When buying a car, what you do depends on whether you
are looking for speed or reliability, not to mention the size of
your wallet. And you’ll also need to think about petrol
consumption, insurance, replacement part costs, resale
value…
In implementing business critical solutions (BCS), deciding
which servers, storage and infrastructure you’ll need to run your
business-critical solutions requires the same kind of thought
processes. Is it speed or reliability the business most needs, and
if people choose ‘speed’, exactly how speedy? And what are the
implications for the IT department of giving them that speed?
Sadly, according to James Governor, principal analyst and
co-founder of analyst firm Redmonk, there still aren’t enough of
these kinds of discussions going on. ‘It’s the age-old problem.
Systems people worry about systems uptime and business people worry
about business uptime,’ he points out.
And that leads to a lack of joined-up decisions. ‘If line of
business is in charge, as if often the case, it may be looking for
the best application, but not thinking about the best architecture.
What’s needed is joined-up purchasing and management,’ suggests
Governor.
Only by battering down the barriers between business and IT (and
between different IT projects) will the best business solution be
worked out. Technology should be bottom of any to-do list for
running a business-critical solution. After all, you could deploy a
state-of-the-art application, server, storage and networking and
not only pay a hefty wad, but still miss out on the business value
jackpot.
‘It’s a classic case of leaping to the answer rather than
looking at the question,’ says Peter Critchley, strategy director
at technology integrator Morse.
The question is: what exactly does the business want to
achieve?
IT people are still too blinded by vendor claims of scalability and
high availability to think of the true business implications,
maintains Governor. They are not assessing what the business needs,
the impact on infrastructure or the development costs.
‘Organisations are not very good at working out the implications
of scalability and availability and mission-critically of systems
they run,’ he says. ‘They also have a tendency to buy something
that’s “five-nines” without thinking about which particular service
that’s for. To get the most out of IT you need to analyse what you
mean by scalability. Four or five nines sounds good, but if you
measure that over a year, that can mean significant downtime. For
some services that can be fine, but for others it’s not.’
The business isn’t interested or impressed by phrases such as
‘five nines’. It doesn’t matter if financial applications go down
over a whole weekend, but a few minutes on the last day of the
financial year could be extremely damaging.
But IT’s response to demands for uptime is often to buy big.
‘Hardware salesmen would prefer it if everyone over-prioritised
their systems. If an application needs scaling, their answer is to
buy more hardware,’ adds Governor. ‘The thinking is that as
hardware is coming down in price, the answer must simply be to
throw more hardware at the problem.’
We’re now mopping up the mess from what Peter Hindle, enterprise
solutions manager at HP UK & I, calls this ‘just in case’
scenario. Rather than buy the six processors needed for a big
capital project, companies would buy 10 - just in case - then add
more for failover and yet more to keep development work separate,
just in case. End result: 30-odd CPUs instead of the six the system
actually required.
That means companies are sitting on vast pools of untapped
capacity. ‘I usually quote that we use about 20%, but most people
it’s nearer to 8-10% and I’ve had people publicly say 6% and
privately lower,’ says Hindle.
Paradoxically, one of the key gripes of IT heads is a lack of
resources where they need it. And of course this will closely be
followed by complaints about IT costing too much. In fact,
three-quarters of IT budget goes on ‘just keeping the lights on’
rather than innovation, according to Hindle.
Rather than buy more tin, successful companies are now looking
at how to get rid of systems and to consolidate and centralise both
their hardware and software. And that is a proving a far from easy
task.
‘Most large enterprises have a 100 or 1,000-plus
business-critical systems – too much of a large landscape to
grapple with in one fell swoop,’ says Michael Allen, global
performance director at services company Compuware. ‘You need to
look at evaluating that, not only measuring end-user availability
and performance, but strengthening stakeholder and business
involvement.’
Consolidation makes total sense, but it is complex to achieve.
When HP started its own internal consolidation programme – reducing
roughly 10,000 applications (roughly, because the company decided
to stop counting at 10,000) down to a more manageable 1,500 – it
hit some problems. Every business process had a variable in some
region or country, which kept adding complexity where it was trying
to strip it away. The answer was to move up a level and streamline
the business processes themselves, not the applications. As Hindle
points out: ‘Sometimes the complexity is not always in the
application or the hardware.’
Allen also believes centralisation of business critical systems
is not always as straightforward as some believe. ‘Many legacy
applications are set up to run on departmental servers and when
they run on the WAN they are just too heavy. So when organisations
try to centralise they are often not looking at the network
aspect,’ he says.
Whether centralising or not, a key weapon for reducing costs and
ensuring that business-critical systems are well supported is
server and storage virtualisation, Server virtualisation partitions
a large server into smaller virtual servers, which means one device
can take the place of many smaller servers.
A November IDC report states that three-quarters of 500+
employee organisations are using virtual servers and more than 50%
are using it to for business-critical systems. By 2009,
virtualisation will be a $9bn market.
Just as important as the virtualisation is service management –
lifting eyes above the infrastructure layer and looking at IT as a
service, popularised by the IT Infrastructure Library (ItIl)
best-practice framework.
Nicholas Carr’s hotly debated article in the Harvard Business
Review a couple of years back contended that IT is becoming a
commodity. While it’s true many technology elements are becoming
commoditised, it’s what you do with them that counts.
Consolidation of IT hardware, software and processes may reduce
costs, but it will only start delivering business value when it’s
totally aligned with your business goals.