for data transfer.
NOTE:
Depending on the storage server model, array con fi guration may not be possible or necessary.
P1 P3 P2
L1
gl0042
Figure 18 Con fi guring arrays from physical drives
Because the read/write heads are simultaneously active, the same amount of data is written to each
drive during any given time interval. Each unit of data is termed a block. The blocks form a set of data
stripes over all the hard drives in an array, as shown in Figure 19 .
S 1
S 2
S 3
S 4
B1
B4
B7
B2
B5
B8
B11 B10 B12
B6
B3
B9
gl0043
Figure 19 RAID 0 (data striping) (S1-S4) of data blocks (B1-B12)
For data in the array to be readable, the data block sequence within each stripe must be the same.
This sequencing process is performed by the array controller, which sends the data blocks to the drive
write heads in the correct order.
A natural consequence of the striping process is that each hard drive in a given array contains the
same number of data blocks.
NOTE:
If one hard drive has a larger capacity than other hard drives in the same array, the extra capacity is
wasted because it cannot be used by the array.
Fault tolerance
Drive failure, although rare, is potentially catastrophic. For example, using simple striping as shown in
Figure 19 , failure of any hard drive leads to failure of all logical drives in the same array, and hence to
data loss.
To protect against data loss from hard drive failure, storage servers should be con
fi gured with fault
tolerance. HP recommends adhering to RAID 5 con fi gurations.
HP ProLiant DL380 G5 Storage Server 47