Disaster recovery is still a difficult item to sell to the
board. The main problem is that all the possible disasters appear
to be extremely improbable.
It is not a good idea, for instance, to make the board think you
spend your expensive time worrying about deeply unlikely things,
such as the Thames flooding or the roof collapsing.
Done properly, disaster recovery is not cheap, and there is the
likelihood you will never have to use it. This is just as well
because, in my experience, the disaster recovery arrangements at
many firms are alarmingly flimsy.
Compare disaster recovery with any other IT deliverable. When you
deliver some minor upgrade to a vaguely useful system, you test it
and try it on users. I know of only a few companies that test
disaster recovery this thoroughly. Yet having a back-up which has
not been tested in a realistic way is a waste of money.
By testing, I do not mean snatching users during a working day and
saying "go to the disaster recovery centre, work there, see how far
you get". Of course, most firms "test" their systems, but the tests
tend to be small-scale and involve few staff, which hardly makes
them realistic.
We all know that systems rarely work first time, even when we have
tested them, and the first days of live running are often more
interesting than we would like. If in a catastrophic event your
business lost half of its staff, would it still be able to
function?
Ever since the 11 September terror attacks in the US, discreet
conversations with the IT heads of other firms have shown little
confidence that things have got much better. One top-tier bank
equipped its disaster recovery centre with the old hardware from a
major upgrade programme and it had not actually got around to live
testing.
So how should IT directors sell disaster recovery to the board?
Just listing problems is not going to fix it. One option is to sell
disaster recovery as other things.
Having hardware lying around is wasteful, so the disaster recovery
site can double as a development and test centre. This works better
if you have multiple sites which can back each other up. You can
use thin-client technologies such as VMware and Citrix to deliver a
working desktop to users, wherever they are.
Load-balance the datacentre across your sites: the rule of thumb is
that if you have enough servers to call it a datacentre, not a
comms room, you need to split sites.
To sell disaster recovery, you need to dovetail it with your
business support, not treat it as a special item, otherwise it will
not work and will cost too much.
Dominic Connor is head of IT at King &
Shaxson