by Simon Firth
For years, Grid computing has been the
province of academic and government researchers,
who have been hooking together individual
computers around the world to perform compute-intensive
These virtual supercomputers, which can
quickly process vast amounts of information,
have helped produce breakthroughs in meteorology,
physics, medicine and other fields. Scientists
have even harnessed the power of the Grid
in the SETI@home search for extraterrestrial
Now researchers at HP are working to bring
the benefits of Grid computing to the corporate
The promise of grid in
Grid computing will let enterprises use
their IT resources more efficiently. It
will allow collaboration within a company on any kind of
compute task of any size or duration. It will make
it possible to set up ad hoc, secure IT
projects between geographically disparate
units, and even across organizational boundaries.
"Companies want to better utilize
the technology they have," says Greg
Astfalk, HP's Chief Scientist and a leading
proponent of grid computing within HP. "They want to interoperate and collaborate
better and more efficiently and securely
with partners. Grid addresses those needs."
But even though the promise of Grid computing
makes it one of the most talked-about ideas
in corporate IT today, the barriers to creating
an enterprise Grid are considerable.
An enterprise grid must be
smart enough to navigate boundaries
such as corporate firewalls and networks,
and at the same time, must be reliable enough
to process mission-critical applications
such as company payrolls and credit card
"Scientific applications can be scheduled
in 'batch' mode or are deadline-driven,"
says Sharad Singhal, a principal scientist
in HP Labs. "By contrast, commercial
applications run continuously over long
periods of time and have more stringent
requirements on reliability, security and
Grid resource allocation
Hard technical problems remain
before enterprises can make full use of
Grid capabilities. One problem being tackled by HP Labs is determining how to connect multiple data centers to the
To address that, researchers designed the Grid Resource
Topology Designer, a simple, graphical user
interface for designing Grids of UDCs or
data centers. Topology Designer allows
users to easily and securely "draw" resource needs, then submit the requirements
to the Grid for fulfillment. Working with
the data center, it automatically decides
on the appropriate resources to deploy to
fulfill the service-level request.
"Say you wanted to deploy an application
in two data centers," says Sven Graupner of HP Labs, one of the creators of Topology Designer.
"The description is submitted using
Grid protocols to the Grid, then to the
data centers, or a broker between data centers,
and it will come back and say if the job's
been accepted or not."
Although still at the prototype stage,
Topology Designer illustrates that it's
possible to build application environments
that work across data centers. "This
is exactly the vision the Grid summarizes
under its notion of a virtual organization
or collaboration," Graupner notes.
A related technology being developed by
HP Labs, the Adaptive Control Interface,
uses standard Grid infrastructure in the
UDC to create an adaptive control system
for adding or subtracting servers from an
Another hurdle for the enterprise grid
is the lack of a broadly accepted way
of deploying and configuring software and
hardware to run over it. A team in HP's
Bristol, UK, lab built a framework -- the
Smart Framework for Object Groups, or SmartFrog
- that provides rules for how resources
on a Grid are configured, run and managed
through their operative life cycle.
"Configurability and deployment are
critical aspects of the technology
today. Adaptive infrastructures, and utility
computing in particular, are complex
concepts that can be significantly simplified
by automated deployment tools such as SmartFrog,"
says Patrick Goldsack, one of the designers of SmartFrog.
Researcher Dejan Milojicic is working
with the Global Grid Forum to try to incorporate
SmartFrog concepts into an industry standard. "We've had a lot of interest from
the Grid community," he says. "Other
companies have expressed interest in making
reference implementations -- so we're very
happy about that."
This year, HP Labs put SmartFrog to work
configuring a service for rendering complex
computer graphic images. A team created
an interface to an HP Utility Data Center
in HP Labs Bristol (UK) for a local animation
company and commissioned a high-quality,
animated short film, "The Painter,"
which thanks to HP technology, was rendered
in a fraction of the usual time.
Managing resources, services
in a uniform way
Another missing piece needed to bring Grid
to the enterprise is a uniform way of managing
the wide range of computing resources in
an organization, which may include multiple
operating systems, platforms, data formats
HP's Web Services Management Framework (WSMF) could be the answer, providing a
management framework for all types of IT
to the Grid, WSMF offers a common approach
for managing all components of a Grid deployment,
including resources and services. Because the framework is itself built on Web services, it can deal with such issues as federation,
distribution, manageability and standards, according to
Homayoun Pourheidari, one of the architects of WSMF.
"Grid wants to render everything in
IT as a Grid service," explains Chief
Scientist Astfalk. "It could be a computer,
a storage array, a file, a database. It
could be a record in that database. It could
be an application program or it could even
be a scientific instrument."
Using open protocols and interfaces, he
says, makes it possible create a way "to
describe all those services, to register
them, provision them, bring them to life
and tear them down -- all the types of manipulations
you'd expect on a service."
Engineers in HP's OpenView team in the
HP Software Global Business Unit and HP
Labs researchers recently worked together
to make WSMF Grid-compatible, integrating
it with the Global Grid Forum's Open Grid Services Infrastructure standards.
"We're trying to create standard interfaces
to manage and control all sorts
of resources -- from printers to firewalls
to applications," says Akhil Sahai,
an HP Labs researcher involved in the project.
WSMF was recently submitted to the OASIS
Web Service Distributed Management Technical Committee
toward the goal of achieving a standard
management solution for the industry.
HP's commitment to Grid
Inside and outside the lab,
HP is putting its energies
toward Grid computing, working
all of the company's products.
Products ranging from the smallest
handhelds, printers and PCs
to the most powerful storage
arrays and supercomputers will
be able to connect with and
serve as resources on a Grid.
HP's Grid strategy has three primary thrusts.
Most importantly, Grid computing enables HP's
vision of the Adaptive Enterprise -- where
information technology is a highly efficient, flexible service that is agile enough
to change in line with a corporation's business
and its business environment.
"If you think of the Adaptive Enterprise
as set of bricks," says Astfalk, "and
each brick being a solution piece, the mortar
that we envision holding these pieces together
Another benefit to Grid-enabling HP's systems
is working with customers and partners already
using Grid, primarily those in scientific
and technical computing.
In the long term, Astfalk foresees the
Grid making a huge contribution in a third
area, the space where home networking meets
entertainment media. In these cases, there's typically a content
provider that allows access only to subscribers.
The provider may also require that materials
are viewed on a device with certain characteristics,
such as high-grade security.
"All this could be done by virtual
organizations in a loosely coupled service-based
economy," says Astfalk. "And that's
HP is involved in a number of Grid
collaborations and initiatives. These include:
- Global Grid Forum, platinum sponsor
- CERN openlab for DataGrid applications
HP Collaboration and Competency Network -
a forum to facilitate wide-ranging collaboration, innovation, discovery, and competency sharing between HP, HP partners and customers in high-performance technical computing (HPTC). One key focus of the network is computational and data Grids.
GridWeaver - a project with the University of Edinburgh School of Informatics and EPCC under the UK e-Science Grid Core Programme,
focusing on the configuration and management of large-scale computing fabrics, with an emphasis on fabrics that are part of utility or Grid computing infrastructures.
joint effort with
Universidade Federal de Campina Grande in Brazil to research and develop solutions of usage and management of computational grids. Builds on the MyGrid work.
GECEM (Grid-Enabled Computational Electromagnetics) with Cardiff University, Swansea University, BAE SYSTEMS and the Institute of High Performance Computing in Singapore.
GriddLeS: Building Grid Applications from Legacy Software - a project with Monash University that builds on previous work on Gridifying applications, especially legacy Fortran codes and brokering for resources.
GridLab - a European Union-funded project to
develop an easy-to-use, flexible, generic and modular Grid Application Toolkit enabling today's applications.
CERN OpenLab - research to develop data-intensive Grid technologies to be used by the worldwide community of scientists working at the next-generation Large Hadron Collider.
Pittsburgh Supercomputing Center- HP engineers and researchers, both from the High-Performance Technical Computing (HPTC) Division as well as from HP Labs, will work with PSC personnel to run tests and realistic benchmarks to help ensure successful elaboration of the TeraGrid.
Georgia Institute of Technology - National Science Foundation-funded project involving middleware for ubiquitious computing with applications for Grid
Global Grid Forum
Committee - Martin Walker
GGF Market Awareness
Committee Leadership Council - Sara Murphy, HP Marketing Manager
for Grid Computing
Allocation Protocol working group - Jim Pruyne,
researcher, HP Labs, co-chair
Configuration, Description, Deployment and
Lifecycle Management working group - Dejan
Milojicic, researcher, HP Labs, co-chair
User Program Development Tools
research group - Susanne
Balle, HP Technical Computing Division R&D
Compiler and Tools Architect, founder and
Open Grid Services Architecture
Working Group - Jeffrin Von Reich, project
UK Particle Physics and Astronomy Research
Council (PPARC) - Paul Vickers of HP Labs Bristol is on the
Grid Steering Committee
UK e-science Security Task Force - Antonio Lain of HP Labs Bristol is a member
UNESCO (United Nations Educational, Scientific and Cultural Organization) - improving scientific research in southeastern Europe by providing Grid computing technology to four universities.