[LINK] Supercomputing for everyone

[tomk]

On 19/09/2012 9:59 a.m., Jan Whitaker wrote:
> At 01:21 AM 19/09/2012, stephen at melbpc.org.au wrote:
>>> It looks to me like yet another cloud solution looking for a problem.
>> Take your point Bernard. Mind you, thinking about it, being cloud-based,
>> data will be in/of 'known' formats, so, if our many cloud-providers work
>> hard regards their interface & modules (eg a wind-tunnel module) whereby
>> one simply plugs-in data, and plays what-if, one can see it being indeed
>> a valuable and hither-to unavailable SME design and development resource.
> I didn't give this much thought when posted, but what Stephen is
> pointing out could also be extended like this. If we are graduating
> people with super-computer training and access in universities, it
> would be great for start-up businesses challenged by access to
> capital to gain access to these sorts of resources. It may not be a
> huge number in the early days and maybe not optimally, but as large
> computing power networks for engineering problems become available,
> people will figure out ways to use them. In a way it's a 'open it up
> and they will come' approach.

With the exception that with GPU processing around the 88K cost per 
Petabyte, the supercomputers that one can build in ones backyard shed 
are a lot cheaper than the living costs for even one post grad or 
candidate student who might be moonlighting part-time at an SME.
Or looking at it from a different perspective : The consumer edge of the 
smartphone network now exceeds an exabyte.
As the P2P Smart Grids start to link all of those ARM based handhelds, 
the concept of a service industry being built on the back of providing 
super-computer services seems almost laughable.