7
JUNE, 2005
The
utility business model of the future
While
there were several important and interesting products being launched at
HP's ENSA@Work event, including an Itanium-based implementation of the
much-loved Tandem Non-Stop server system, there were other demonstrations
that played the equally important role of setting out a potential path
for architects to follow in the future. Not least amongst these was the
outline by John Manley and Martin Sadler, directors of HP's Bristol Research
Laboratories, of a potential business model for future utility computing
infrastructures.
This pair are interesting in that they are amongst the few people in the
IT industry who not only understand a great deal about the technology
and issues but are also allowed - indeed encouraged - to discuss them.
This makes them particularly interesting for solutions architects, as
they are allowed to publicly day-dream about future trends. OK, day-dream
is a gross over-simplification, for what they showed at the event set
out a business model based on a real pilot of a utility service in action.
This pilot is known as SE3D (`seeding' the development, geddit?) and targets
the UK's community of animators. This is an area of film where the UK
excels (Wallace and Grommit being the most obvious example) and there
is a strong pool of UK talent, many working with computer-based graphics
systems. The biggest problem they all face is the final rendering of the
animation outlines into the finished art. This is a complex computational
problem that requires true levels of brute IT force to complete. Major
animation houses spend millions on rendering server farms, and millions
more updating them regularly.
Small animators cannot afford this, so the plan was hatched for an animation
competition where a selected group of animators could exploit the huge
IT resources in the HP Labs' network of utility datacentres. The underlying
objective, however, was to test out the management of such a service where
each animator had an account with the datacentre and could rent time on
it, as well as then evaluating the consequent business model.
Each animator was given a notional block of `Monopoly' money as their
budget with which they could bid for time slots on the servers, as and
when they required it. This led to the development of a market in utility
time. Animators that know well in advance that they will require a time
slot can bid early, and get a price based on demand - probably low. They
can then use the time, sell it to other animators if they subsequently
don't need it; or if they need to, buy more late in the day at a much
higher, demand-driven price.
According to Manley and Sadler HP is learning a great deal about managing
such time-sale markets, as well as the management of the resources. For
example, the exercise is managed from Bristol, but the servers being used
are at the company's Palo Alto Labs. This is deliberate, as one of the
key tasks of any future utility service will be the management of third
party resources that are integrated into a single, deliverable service.
Many pundits agree that this is the way IT resources will be delivered
to companies in the future, with the only argument being as to when it
might occur. It certainly will not be tomorrow, and is more likely to
not even show its face for another five years. But in 10 years it could
be the way many companies operate, and is therefore a business and operational
model that architects should now be examining to determine when and where
it will fit their own company's plans. Manley and Sadler can be tracked
down via the company website. www.hp.com
|
