Grid Computing article – MIS magazine

Grid Computing

Computers unite
MIS Australia
By Sarah Stokely

The time is coming when millions of computers will be connected by grids that enable each to feed off the others’ resources.

When Bill Gates delivered the keynote speech at the Supercomputing
Conference in Seattle last November, he became the latest vendor
heavyweight to step into an industry-wide debate on the use of grid
computing within information technology.

Gates, using his keynote address to talk up Microsoft’s vision of a
$10,000 supercomputer under every desk, is the latest voice to join the
chorus singing the virtues of taking high-performance computing beyond
computational modelling into the realms of web services, business
intelligence and data mining.

While vendors continue to push the dream of harnessing grid computing
in business, it is clear that the largest strides for this technology
continue to be made in academia and scientific research.

In Europe, work is under way on the world’s biggest computer grid, the
LHC Computing Grid Project. It is being built to support the Large
Hadron Collider which, when completed, will be the world’s largest
scientific instrument. The LHC will enable scientists to explore the
structure of matter by forcing protons and ions into head-on
collisions.

Based at the European Organisation for Nuclear Research (CERN) in
Switzerland, the LHC will be capable of processing more than 15,000
terabytes of data per year. The LHC Computing Grid will eventually link
150 supercomputers across the world to help process these huge amounts
of data, the world can pool their resources and conduct research on any
number of topics.

Moving towards e-research involves new methods and technology, Abramson
says. “The idea of the grid is not just high-performance computers but
very large data stores and scientific instruments,” he explains.

Abramson likens the access to distributed resources that grid computing
provides to the use of the internet to tap into global information
using a single protocol, http.

He says grid computing’s use in Australia is starting to spread further
afield than academia, aided by the development of powerful software to
harness its potential.

Nimrod, a Monash-developed parametric modelling system designed for use
across a range of scientific fields, is an example of what Abramson is
talking about.

“Nimrod lets you explore design options. It can be anything from designing a car to drug modelling,” he says.

“The trick is to produce generic tools to help solve particular
problems. If a solution is messy, scientists won’t bother using it.”

Beyond academia
Grid computing has already entered industries that require huge
computation power for solving complex problems. These include
biotechnology, where it has aided research into mapping the human
genome and protein folding.

Another application has been in the automotive design industry. Car
makers including Ford, Toyota and DaimlerChrysler use supercomputers to
simulate car crashes and model aerodynamics.

What about the broader applications of grid computing? Vendors are
talking up the possibility of moving beyond computational grid
computing to data-centric grid computing.

When speaking of the benefits for business, vendors tend to extol the
grid’s ability to harness computational power across geographic and
network boundaries, share resources and to take advantage of system
downtime during off-peak hours to maximise returns on investment.

Commercial reality
“The new
challenge is seeing what grid computing can do in the commercial
world,” says Bill Caelli, the assistant dean for strategy and
innovation in the IT faculty at Queensland University of Technology.
The prospect of being able to move “the program to the data, not the
data to the program” has the potential to cut costs, he says. However,
legal and privacy issues would need to be considered.

Research analyst Kevin McIsaac, of Intelligent Business Research
Services, believes that “in terms of classic enterprise work, there’s
not a lot being done in Australia”. But he says grid’s ability to shift
workloads to different machines could have potential applications for
data warehousing.

Gartner Group research vice-president Phil Sargeant says grid computing
is not really on the Australian corporate landscape yet.

“It’s pretty over-hyped,” he says. “Grid computing is very much [an]
area of academia more than anything else today. If you’re talking about
grid as it relates to commercial usage, it’s in its very early days.”
Users of grid technology tend to be in industry niches such as
biomechanics, biometrics and financial modelling, areas that focus on
“number crunching” tasks, says Sargeant.

Wider adoption of grid technology would involve an evolution towards a more grid-compatible architecture, he says.

“One of the things about grid technology is it’s really applying a
large number of resources to a single task. You really need to have the
applications that can take advantage of that architecture.”

In contrast, the typical enterprise IT shop is geared towards smaller
numbers of large mainframe computers or servers. Current applications
are therefore geared to those kinds of IT environments.

To take advantage of grid, you’d need to “parallelise” the IT environment, says Sargeant.

These sentiments are echoed by McIsaac: “I think there’s interest, but
the problem is your standard enterprise workloads are typically
transactional support and reporting, which are not that amenable to
grid computing.

“Grid is about taking a workload and breaking it up across a large
number of computers. It needs workloads that can be broken down [into
smaller blocks].”

CIOs who are evaluating grid computing would need to ask themselves if
their organisation had workloads that could usefully be distributed inComputers unite
By Sarah Stokely MIS

The time is coming when millions of computers will be connected by grids that enable each to feed off the others’ resources.

When Bill Gates delivered the keynote speech at the Supercomputing Conference in Seattle last November, he became the latest vendor heavyweight to step into an industry-wide debate on the use of grid computing within information technology.

Gates, using his keynote address to talk up Microsoft’s vision of a $10,000 supercomputer under every desk, is the latest voice to join the chorus singing the virtues of taking high-performance computing beyond computational modelling into the realms of web services, business intelligence and data mining.

While vendors continue to push the dream of harnessing grid computing in business, it is clear that the largest strides for this technology continue to be made in academia and scientific research.

In Europe, work is under way on the world’s biggest computer grid, the LHC Computing Grid Project. It is being built to support the Large Hadron Collider which, when completed, will be the world’s largest scientific instrument. The LHC will enable scientists to explore the structure of matter by forcing protons and ions into head-on collisions.

Based at the European Organisation for Nuclear Research (CERN) in Switzerland, the LHC will be capable of processing more than 15,000 terabytes of data per year. The LHC Computing Grid will eventually link 150 supercomputers across the world to help process these huge amounts of data, the world can pool their resources and conduct research on any number of topics.

Moving towards e-research involves new methods and technology, Abramson says. “The idea of the grid is not just high-performance computers but very large data stores and scientific instruments,” he explains.

Abramson likens the access to distributed resources that grid computing provides to the use of the internet to tap into global information using a single protocol, http.

He says grid computing’s use in Australia is starting to spread further afield than academia, aided by the development of powerful software to harness its potential.

Nimrod, a Monash-developed parametric modelling system designed for use across a range of scientific fields, is an example of what Abramson is talking about.

“Nimrod lets you explore design options. It can be anything from designing a car to drug modelling,” he says.

“The trick is to produce generic tools to help solve particular problems. If a solution is messy, scientists won’t bother using it.”

Beyond academia
Grid computing has already entered industries that require huge computation power for solving complex problems. These include biotechnology, where it has aided research into mapping the human genome and protein folding.

Another application has been in the automotive design industry. Car makers including Ford, Toyota and DaimlerChrysler use supercomputers to simulate car crashes and model aerodynamics.

What about the broader applications of grid computing? Vendors are talking up the possibility of moving beyond computational grid computing to data-centric grid computing.

When speaking of the benefits for business, vendors tend to extol the grid’s ability to harness computational power across geographic and network boundaries, share resources and to take advantage of system downtime during off-peak hours to maximise returns on investment.

Commercial reality
“The new challenge is seeing what grid computing can do in the commercial world,” says Bill Caelli, the assistant dean for strategy and innovation in the IT faculty at Queensland University of Technology. The prospect of being able to move “the program to the data, not the data to the program” has the potential to cut costs, he says. However, legal and privacy issues would need to be considered.

Research analyst Kevin McIsaac, of Intelligent Business Research Services, believes that “in terms of classic enterprise work, there’s not a lot being done in Australia”. But he says grid’s ability to shift workloads to different machines could have potential applications for data warehousing.

Gartner Group research vice-president Phil Sargeant says grid computing is not really on the Australian corporate landscape yet.

“It’s pretty over-hyped,” he says. “Grid computing is very much [an] area of academia more than anything else today. If you’re talking about grid as it relates to commercial usage, it’s in its very early days.” Users of grid technology tend to be in industry niches such as biomechanics, biometrics and financial modelling, areas that focus on “number crunching” tasks, says Sargeant.

Wider adoption of grid technology would involve an evolution towards a more grid-compatible architecture, he says.

“One of the things about grid technology is it’s really applying a large number of resources to a single task. You really need to have the applications that can take advantage of that architecture.”

In contrast, the typical enterprise IT shop is geared towards smaller numbers of large mainframe computers or servers. Current applications are therefore geared to those kinds of IT environments.

To take advantage of grid, you’d need to “parallelise” the IT environment, says Sargeant.

These sentiments are echoed by McIsaac: “I think there’s interest, but the problem is your standard enterprise workloads are typically transactional support and reporting, which are not that amenable to grid computing.

“Grid is about taking a workload and breaking it up across a large number of computers. It needs workloads that can be broken down [into smaller blocks].”

CIOs who are evaluating grid computing would need to ask themselves if their organisation had workloads that could usefully be distributed in that way, says McIsaac.

Security concerns
Caelli is one of Australia’s foremost computer security experts and is frank when discussing grid computing’s potential security risks.
“The question really is: are we building a castle with quicksand?” he asks.

“Essentially, grid computing assumes [you] can trust the software or the operating system … The question you need to ask a CIO is would they really believe this? With the operating system as the critical link, the answer would have to be a resounding no.”

Any move to open up an organisation’s computer network to a grid would need to be done with extreme caution.

“As a security person, what worries me is that grid computing itself could become an attack vector,” says Caelli.

The risks of participating in grid computing would vary according to the scale of the grid, Caelli says. “In a global grid, I have no way of knowing who the other grid participants are.

“There is nothing to stop a grid applet or application taking over your machine, especially if it has admin privileges, which any home computer set up for internet shopping would have.”

Caelli points to the recent revelations about Sony BMG Music Entertainment embedding anti-piracy software in some of its music CDs, which installed root kits on listeners’ PCs, as a reminder that our ideas about trusted sources of information make us fallible.

“How do you know your [grid application] isn’t installing a root kit?” he asks.

CIOs would need to adopt a risk policy to ensure that appropriate security procedures were in place as well as legal agreements with outside parties.

This could include the use of digital signatures to verify that grid applications had come from a trusted source.

Settling standards
As a technology with emerging standards, grid computing is still subject to a fair degree of confusion in the market, Abramson says.

“If you ask 10 people, you’ll get 20 different definitions.” While vendors might contribute to this confusion by using many different names and definitions of grid computing, they also tend to support the evolving open standards, he says.

International grid communities, including the Global Grid Forum, have been working since 2002 to develop open grid standards architecture.
“There is a very strong push for open standards,” says Abramson. “You’ll see vendors are falling in line with that.

They have no choice, they have to be able
to inter-operate, just like the web.”

Despite vendor hype, commercialisation of grid computing is not something we can expect to see in Australia during the next year, Sargeant says. “Most of the multinational vendors have initiatives in the grid computing space.

Standards are being formulated, it is progressing, but it hasn’t progressed sufficiently.”

He believes uptake of grid computing will be helped as standards evolve to address issues such as workload management and security.

The pace of refresh rates of enterprise platforms would also impact on the uptake of grid, as a shift away from monolithic legacy platforms towards a more distributed architecture would come only when organisations were considering different architectures to replace legacy systems.

A decision to move towards grid would also be informed by the applications that are developed for it. “I’m not suggesting that this isn’t going to happen, just that it will take a while,” Sargeant says.

1 Comment

  1. damien_wise Said,

    March 6, 2006 @ 12:26 pm

    Very nice article (been looking forward to reading it for ages). It really captures the wide variety of opinions in the industry on this topic.
    Congrats on getting your freelancing work happening, too!

    Phil “not really on the Australian corporate landscape yet” Sargeant ought to see what a large Australian telco recently used to re-platform one of their brand-name systems with.
    Is it just me, or does everyone from The Gartner Group sound like a tosser or a Microsoft mouthpiece (or both) ?

    For my money, David Abramson nails it.

    (*yawn* been there, done that, wrote the manual. 😛 )