by Steve Towns*
"Nothing endures like change." Heraclitus, a
Greek philosopher, said it in the 6th century, but it could
easily be any business manager speaking today. What's different
now is that change in business involves more than people
-- it involves IT systems. And those don't change so easily.
Introducing a new product? Making an acquisition? Merging
two divisions? Ready for a costly IT headache?
"Organizations today are constantly in the middle
of change,” says Sharad Singhal, a researcher in
HP Labs. “So the question for us was: How can we
make IT more responsive to those changes?”
Researchers may have an answer in an experimental tool
that aims to turn IT systems into flexible resource utilities
by delivering on-demand computational resources to applications
or services.
The research is already making its way into products.
The HP OpenView Automation Manager, announced in November,
is the first automation product to use critical business
performance information to allocate IT resources.
The software, based in part on business intelligence technology
developed in HP Labs, adjusts the configuration of IT services
and applications based on changes in demand, automatically
reacting to peaks in order processing, for example, or
adapting to a surge in e-mail traffic.
Automation Manager and future products emerging from the
research could reduce the risk of complex IT projects and
shave weeks from the deployment process.
Researchers in the lab designed the software to capture
user requirements, assess the available IT resources, account
for the existing workloads and other constraints, and create
system configurations that meet those requirements.
Easier said than done. The tools must not only understand
the capacity and behavior of sophisticated hardware such
as servers and storage, but they must also determine capabilities
of multiple operating systems and sort out conflicts among
them -- all while complying with hundreds of constraints
required to protect the operation of key systems.
“Typically you would need to gather a team of experts
-- on networking, servers, storage and applications --
to figure out what needs to be done,’” says
Singhal. “Using that approach, designing and deploying
something like a new Web application could take about 12
weeks."
“These new tools look at all of that, reconcile
the information and quickly bring the system together,” he
adds. “We would like to bring that time -- including
all the human processes that go with testing -- down to
a week or less.”
To accomplish that, the researchers developed a collection
of integrated technologies that perform three key functions:
• Policy-based resource composition - allows users
to model system composition rules and best practices,
and gives them the ability to automatically create custom
configurations that conform to those policies. This shortens
the time spent designing IT environments and reduces
the likelihood of error.
• Capacity management - provides scheduling and
capacity management algorithms that track complex resource
demands and react accordingly. This lets operators manage
infrastructure use and meet quality-of-service requirements.
• Resource assignment - uses mathematical programming
techniques to ensure that resource requirements such
as network bandwidth are met for new applications, and
that adding those applications to shared resources does
not cause system bottlenecks.
 |
|
These components mask the intricacy of large IT environments
by giving IT managers a relatively simple drag-and-drop
interface for designing new functions. This doesn’t
completely eliminate the need for skilled people, however.
The tool assembles the pieces; IT professionals check the
results.
If the results aren’t quite right, operators tweak
the requirements to get a different answer.
By reducing the cost and difficulty of IT changes, these
tools could make it easier for businesses and other large
organizations to explore and adopt better business processes.
“The amount of time managers spend agonizing over
business-level change will shrink because they know if
they make the wrong decision, they can go back and change
it tomorrow,” Singhal notes. “They haven’t
set into motion a bunch of wheels that can’t be stopped.”
The new technology grew from Utility Data Center research
at HP Labs. The Utility Data Center concept treats IT resources
in an organization’s data center as one pool of computing
power. Computing resources are configured and allocated
to handle a particular task, then reconfigured on the fly
to tackle the next job.
Instead of operating dedicated payroll systems or accounting
systems, for example, a business could apply computing
power to those tasks when needed, and direct those resources
elsewhere when finished. The experimental software facilitates
a Utility Data Center approach by automating the entire
life cycle of computing tasks -- the design, deployment,
operation and decommissioning of each computing job.
But it is just as effective with traditional, dedicated
systems because it collects data about the underlying infrastructure,
and configures systems based on an organization’s
IT capabilities, taking into account whether resources
can be shared or not.
Ultimately HP Labs research may allow enterprises to use
their IT resources with unprecedented effectiveness.
For instance, businesses could rapidly allocate resources
to create custom infrastructure for marketing promotions,
or speed up test and development environments where infrastructure
has to be constantly reconfigured. What's more, gaining
an understanding of how different workloads are placed
on applications could make it possible to reassign applications
to run on infrastructure that would otherwise be idle.
The tool also could facilitate IT consolidation efforts
and help businesses make more effective use of their computing
infrastructures because it would allow them to determine
what proportion of resources each application requires
whether it's possible to juggle those resources to operate
more efficiently.
Internally, HP is using the software's capacity management
components to consolidate multiple servers into a shared
utility to support business applications. Like many large
companies, HP maintains hundreds of internal business applications,
each of which traditionally required its own server. Treating
these separate servers as a single, shared computing resource
reduces the number of servers needed, which lowers licensing,
support, management and hardware costs.
HP researchers also used the technology to help design
an IT consolidation initiative for a large financial services
firm. The firm plans to move users from standard PCs to
terminal-based “virtual” desktops supported by centralized
servers and storage. Using components of the software,
researchers modeled how the proposed system would react
to various demands and how resources could be allocated
most effectively.
“We have to simplify the process of making changes so
that systems quickly accommodate new business requirements,” Singhal
said. “We’re trying to make sure that IT resources support
the kinds of things an enterprise needs to do.”
* Steve Towns, editor of HP Government Solutions magazine,
wrote an earlier version of this story for the magazine's Winter
2005 issue . The original story was modified for HP
Labs.
|
Printable version
