 |
|
DAILY NEWS AND INFORMATION FOR THE GLOBAL GRID COMMUNITY / MAY 26, 2003: VOL. 2 NO. 21
|
Special Features:
THE COST OF COMPLEXITY AND THE USE OF GRIDS
By Mark Tolliver, chief strategy officer for Sun Microsystems
The tech industry today faces a challenge with a wry twist: Our own impatience
-- our desire to change the world now -- has been surpassed by the
impatience of CFOs looking for instant ROI.
Some might say that we've become the victims of our own grandiose promises and
too-futuristic thinking. Remember when we thought stores were an endangered
species? We thought inventories, too, would soon be extinct, thanks to the
super-efficient, just-in-time supply chain.
So what happened? We still have stores -- but they all have Web sites and sell
their wares online as well as in person. And we still have inventories -- but
they're getting smaller as we learn to network with our suppliers.
So, in the disparity between hype and reality, we see ... progress. In fact, I
believe the tech industry, rather than falling victim to futuristic thinking,
will be revitalized as a direct result of our tendency to look ahead.
In a way, our own hype enabled us to see what would be needed. The rapid
growth of the Internet, combined with our vision of just how far it could go,
caused us to rethink the way we design and manage networks -- and the changes,
now beginning to come to market, will make CFOs smile again.
Why? Because we're attacking the biggest drawback of technology: It's too
complex and costly.
Processing power, storage capacity, and bandwidth may be cheaper than ever,
but the bulk of any company's technology budget -- about 70 or 80 percent --
goes somewhere else: maintenance.
Changing that equation is our number one goal.
The number of systems a single administrator can manage today is somewhere
between 15 and 30; it should be 500. System utilization rates are currently
around 15 percent; they should be 80 percent. The time it takes to deploy a
new network service is generally measured in weeks; it should be days.
More hype? Not at all. If history is any indication, it's simply a list of
attainable goals.
In the early days of computing, system administrators had to manually schedule
programs to run on specific processors. Today, software automates such tasks.
In 1960, programmers wrote applications using languages that reflected the
fundamental instructions used by the computer hardware itself. Today, using
fourth-generation languages and technologies, programmers can write a single
statement to execute millions of instructions.
My point is this: For more than 40 years, the industry has created tools for
dealing with increasingly complex tasks. We're doing the same thing today.
Take computing "grids," for example. Like the Internet, grids were once the
realm of universities and laboratories, but now they're showing up in
commercial settings -- Sun alone powers more than 7,000 of them.
Linking departmental and even enterprise-wide resources, then allocating them
based on business goals and priorities, grids put computing power where it's
needed most.
And that's just the beginning. The industry is poised to move to even higher
levels of abstraction, virtualization, and automation. That's where the big
breakthroughs will come, not only in increased utilization but, more
importantly, simplified management.
This long-term vision is already taking shape. (Our own N1 roadmap, for
example, includes the phased roll out of software engines for virtualization,
provisioning, and policy automation.) It will pick up steam as we continue to
bring down management costs, freeing up funds for new projects -- with
quicker returns.
|