July 18, 2005
Defining Grid a Hot Topic at VIP Summit
By Derrick Harris, Editor
If the attendees of the GRIDtoday VIP Summit take one thing away from
the event's first day, it might very well be that Grid computing --
both the technology and the market -- is desperately in need of a
standard definition.
In a morning panel led by HPC veteran Mike
Bernhardt, a quartet of industry analysts were asked a battery of
questions designed to shed light on a market mired in confusion. For
example, when asked if the various definitions of Grid computing act as
a barrier to accelerated market acceptance, all four panelists -- Dan
Kusnetzky and Addison Snell of IDC as well as Martin McCarthy and Steve
Wallage of The 451 Group -- seemed to agree that while the definitions
might not be a barrier, they definitely muddy the waters for everyone
involved.
According to Kusnetzky, he is tracking 14 different
buzzwords that people he speaks to equate with Grid computing,
including technologies such as blade computing and network-attached
storage (NAS). There needs to be clarity, he said.
IDC's Snell
pointed out, however, that in the HPC market, there is little confusion
among end-users as to what constitutes a Grid. Users know what they
want, he said, but vendors do suffer as a result of the confusion when
they try to lay out Grid packages that define what, exactly, customers
can expect from a Grid infrastructure.
Just how much does this
confusion affect vendors? According to The 451 Group's Wallage, it can
be pretty intense. He remarked that start-ups that use the term "Grid"
in defining what they do are often "dead in the water." Many will use
other terms, such as "next-generation computing," as they don't want to
be directly associated with Grid, he said.
Even among Grid's
biggest proponents, there is an obvious disconnect. The 451 Group's
McCarthy cited as an example the Oracle Grid, which he said is not even
a Grid at all. In fact, when moderator Bernhardt posed the question of
how Grid will affect the cluster market, McCarthy quipped, "I think the
Oracle Grid will definitely drive the market for clusters."
Although
the panelists agreed that clusters and Grids are not competitive, there
still is confusion. Perhaps it is because some perceive a main draw of
Grid computing to be its ability to harness the power of all the
connected processors -- something for which clusters are specifically
designed. In addition, clusters often act as nodes on larger Grids -- a
relationship that can be hard to grasp for potential users not well
versed in those technologies.
This topic came up again in a
presentation by GGF chair Mark Linesch, when he acknowledged the line
between clusters and Grids can be fuzzy at times. "How many clusters
could a clusterer cluster if a cluster could cluster clusters?" he
joked. "And when is that a Grid?"
If all this discussion about
the difficulty in defining Grid computing, and the ill effects it can
have on the market, wasn't enough, the point was drilled home during
the final presentation of the day. DataSynapse's chief marketing
officer, Kelly Vizzini, laid out five reasons why more companies are
not adopting Grid computing. As expected, her first reason is an
overall lack of understanding among potential users. Vizzini then,
however, suggested instead of wasting time on definitions, vendors in
the Grid space should spend their energy answering questions about what
Grids can do for a company.
This suggestion is in stark contrast
to suggestions made by Bernhardt during the panel earlier in the day.
His advice: vendors should go to great lengths to define how you define
Grid. Upon closer inspection, though, the two ideas might not be that
different, after all. If vendors can present prospective users with a
concrete definition of Grid -- what it is, what it isn't, what it can
do and what it can't do -- those users should be able to decide whether
Grid is truly the solution for their problems.
So, what can the
Grid community take away from this important discussion? The answer is
in how it presents the technology to end-users. Whether one standard
definition of Grid computing will ever emerge remains to be seen --
although events like the GRIDtoday VIP Summits could go along way to
making this happen. The real challenge remains making users understand
all that Grid has to offer and how it can help them achieve their goals.