Information Technology Reference
In-Depth Information
Chapter 14
Reliable Storage
A stitch in time saves nine
{EnglishProverb
Highly reliable storage is vitally important across a wide range of applica-
tions from businesses that need to know that that their billing records are safe
to families that have photo albums they would like to last for generations.
So far, we have treated disks and flash as ideal nonvolatile storage: data
stored there will remain there forever, until it is overwritten. Physical devices
cannot achieve such perfection|they may be defective, they may wear out, or
they may be damaged|and lose some or all of their data.
Unfortunately, the limits of physical devices are not merely abstract con-
cerns. For example, some large organizations have observed annual disk failure
rates of 2% to 4%, meaning that an organization with 10,000 disks might ex-
pect to see hundreds of failures per year and that important data stored on a
single disk by a naive storage system might have more than a 30% chance of
disappearing within a decade.
The central question of this chapter is: How can we make a storage system
more reliable than the physical devices out of which they are built?
A system is reliable if it performs its intended function. Reliability is related
Definition: reliable
to, but different than, availability.
A system is available if it currently can
Definition: available
respond to a request.
In the case of a storage system, the storage system is reliable as long as it
continues to store a given piece of data and as long as its components are capable
of reading or overwriting that data. We dene a storage system's reliability as
Definition: reliability
397
