Canada to Launch Countrywide Virtual SuperComputer

LadyCatra writes "A serious shortage of world-class computing power in Canada prompted University of Alberta scientists to create the next best thing -- a countrywide, virtual supercomputer. On Nov. 4, thousands of computers from research centres across the country will be strung together by a U of A effort to create the most powerful computer in this country. The full story is here"

Comments

Sun is Right

[e8johan]

"The Network is the Computer"

It would be nice to see a worldwide system. If this is going to work there must be some CPU time quota system, perhaps a quota that can be bought and sold. This could make it interesting for ordinary home users to join (earn quota, sell quota, make $$$). There are many projects in the academic world that could never make a SETI@home launch, since the research is to boring. Still, we need to use all that idle time buring away across the world.

Custom solution for a specific task?

[jukal]

The article does not seem to mention whether they use a ready made grid/distributed computing platform or are they whipping up it themselves? Or am I blind? Does anyone know more about this? And what do they mean by:

"The computers will be linked by the Internet, but involve a simple networking system, Lu said. Keeping the linkage as simple as possible was the goal."

Based on the article I would assume that they have made a custom tailored system (if not kludge) for one specific purpose ("for calculating energy shifts as two molecules are manipulated around 3-D space") - and not a platform which could be easily tailored and managed to solve different kinds of tasks with different kinds of relationships between the tasks.

Ohh, I could also link my grid computing links.

Big Iron.

[jericho4.0]

I thinks they're talking about linking together several (5-20?) large computers over fat pipes, rather than many small ones. Although seeing that all of Canada's reasearch computing power is less than that of the University of Southern Florida, that might not mean much.

And the innovation is where?

[cybercomm]

The computers will work jointly on a molecular chemistry research question that would take a single computer as long as six years to complete. Jonathan Schaeffer and Paul Lu, professors in the U of A's department of computer science, expect their virtual supercomputer will do the work in one -- one day, that is.

So how is this different from DC or SETI?

Why build a new, separate system?

[magnum3065]

rather than joining a currently existing project? I'm a student at the University of Virginia and we have a project like this that's been going on for 5 years now: http://legion.virginia.edu/

They talk about how they feel that Canada should be pursuing its own supercomputing, but why not join up with other universities that have been pursuing similar projects and give Canada access to the computing power of other countries as well? Isn't the goal here for people to work together for mutual benefit? I don't understand why they feel the need to isolate their Canadian initiative, rather than giving Canada the access to computing power far greater than they can acheive on their own.

Check out photos of UVA's branch of Legion: http://legion.virginia.edu/centurion/Photos.html
(I think these are a little out of date. There's a bunch of rack-mount machines in there now too)
This room has big glass walls, and everytime I walk by it I wish I had a room like it.

Re:Wow

[Jezza]

Well I guess the first thing that needed are developer tools to ease the creation of programs to run on the platform - I guess a new project type in Project Builder would help and maybe even some language additions to allow people to more easily create programs. There are a number of challenges involved with creating programs of this type, how do nodes communicate? What happens when a node goes away (someone starts to use the computer for instance) what happens when a new node becommes available? And of course how easy is it to deploy these programs? What you'd like to do is "feed" these programs in via some kind of queue, and allow that queue to be reordered - how does that work? You possibly want to prevent the machines from sleeping or being shutdown, this will also need some UI changes - maybe a machine needs to be shutdown for an upgrade or simply to be moved, how do you over ride the settings? You might also want to see how the programs impact the network (you can imagine that a program could swamp the network with IP traffic if you weren't careful) some form of debuging software that could run on a single machine to simulate it's deployment would also be useful.

Of course Apple have some good tools here - perhaps Rendezvous (Apple's dynamic discovery or services over IP) could help. These such tools could help make it much easier to provide "community supercomputers". This would be especially useful in higher education, a place where Apple has been traditionally strong.

Re:Distributed computing?

[FTL]

>Why didn't they just make a client program for distributed computing so the entire country/world could help out?

Because there will always be creeps who won't play fair. Much of the work that SETI@home does is security, combatting those who would submit false or abreviated results in order to get higher stats. UofA want to do real computing on a variety of applications. They've concluded that it is more efficent (for their purposes) to go for a small pool whose results they can trust, than to go for a large pool whose results they have to check and double-check.

Each approach has significant advantages and disadvantages. It depends on the type of work you are interested in performing.