Information Technology Reference
In-Depth Information
multilocalization. This refers to the capability of
generating
a system for
supporting a particular locale at runtime.
High Availability versus Fault Tolerance
The availability of a system is defined as its uptime, as a percentage of
the total time being considered for measurement. High availability, as the
name implies, indicates that the system is available “most” of the time.
Actual percentages can vary but the figure is usually in the 90s. Sometimes,
we talk of “five nines availability” when referring to a system that is up
99.999%. In a year, this would mean a total downtime (both planned and
unplanned) of a little over five minutes.
Interestingly enough, availability is only measured in terms of an
application or computer usage. It does not refer to the business impact
that downtime or an outage might have. A computer outage of a few
seconds might cause an irrecoverable business outage in terms of rework
that may have to be undertaken (e.g., in online stock trading). The system
may still qualify for the five-nine availability criteria while being down
for five seconds at a time, sixty times a year. High availability should be
defined from the viewpoint of the customer, not that of the system.
If a system is experiencing repeated failures, although the durations
are small, the user is justified in losing trust in the system. In the case of
online users who have been assured that they can reach their data at any
time, such lack of access can undermine the business proposition.
High availability must be designed into the system through failover,
redundancy, or a number of other techniques. If that is not being done,
then the expectation of high availability should not be set.
It is also likely that the term “high availability” may be applicable only
to a subset of the active systems. For example, there may be a high
availability failover solution installed, which fails over machines but does
not guarantee that the application comes up on the new server without
interruption or manual involvement. From the vendor and the OS (oper-
ating system) point of view, this might be a high-availability solution.
From the customer point of view, it is not.
A fault-tolerant system is one that will continue to operate, possibly
at a reduced level, rather than terminating abnormally when some com-
ponent (or components) of the system fails. The term is most commonly
used to describe computer systems designed to lose no time due to
problems either in the hardware or the software. This is achieved in either
of the two ways: (1) by using fault-tolerant components, or (2) by
instituting redundancy by using additional or duplicate components that
can start if the primary one fails. Tandem Computers (now part of HP)
