Another issue related to reliability and fault tolerance but not the same as either of them
is availability. This term refers simply to the ability of users to continue to
get access to their data. If people are using the array to do their work, then it is
"available"; if everyone's gathered around the water cooler talking about how
much money some guy won last night on "Who Wants To Be A Millionaire?" then it
is "not available". 
For many businesses that implement RAID, availability is the "bottom line"
reason why they decided to spend the money for that technology. (Data protection is
usually just as important, but you can get pretty good data protection through backups,
saving the cost of all the RAID hardware, and you
still need the backups anyway.) For some businesses, the cost of losing access to
their data for an hour can exceed the price of the entire RAID system.
The availability of an array depends on several factors and features:
- Hardware Reliability: The higher the reliability
of the hardware in the array, the less the chances of there being a hardware problem in
the first place; this is the best way to keep availability high, of course. The
reliability of support hardware is just as
important as that of the drives in the array.
- Fault Tolerance: RAID arrays and implementations that have higher fault tolerance provide higher availability. Again
here, the fault tolerance of support hardware is
also important.
- Hot Swapping: RAID arrays that include the ability to hot swap drives eliminate the need to shut down the
system in the event of a hardware failure.
- Automatic Rebuilding: If the system runs continuously, the ability to
automatically rebuild the drive while the array continues
to operate is essential.
- Service: If an array goes down, availability is gone until the hardware
fault is corrected. Though this is uncommon in a properly-configured RAID system it does
occur, and when it does, availability rapidly boils down to how fast the hardware can be serviced.
If high availability is important then it can be achieved--at a cost. There are in fact
companies that design special, highly-fault-tolerant systems that can withstand just about
anything not involving the use of explosives. Very high-end solutions for
availability-critical applications go beyond RAID to the use of complete redundant
systems, often in different locations--these can even handle explosives!
Next: Reliability of Other System Components
