According to the crystal-ball gazers - those inhabitants of ivory
towers for whom the rigours of business-critical system support are
no more than a trivial distraction - IT applications of the
not-too-distant future will be available at the flick of a switch
(or of a mouse).
We will no longer have to think about response times; if a system
of any description cannot be 'instant on', there will be hell to
pay. End of story.
Of course, we'll be able to cope with that, won't we? Every
month, it seems, our most powerful and flexible enterprise systems
take another step towards the nirvana of 100 per cent availability,
and all the way down the line, smaller and lower-cost processors
are rolled out with marginally (or in some cases, substantially)
more resilience. A little redundancy here, automated failover
there, sprinkle in some extra fault tolerant features - and before
you know it, you've added an extra '9' to the 99.xx availability
you were offering already. At the same time, the networks are
gaining resilience, with substantial enhancements in the more
troublesome areas of technology (such as cacheing in routers),
elimination of single points of failure, and vast improvements in
bandwidth and network management to reduce bottlenecks to a
minimum.
But is availability enough? Implicit in the whole concept of
availability is the idea that a service is either there, or it
isn't. The quality of a service, rather than its mere presence, is
even more important, and is becoming all the more so, as user
expectations of e-business applications increase. As Fred Moore
puts it, in his report Scaleability Considerations and Trade-offs
(www.xephon.com/scale.html): 'Measuring the availability of a
computer or server alone is no longer adequate, as it describes
only the presence or absence of a service. QoS describes how well
these platforms operate under normal and adverse conditions. If a
store has 20 checkout counters and only one is operative, the store
is still available to its customers but its ability to serve them
(the quality of service) is greatly diminished or degraded. Vendor
claims on availability vary, but determining a value that you can
assign to a certain computing platform is crucial to selecting a
platform. Does the QoS meet your critical application requirement
needs? With e-business, the internet and intranets empowering a
greater number of people daily, enterprises will not be able to
survive much longer without near 100per cent availability
levels.'
Of course, the real danger with the hype surrounding 'instantly
available' applications is that we are chasing a moving target,
just like the roads that magically generate more traffic every time
you widen them. Quality of service has to be a balance between what
is expected of a potential application, and what is deliverable. As
an industry, we have traditionally been far better at improving
system performance than in managing the expectations of our
customers and business managers. Yes, we need to move towards our
goal of an IT infrastructure with the instant access of a PalmPilot
calendar, and an architecture as unobtrusive and seamless as the
pipes that supply our drinking water. But at the same time, we need
to make sure that the end user knows how much of that vision is
actually available today.
Even within the relatively stable world of internal accounting
systems, it's astonishing how often service level agreements are
impractically vague, or neglected altogether. As we move into the
territory of 'extranets' with other companies and web based
applications, where many of the technical components are outside
the direct control of the IT department, the idea of agreed service
levels becomes far more complex (but even more critical).
E-business has brought with it the optimistic philosophy that,
given enough imagination, anything is possible, and there's no
doubt that the most successful implementors believe passionately in
this principle. But to realise the vision, and deliver the quality
of service that today's users expect, business strategists and IT
managers need to work more closely than ever before, and must
formalise their expectations of one another. n
Mark Lillycrop is director of research at market watcher Xephon
(www,xephon.com)