Is Grid Computing the next 'big thing'?

Grid computing (GC) is a high concept, a computing grand
challenge, and the "next big thing." What GC is not
is a killer application that end-user enterprises need to go out and get. In
fact, an IT department couldn’tbuy
or build one if it tried. It is the latest hot thing to hype, and like moths to
the flame, vendors, professionalservices
firms and even some industry analysts are drawn inexorably to it. Caution and
selective debunkingare clearly
appropriate. Yet as infrastructure, the high concept of the computing (and data)
grid will succeedin driving
technology innovations, leveraging Web Services and, ultimately, redefining the
limits of thepossible.

One reason we are talking about this now, aside from old
marketing campaigns having run out of gas, is grid computing
has established a track record in scientific collaboration. Examples include
project-level gridconstructions
from organizations such as NASA, the Argonne Super-Computing Lab (University of
Chicago), Max Planck Institute or the
European high energy physics laboratory at CERN where petabytes ofastrophysics
data or the data output from atom smashing require crunching. A high-profile
scientific example,of which the
average person is likely to have heard, is of grid processor cycles as part of
SETI@home, the

Search for Extraterrestrial Intelligence (http://www.seti-inst.edu/).
Bioinformatics is also an active area ofgrid
construction and subsequent collaboration between research centers in different
governmental andprivate
enterprises into which vendors such as IBM
(DB2), Oracle
(10g) and Teradata
(Teragenomics) areactively
positioning products and services with a grid flavor to them. IBM is promoting
grid solutions inaerospace,
automotive, electronics and finance, most with a significant scientific
component, high performancehardware
and a generous admixture of custom consulting services. Obviously, new
challengesawait as the concept
and the corresponding technology migrate in the direction of commercial
applications.

Unlike "on demand," "unbreakable" and
"demand more," the computing grid is something to which the average
consumer can relate. After the power failure of Aug. 14 in
the US and Canada (and in Italy
on Sept.27), everyone knows what
a grid is —– it is a metaphor for shared resources. Of course, the average
personwill think of the
electrical grid. This association is less fortunate than it was a few weeks ago.
The computinggrid is an
extension of the metaphor in the direction of shared computational services, and
it suffers fromsome of the same
problems – too many moving parts, too hard to monitor, not enough standards
andsometimes too many
incompatible standards.



Call it monumentally unfortunate timing, but no end-userenterprise
can be expected to have much patience for a marketing campaign that promises a
grid or acomputer system that
behaves like one and fails unexpectedly, given that we already have an electric
powersystem that does. Further
explanation of what works and what doesn’t and why is needed. Send in the
plumbers! That is, send in the
infrastructure specialists.

The extreme example that best represents the computing grand
challenge in grid computing is the vision of an application
sharing resources from multiple computers. But think about it. The application
was written toexecute on Windows
and the grid includes Unix, Linux and zOS systems. Where’s the abstraction
layer toenable those Web
Services to be shared? Similar considerations apply to different data formats
– how arethey to be
reconciled? How is security handled? Suppose a resource petitions to join the
grid – how can thegrid
administrator verify that the applicant does not innocently bring a virus
infection with it? Initially thegrid
was conceived to share the vast unused resources of idle processors. But what if
the application ismission
critical – where is the software to negotiate and implement the quality of
service (QoS) to beprovided?
Standards such as the Globus toolkit are coming online, but uncertainties
outweigh what is known.

As with most new concepts and new technologies, the terminology
is in flux, and different collaborators use the
language in diverse ways to emphasize the importance of their own contributions.
Since "grid" isbasically
a metaphor, it is hard to say that any given extension or spin is wrong in
itself. However, some useswill
be more mainstream than others. The original problem requiring the construction
of grids in the scientificcommunity
was driven by (1) the need for collaboration across organizations, (2) the
opportunity to poolcomputing
resources to exploit unused processor cycles and (3) data volumes and
computationalrequirements that
exceeded the capacity of existing supercomputer resources.

This results in the
basicdefinition of the computing
grid as the coordinated sharing of resources and coordinated problem solving in
dynamic context of multi-institutional
virtual organizations (see www.globus.org).
This immediately narrowsthe
definition significantly, though alternative use cases may well bring other
examples back as proto-grids orquasi-grids.
Thus, one can call a clustered computer a grid, but it will not be
multi-institutional. One can callpeer-to-peer
file sharing a grid, and it will be multi-institutional, but in P2P the
participants are not workingon
solving the same problem or application. Likewise, computing resources (file and
processor) distributedacross a
campus or enterprise may be called a grid, but they are not multi-institutional
nor are the resourcesusually
attacking the same problem.

Examples of collaboration in the commercial market that can
exploit grid computing include vendor managed inventory,
where large volumes of point of sale data require aggregation into a data
warehouse, possiblyacross
multiple sites, that must be accessed by multiple suppliers. This is
fundamentally different than using aWeb
browser and Secure Sockets Layer (SSL) to look through someone’s firewall at
what is basically acentral data
warehouse. Service bureaus and ASPs are also strong candidates to benefit from
shareddistributed resources that
are provisioned across organizational boundaries.



Another example: Financialservices
present requirements for fund and information transfers between different
organizations across highly secure networks. The global digital economy is the
ultimate virtual organization in which all playershave
an interest. Right now these operations are being handled by file and data
packet transfers betweenfinancial
institutions, including the Federal Reserve Bank(s). The idea of function
shipping the computationalanalysis
rather than the data, in effect turning the system inside out, will give you
some idea of just how high– and how far into the future – the concept of a
financial reconciliation grid really is.

Recommendations

Grid computing is a research project, not a requirement for
end-user organizations. Absent a specific high performance application
requiring collaboration across organizations – think of aerospace, automotive
engineering or genetic design –
commercial enterprises should monitor developments and research, but take
no further action. Vendors such as IBM,
Oracle, Microsoft and
the infrastructure vendors (Sun,
EMC, HP,
etc.) will find opportunities for
technology innovation and marketing hype within the high concept of thecomputational
grid. Investors will find opportunities to roll the dice on technology
innovation and marketinghype.



 Hackers will find a new source of mischief. The grid is not Oracle RAC, it
is not IBM ICE or MicrosoftDNA.
It is a vision of shared computational, storage and application resources,
spanning the virtualorganization,
in a highly heterogeneous environment, and being demoed now at academic,
scientific anddistributed data
centers. IBM has the most realistic concept of the grid and is telling the truth
about

(1) thecomplexity
of the challenge of getting different technology stacks to collaborate over the
grid "on demand,"

(2) the need for integration of heterogeneous data, processing
and applications computing resources and

 (3) the
requirement for work on standards such as the Globus toolkit and a next
generation of Web Services.

Beskeptical
of tools and technology that are repositioned as "grid offerings" in
order to make them soundattractive;
but be prepared to be amazed once again by technology within five years as the
bounds of thepossible are
expanded by innovations spun off from the initiative, even if not all of the
promises can be kept.