Point-in-time (PIT) solutions are what most companies will
implement due to cost. They will define their Recovery
Point Objective (RPO), which basically states how much data
can be lost. The RPO also dictates the type of offsite /
remote storage that will be required to meet that objective.
For example, if a company states that it is only acceptable
to lose "up to 1 hour worth of processing" then offsite
tapes are not enough. If they want "within 5 minutes of the
failure" then some type of replication or SAN will be
required. The more current the data the more costly the
solution.
What really complicates things is when there are numerous
heterogeneous systems (e.g., NT/2000 & Unix servers)
running different (or different instances) of software
(databases, ERP systems, even mail systems). It can be very
difficult to synchronize these systems to an exact point in
time (without realtime recovery).
At the least you need synchronized system clocks. Then
you need a way to frequently generate & flush journal /
redo information. Generally you will also need to recover
the "work in process" areas (e.g., transaction log files).
Certainly not impossible, but there is much more to
consider when working with the types of mission-critical
systems that most businesses use.
Finally, remember your Recovery Time Objective (RTO). It
may not make a lot of sense to plan on recovery up to the
point of failure if it will take 2-3days to restore the
business systems.
