| Hitting the proverbial wall
Despite enormous advancements in computing power over
the past few decades, today's personal computers and servers
can still hit the proverbial wall when a demanding application
or set of programs wind up running at the same time. In
fact, while some systems are sitting idle others are grinding
through every last bit of processing power in an attempt
to keep up with the workload. This imbalance of resources
costs many organizations enormous amounts of money and
time. It also impairs an organization's ability to solve
complex tasks.
Although most companies cope with such imbalances by continually
upgrading systems and adding additional processing power,
a growing number are now reexamining conventional thinking.
They are turning to grid computing to create a more efficient
IT infrastructure.
Grid computing uses specialized software to transparently
and securely share IT resources across geographically dispersed
sites. When particular systems or tasks require greater
computing power, they're able to tap into systems that
are underutilized or unutilized at that particular time. "Pooling
and sharing computing resources has a profound impact on
a business," says Tom Hawk, general manager of grid
computing at IBM.
Although grid computing isn't a new concept - computing
experts envisioned it more than 30 years ago and it has
been used by research and academic institutions for over
a decade - it is only beginning to take root in the mainstream
corporate arena. Today, firms in life sciences, engineering,
financial services and manufacturing are turning to it
to solve complex problems and utilize IT resources to maximum
advantage. "Grid computing can create an IT environment
that is resilient and flexible. It can smooth out the technology
bumps that most organization's face," Hawk explains.
To be sure, it's garnering the attention of CIOs and other
decision makers. According to South Hadley, Mass.-based
market research and consulting firm Grid Technology Partners,
the use of computing grids will grow from $180 million
in 2002 to $4.1 billion by 2005. In addition, less than
one percent of companies had tapped into grid computing
in 2002. The consulting firm estimates that the figure
will swell to 10 percent by 2005. More than 40 per cent
of large companies (greater than 10,000 employees) will
use the technique by 2005.
Relieving gridlock
The appeal of grid computing is undeniable. IBM research
indicates that a typical mainframe computer is idle 40
per cent of the time. A Unix server actually "serves" data
less than 10 per cent of the time. And most PCs do absolutely
nothing for 95 percent of a typical day. Yet, at the
same time, key applications cannot grab enough processing
power from a single computer to run optimally. In the
end, it's like a factory that runs at only 20 percent
of capacity because conveyor belts run too slowly or
workers can't load trucks fast enough to keep up with
production.
Virtualization, the ability to create additional resources,
changes the entire equation. For example, using grid computing,
pharmaceutical giant Novartis performed 3.18 years of simulations
in a mere seven days. Meanwhile, telecommunications reseller
Cognigen doubled its computing power at a mere fraction
of the cost.
Using systems connected over an Internet Protocol (IP)
network, it is possible to connect a heterogeneous group
of PCs and servers - including those running on Unix, Linux
and Windows - and boost processing power by orders of magnitude.
What's more, it is possible to do using open standards
and protocols, such as the Open Grid Services Architecture
(OGSA).
IBM, which entered the grid marketplace in 2002, has found
that grid computing is most
valuable for organizations focusing on certain key areas.
These include:
• Large-scale analysis and in-depth analytics
A diverse array of industries, including financial services,
life sciences, agricultural chemicals and petroleum,
can use grid computing to sift through mountains of
data
and solve real
business problems.
• Design collaboration
Aerospace, automotive and electronics
can virtualize disparate systems, share complex design
tools, and distribute workflow across partner communities.
• Engineering Design
Aerospace, automotive and others can
speed the design process and decrease time to market
through the use of grid tools.
• Research Collaboration
Research facilities, think-tanks
and universities can share raw data among researchers
and tackle complex mathematical equations. They're also able to share data securely.
• IT Optimization
Organizations in an array of industries
can increase computing capacity without constantly adding
to their IT costs.
Based on these five areas, IBM has introduced a suite
of industry-specific grid computing offerings designed
to help organizations improve operating efficiency, reduce
capital expenses, enhance productivity and adapt more quickly
to changing business conditions.
Hawk believes that grid computing will also pop up in
e-business and e-commerce applications, customer relationship
management programs, business intelligence tools and other
areas over the next few years. "Grid computing faces
many of the same obstacles that the Web did. It is necessary
to get used to the idea of open standards and interoperability
across heterogeneous systems," he explains.
Net gains
Many of the challenges facing grid computing have more to
do with cultural issues than technology. Grid computing can
force an organization to examine, among other things, who
manages and controls various IT resources. What's more, it
can alter workflow and data management across the organization.
In some cases, a particular data set or program may reside
on an individual computer for only a few seconds. That also
can force a reexamination of workflow and accounting or billing
systems.
Although security is an ever-present concern within any
company, advances in authentication and certification make
grid computing a safe proposition. Moreover, the grid software
ensures that others cannot read files or data on other systems.
Still, it is clear that some workers still aren't completely
comfortable with the idea of others using their computer's
processing power, even if it's in a stealth manner. Likewise,
privacy can emerge as a sensitive issue, particularly in
the life sciences and medical arena. "It's something
that all organizations need to manage," Hawk says.
Nevertheless, the future of grid computing looks bright.
Some believe that, within a few years, organizations might
band together to create a "GridNet" similar to
the Internet. It would allow different companies and organizations
to pool and swap resources beyond their own walls. Consulting
firm Grid Technology Partners believes that such a TeraGrid
could one day operate alongside the Internet as an entirely
separate entity.
Concludes Hawk: "Grid computing is a natural evolutionary
step in the computing paradigm. Within a few years, it will
become the fabric of an efficient computing model, just as
a Web browser or e-mail program is now part of a desktop
computing experience."
Samuel Greengard is a Burbank, CA-based writer who specializes
in business and technology.
|
