Page 87 -

P
58 58 P a r t I I : a r t I I : C C o n s u m p t i o n I s s u e so n s u m p t i o n I s s u e s

Drives powered down, waiting to be accessed Actively spinning drives

FIGURE 3-6 In a MAID deployment, only a quarter of the disks are spinning at any one time.

the power used to run them as well as reducing the generation of heat, which in turn reduces
cooling costs.
MAID differs from RAID in that it has increased storage density and is much less expensive,
thus saving power and the need for cooling.
MAID comes with some compromises, however, such as increased latency, lower
throughput, and much lower redundancy. Although a MAID solution can be slow—data
access can take a few milliseconds up to 10 seconds—it is much faster than tape, which can
take 60 seconds or longer to access data.
Also, because large hard drives are designed for constant spinning, continual shutdown
and reactivation threatens their reliability. Drives that are designed for repeated spin-up/
spin-down cycles are much more expensive.
The MAID architecture really developed because of the introduction of SATA drives that
are designed to be powered up and down. In a large deployment, MAID allows a dense
packaging of drives, and typically only 25 percent of the disks are spinning at any given
time, as illustrated in Figure 3-6. This helps with the problem of throughput.
The appeal of MAID is also apparent when you consider their use in large environments.
SATA drives are rated at 400,000 hours of life. If a datacenter has 1000 drives powered on all
the time, a drive would fail every 18 days. Clearly this is no good, so employing a MAID
system quadruples the drive’s life to 1.6 million hours.
Power-managed RAID
The idea behind RAID has always been to safeguard your data. But employing multiple,
always-on drives jacks up your power consumption. To deal with this issue, a new form of
RAID has been introduced.