The level of understanding of grid computing as an idea
has matured to the level where users are now able to understand the
different types of grid computing that are available to
them.
Quocirca's latest grid research shows that while 18% of users
see that their first foray into grid computing will be via a highly
strategic enterprise grid project, a similar number see a
departmental, or cluster grid as the way forward.
There are two main reasons for people looking towards a
departmental grid. The first is technical: it makes a great deal of
sense to try out your first grid experience in a highly constrained
environment - a single application being able to call on
virtualised resources as required to fulfil its needs.
The second is political: one of the main hurdles to grid
adoption is trying to identify ownership and financial resources to
fund the grid project. By aiming the grid project at a single
department, both of these political issues can be sidestepped.
The prime uses for departmental grids will be resource-intensive
applications that have peaks and troughs in their usage.
Manufacturing could be one area to benefit, where computer aided
design systems need large processing resources for number
crunching. Other applications could be in pharmaceuticals, where
molecular modelling needs intensive resources, and in the finance
sector for mathematical modelling.
Each of these areas is the low-hanging fruit of the grid world -
they are high-performance computing systems. But is there a role in
the more commercial environments for departmental grids?
The answer is yes, and some suppliers are beginning to make
moves in this area. A prime example is IBM, which has taken its
blade server technology and created a low-end grid system called
Grid and Grow.
This box comes with Websphere Application Server GE (Grid
Enabled) already installed. This makes any application
grid-capable to an extent, and the blades dynamically allocate
resources to the application or applications installed on the
environment as required.
The neat thing is that should you want to grow your grid, this
is what Grid and Grow has been made for. If you want more
scalability, just plug in another box. If you want to run more
applications, put in a different box, and each will be aware of the
other and will share resources between the two systems
automatically.
This product has a low entry price and is aimed at the
mid-market. For departments wanting to look at how grid can help
them, it would seem to be an excellent way of testing the waters.
Not only does it provide a simple means of grid-enabling existing
applications, but it also provides flexibility going forward, with
each new box being capable of being part of a greater system.
We can expect similar systems to come from other suppliers now
that grid is seen as being the next evolution of the infrastructure
following on from web services and service oriented
architectures.
It is a frightening thought that most companies are running at a
CPU utilisation of less than 10%, and a storage utilisation of less
than 30%. This affects the number of servers being bought. With the
need for effective and efficient infrastructures being pushed by
business management, grid is an answer and departmental grids would
seem to be the right entry point.
Clive Longbottom is head of research at analyst firm
Quocirca
www.quocirca.com/report_grid
index3.htm