 |
|
DAILY NEWS AND INFORMATION
FOR THE GLOBAL GRID COMMUNITY /
|
Special Features:
A LOOK AT INTEL'S GRID
STRATEGY
By Mike Bernhardt
On May 24, GRIDtoday 2004 (Gt'04), a conference focused on the commercial
applications of Grid computing, will kick off in Philadelphia.
Intel is one of the leading players in Grid computing and is also one of
the
sponsors of Gt'04.
Gt'04 conference program director, Mike Bernhardt, caught up with Intel's
director of Grid strategy and business development, Robert Fogel, to get his
perspective on the growing momentum of Grid computing and to learn more about
Intel's role in this exciting market segment.
Bernhardt: Why have we seen such a tremendous surge of interest in Grid
computing over the past year? What is driving this momentum?
Fogel: Inspired by the tremendous growth and impact of the Internet, Grid
is
also seen as having this kind of potential. We see three major world-wide
market waves afoot. The first wave is the national, government and academic
HPC arenas, and is already well-established as evidenced by the NSF TeraGrid
in the U.S., the European Union DataGrid, the China Grid, as well as many
others. These are the true pioneers, thought-leaders and early adopters of the
very latest technology. The momentum here is primarily driven by an insatiable
appetite for compute power and the need for large-scale,
geographically-independent collaboration.
The second wave, which is beginning to gain momentum now, is primarily
made-up
of commercial enterprises particularly in financial services, pharmaceutical
and manufacturing markets where this likewise early adopter HPC community sees
Grid as playing an essential role in their ability to remain competitive.
Grid will make it possible for them to do what could not be done before, such
as real-time client services, new drug discoveries, or going straight from
design to production; and, these capabilities are made possible by
significantly lowering TCO, as well as increasing utilization of compute and
information resources.
The third market wave of Enterprise IT Grids is also primarily made-up of
commercial enterprises, but addresses the sharing and management of all
enterprise compute and information assets including hand-held devices,
laptops, appliances, applications, databases, visual displays, edge servers,
mid-tier servers, as well as back-end servers. This market is in the very
early stages of development and will depend heavily on well defined standards,
flexible building blocks, end-to-end solutions, and a broad range of services
in order to realize the vision of a fully distributed infrastructure. The
momentum for this third wave will primarily be driven by a new opportunity for
IT that lets IT redefine itself from being in the background, to being an
integral part of the business value that an enterprise can offer to its
customers.
Bernhardt: What is Intel's interest in Grid computing?
Fogel: Grid is an important evolutionary step that continues along the path
from proprietary infrastructures to commercial-off-the-shelf building blocks
and solutions that ultimately give the end user maximum value, choice and
flexibility. The real opportunity for Intel is to offer end-to-end
architecture solutions that include mobile devices, laptops, workstations,
networks, servers and large-scale SMP products. Grid is going in a direction
that is very consistent with Intel's vision of autonomic and pervasive
computing based on modular, interoperable components that are fully
distributed across an infrastructure. Intel is deeply committed to driving
standards toward a convergence that enables volume economics and thus drive
global proliferation.
Bernhardt: How does the growing momentum and acceptance of Grid computing
impact Intel?
Fogel: As Grid gains momentum and is more widely accepted across multiple
vertical market segments world-wide, the unique and added value that Intel can
provide will continue to move deeper down the Grid solution stack. This value
may include reliability, manageability, security and software tools.
Bernhardt: Does Intel have channel partners that are specific to delivering
Grid solutions or applications based on Intel technology? Can you name some of
them?
Fogel: There are currently many companies that provide solutions based on
Intel architecture products. Some of these companies include HP, IBM, Oracle,
SGI, Platform, DataSynapse, United Devices, Entropia, Enigmatec, Axceleon and
others.
Bernhardt: From the HPC perspective, will large clusters and supercomputers
eventually become computing resources on departmental and enterprise
Grids?
Fogel: Yes, and in the case of static clusters, many of those will move to
cluster Grids, in which the nodes of a cluster are dynamically configured by
Grid middleware to operate as a virtual cluster for a finite period.
Bernhardt: Do you think many large organizations are currently looking at how
to integrate their cluster and Grid strategies?
Fogel: For many organizations, their Grid strategies are actually growing
out
of their existing cluster (or HPC) strategies. Grid offers the opportunity to
share, and thus increase, the utilization of clusters which can in turn
increase the total available compute power. So, Grid strategies are often a
natural extension of an existing cluster or HPC strategy.
Bernhardt: Do you think the decision makers for adopting "clusters" into the
enterprise are the same decision makers that will adopt "Grid" into the
enterprise?
Fogel: The decision makers who deploy clusters today are often at the
line-of-business level, such as an engineering department doing crash
simulation, and thus corporate-level IT may not be involved. It is more likely
that corporate IT will be the primary decision maker as Grid spreads across
the enterprise and thus even clusters will become IT managed resources.
Bernhardt: What is the role of Intel's Itanium processors, current or next
generation, in enabling the adoption of Grid computing?
Fogel: The Intel Itanium processor is a true 64-bit architecture and is
designed to handle applications that are very computationally intensive, and
that have large memory requirements. So, Grid adoption, especially in the
context of the enterprise, must address a broad range of applications from
those that can run on desk tops (based on Pentium 4 and Intel Xeon) to those
that require much greater capabilities from the underlying platform while
still being able to rely on the availability of non-proprietary
commercial-off-the-shelf technology.
Bernhardt: Who are some of the leading commercial organizations that have
successfully implemented a Grid computing environment using Intel architecture
platforms or other Grid-enabling technology from Intel?
Fogel: There are significant numbers of commercial organizations having
successfully deployed Intel-based Grids; however, many consider their use of
Grid technology to be a closely guarded secret and thus less willing to
identify themselves. A small sample of commercial organizations who have
publicly disclosed their success with Intel-based Grids include Charles Schwab
Corporation, Intel's own EDA Grid, Novartis AG, Shell, Butterfly.net, and IBM
Global Services. There are also non-commercial organizations that have
successful Intel-based Grid deployments, and either have commercial programs
or ties. Some of these organizations include the American Diabetes Association
(ADA), the NSF TeraGrid which is built on more than 3000 Itanium processors,
CERN and the European Union Data Grid, CURE, BIRN, China Grid, Canada's
WestGrid and many others
Bernhardt: Where does Grid computing business development fall within the
Intel organizational structure?
Fogel: Generally speaking, business development is the responsibility of
Intel's Business Development Organization, or BDO; however, there are many
other organizations within Intel that support the BDO. Some of these other
supporting organizations include the Enterprise Products Group, the Parallel
and Distributed Solutions Division, Intel Solution Services, Software
Solutions Group, Solutions Marketing Development Group and many others, as
well.
Bernhardt: Has Intel invested (Intel Capital) in any Grid middleware or Grid
solution companies?
Fogel: Intel Capital is Intel's primary investment arm and has made
investments in DataSynapse, United Devices, Scali, plus others related to
Grid. There are also other Grid-related investments in the pipeline.
Bernhardt: How important is the role of the GGF (Global Grid Forum) in driving
the adoption of Grid computing to the commercial enterprise?
Fogel: Standards are essential to enabling the volume economics that in
turn
drive global proliferation; and, the GGF is establishing itself as the
definitive world-wide standards body for Grid. The GGF also serves a very
important role to provide leadership throughout the entire community of
solutions and building blocks suppliers, and across other standards groups
such as the DMTF (Distributed Modular Task Force).
Bernhardt: It is often said that the abundance of marketing hype and lack of
consistent definitions are causing tremendous confusion and are in fact a
barrier to wider adoption of Grid technology. We recently surveyed several
dozen CxO-level executives of Fortune 500 firms regarding their Grid computing
strategies. Every single executive told us that Grid computing was on their
road map -- but a large majority also stated that they had no idea what that
really means at this time. What is your perspective on this?
Fogel: Although Grid has significant momentum in the academic and
scientific
arenas, it is still in its early stages of adoption within the enterprise; and
even then, it is still mostly an HPC phenomenon. The lack of consistent
definitions causes the hype, and the lack of consistent definitions is the
results of not having complete and widely accepted standards. Since the
academic and scientific arenas are the ones defining the standards, they are
not impeded in the same way that a commercial enterprise is impeded. A
commercial enterprise relies heavily on commercial building blocks and
solutions which in turn rely on relatively mature standards.
Grid is on Fortune 500 road maps because there is faith that Grid can
deliver
real business value, and that Grid has the potential of having the same impact
that the Internet has had on their businesses. This faith is also partly
inspired by the proof-of-concepts that exist in the academic and scientific
arenas, as well as in commercial HPC.
Standards efforts have made significant progress over the past year,
especially with the recent announcement of WSRF in which Grid and Web Services
have merged thus mitigating some of the adoption barriers.
In order for an enterprise to really understand what Grid means in terms of
it's own business planning, proofs of concept will have to used to demonstrate
business value and benefit, as well as used to discover holes in the existing
solutions, building blocks and even the specifications themselves, so that
they can be fortified for the enterprise.
Bernhardt: As far as a Grid computing road map, several analysts seem to agree
that 2004 and 2005 will see the first evidence of a chasm jump as the early
adopters start to report on their Grid implementations. While a few of these
early adopters are already talking about success and measurable ROI, most
agree the standards and security issues have a long way to go. Please comment
on this.
Fogel: The 2004 and 2005 time frame is consistent with the current pace of
converging standards and the development of solution stacks for commercial
applications across a number of vertical market segments. The early adopters
of academic, scientific and even commercial HPC either don't rely heavily on
commercial building blocks and solutions; or, the scope of deployment is
limited enough that many of the standards areas that are still too early for
implementation do not have a significant impact on what they intend to
accomplish in the near term.
In particular, security can have radically different meanings, relevance
and
levels, depending on whether you are talking about being inside the firewall,
or within a single domain, etc. Current security standards are considered
adequate for what is being done today; however, a commercial enterprise for
instance, deploying a Grid infrastructure across multiple geographical sites
and in collaboration with their supply chain and partners, will certainly
require more rigorous security standards.
Bernhardt: Looking forward, can an HPC vendor stake out a credible position
without having a Grid strategy?
Fogel: Given that most HPC end-users see Grid as key to their own
strategies,
it's difficult to see how an HPC vendor could stay in touch with their
customers without having a Grid strategy in place.
Bernhardt: Does Intel have a Grid deployed for internal use? If so, can you
give us a brief description of the physical components of that Grid and how it
is used.
Fogel: Yes. Intel uses Grid technology that has been developed over the
past
10 years specifically for doing processor design and validation. Approximately
30,000 workstations and servers are managed world-wide, and around 3 million
jobs per month are executed. Results have yielded many millions in savings, as
well as accelerated chip schedules, and increased testing for fewer mask
spins.
The current Grid infrastructure is much like a distributed resource manager
similar to those found in commercial products such as Platform's LSF or
DataSynapse's GridServer, but Intel's current Grid technology is highly
specialized for EDA (Electronic Design Automation); and, a new infrastructure
is being planned now for a far more extensive Grid based on industry-accepted
Grid standards.
Bernhardt: Where do you see the most significant advances in terms of types of
Grids? By that we mean departmental, enterprise, application specific (such as
Science Grids), shared data Grids, etc.
Fogel: Today, the low hanging fruit are HPC Grids which include mostly
Compute
Grids but also includes some Data Grids as well; and, the most significant
volume of these types of Grids is mostly in academic environments, national
labs and national government projects. Enterprise HPC Grids are being rooted
today with technology from Platform, DataSynapse, United Devices, Avaki and
many others. The most significant advances will likely be with Enterprise IT
Grids, which are Grids across an enterprise as well as the datacenter and used
for broad-based business computing such as SCM, business processing, ERP,
Finance, etc. Enterprise IT Grids offer the greatest potential in terms of
market segment opportunity and impact. Enterprise IT Grids are also going to
require significant advances in the current state of standards, and the
availability of building blocks and solutions.
Bernhardt: How important is Grid computing to Intel's market segment
development strategy?
Fogel: Grid, is an emerging technology that has the potential for making a
positive impact on academic and scientific communities, as well as the
world-wide economy at large.
Intel takes pride in being on the leading edge of innovation across many
different fronts.
About Robert Fogel
Robert Fogel directs world-wide Grid strategy and business development for
Intel. In this capacity, he works closely with key, strategic players in the
Grid community, as well as coordinates a cross-Intel Grid effort. Fogel has
spent several years in High Performance Computing to promote standard
Intel-based clusters and large-scale SMP servers as the most cost efficient
means for delivering maximum compute power, having the most diverse choice of
hardware and software vendors, and benefiting from a highly scalable
infrastructure. He spent a significant part of his career as a systems
architect for very broad range of development efforts including network
servers, RAID systems, optical storage devices, reusable flight hardware and
software for communications satellites, reproducing pianos, digital
signal/image processing solutions and antenna control systems for NASA's Jet
Propulsion Laboratory Deep Space Network. Fogel received an Electrical
Engineering degree from Cornell University.
Web services, utility computing, .NET, CPU harvesting and distributed
computing are just a few of the technologies that fall under the Grid
computing umbrella. Gt04 -- a premiere enterprise Grid computing conference
targeting industrial and commercial users -- will gather experts, and outline
strategies and road maps for Grid deployment. For more information, visit
www.gt04.com.
Grid computing is here!
|