The theoretical MTBF of any hardware configuration can be
calculated if you have the MTBFs of each component that make up
your configuration. For example, you can calculate the MTBF of
your server if you have the MTBFs of main CPU board, disk drives,
server packaging, etc - a rather daunting task. Most discussions
of server MTBF focus on disk drive MTBFs for a several reasons.
First of all, components with moving parts (such as disk drive
actuators and motors) typically have significantly lower MTBFs
than non-moving components (such as memory chips or main CPU
boards). Because a server's theoretical MTBF is most influenced by
the MTBF of the least reliable component as well as the sheer
number of components, disk drive MTBFs typically dominate the
overall server configuration theoretical MTBF. Theoretical MTBF
decreases in proportion to the number of components that make up
the server, so larger configurations containing many disk drives
by definition have a lower MTBF. Add to that the fact that disk
drives contain data that may be time-consuming or impossible to
recreate, it is easy to see why disk drive reliability dominates
server reliability discussions.
The following examples illustrate the impact of disk drive MTBF
on overall server MTBF.
A server's theoretical MTBF is calculated from the theoretical
MTBFs of the components that make up a the server:
- where
- N = MTBF of each component
- x = the number of components in the configuration
If one component, such as a disk drive, has a significantly
lower MTBF than the rest of the population, its MTBF dominates the
overall server MTBF. For example, examine a server containing one
CPU main board with an MTBF of 1,000,000 hours, and a single drive
with an MTBF of 300,000 hours (we'll ignore the other components
for simplicity's sake). The server's MTBF is calculated as
follows:
= 230,769 hours, close to the disk drive's MTBF
Even if all your components have high MTBFs, the overall
configuration's overall MTBF is reduced in direct proportion to
the number of components in the configuration. For example, the
MTBF of a storage subsystem consisting of two disk drives with
identical 300,000 hour MTBFs is:
= 150,000 hours, exactly half the MTBF of each disk drive
Similarly, a 10-drive configuration MTBF is one-tenth the MTBF
of a single drive, or 30,000 hours, and a 100-drive configuration
is reduced to 3,000 hours.
