[Beowulf] Exascale by the end of the year?

[Joe Landman]

On 03/04/2014 09:04 PM, Prentice Bisbal wrote:
> I find this whole thing dubious as well, but there is a LOT of money
> being made by high-frequency traders. Enough that a $50 billion
> expenditure could probably be made back in a few years. I don't know any

Not quite.

> exact numbers that these HFTers are bringing in (many of them are super
> secretive), but it's definitely billions in profit per year.  As someone
> who lives and works in the NYC metro area, I've seen how much money some
> of these big investment banks have, and the HFTs are bringing in money
> faster than anyone else right now, so $50 billion isn't outside the
> realm of possibility. If anyone can do this that's not a government, I
> would expect it to be someone in the  HFT world.

Actually, those doing HFT are in an arms race working on ever smaller 
margins, against competitors with similar hardware, and "similar" 
algorithms.

I don't know where the conception that they are rolling in cash came 
from, but I could probably guess.

I think the real question is would the system be viable in a commercial 
sense, or is this another boondoggle?

I'd like to see  us get to meaningful and usable exaflop.  I am 
definitely a speed demon at heart.  But I also want to see us get there 
in such a way that we can build more than one, and not have it cost 
$50B/unit, or $1B/year to operate.

I read the Nicole's article on it, but it was somewhat short on details. 
  So I'll remain a skeptic for now.  Though we'd be happy to supply them 
the fastest exabyte of disk they can get on the planet ... :D



-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics, Inc.
email: landman at scalableinformatics.com
web  : http://scalableinformatics.com
twtr : @scalableinfo
phone: +1 734 786 8423 x121
cell : +1 734 612 4615