Trial and error

Victoria University in Wellington has done a small-scale experiment in grid computing, says computing science professor John Hine.

Victoria University in Wellington has done a small-scale experiment in grid computing, says computing science professor John Hine.

“We took a simulation of storm weather patterns and put it on four processors linked over [metropolitan area network] Citylink. It was faster than doing it on one processor, but slower than using four processors locally”, because of the latency of communication.

This, experiment, he acknowledges, is not an indictment of grid computing as a concept; it was a very small and artificial example.

Beyond the technical questions of efficiency and co-ordination “there is a whole bunch of other issues. The models we’re usually looking at involve using the computing power of other organisations, and that raises the question: do you trust them to leave your program alone?” This is the reverse of the normal security situation, which is handled by a firewall keeping out strange programs. Here the hosts are welcoming your program in, and you have to be wary of them, he says. But of course, the conventional risk of your program compromising the host’s system is still there too.

Most of the experiments so far done with grid computing have been in a collegial, mutually trusting environment.

“Turning any program into a parallel version is a challenge; not all programs are ‘parallelisable’.”

Questions of scale also arise in a comparatively small environment like this country. There is no question that problems exist for which grid computing is a solution; “the question is, do they exist in New Zealand?”

With computer power proverbially doubling every 18 months and the cost of parallelising programs being “non-trivial”, the business decision might for a long time come down in favour of “waiting another few months for a bigger box”.

Animation company Weta, with the huge calculations needed to create special effects for Lord of the Rings and other movies, might be expected to be a potential customer for grid computing. “We’ve toyed with the idea on occasion,” says systems administration head Shane McEwan; “we’ve thought about it as a cool thing to try out, but we’ve never used it. It’s not really practical for us because the nature of our work is pretty sensitive.” This makes it inadvisable to entrust it to computer systems owned by others.

Also, Weta uses a lot of proprietary software, which will not run on a wide variety of platforms, and this makes grid computing less practical, he says.

Join the newsletter!

Error: Please check your email address.

Tags Grid Computing

More about Victoria UniversityVictoria University

Show Comments
[]