[geeks] fwd: IBM supercomputer dual-boots Windows and Linux

[Sandwich Maker]

" From: Shannon Hendrix <shannon at widomaker.com>
" 
" On Jun 19, 2008, at 18:35 , Sandwich Maker wrote:
" 
" > " From: "Jonathan Katz" <jon at jonworld.com>
" > "
" > " []
" > "
" > " I'm still at odds with calling the cluster of whatever-1U-du-jour
" > " hooked together with Gigabit Ethernet or Myrinet a supercomputer  
" > even
" > " if the top500 list does. To me a super computer is a big monolithic
" > " beastie that has a single OS image on it.
" >
" > i can live with hardware clusters [though i think linking machines
" > with a network is rather stretching it],
" 
" Down this path lies madness...
" 
" I don't see what the connections have to do with anything, since "a  
" computer" here and there has used networks for system I/O for decades  
" now.  Most PC motherboards do it these days too.
" 
" Also, an awful lot of "a computer" systems run more than one CPU and  
" often not all of them run the same OS.
" 
" Is a Mac Pro with an SMC controller, a RAID controller, and several  
" drives "a computer"?
" 
" After all:
" 
" The SMC controller is a tiny CPU with a tiny OS.
" 
" The RAID controller is a fast CPU with a non-trivial OS.
" 
" Each SATA drive has a CPU and an OS, and some of them even have more  
" than CPU on those little controller boards.
" 
" Even the nVidia graphics card in mine has a small object-oriented  
" operating system in it complete with threaded process and I/O  
" scheduler, etc.
" 
" I've been amazed at the devices where I've found ROM chips that have  
" VxWorks stored on them, which were part of "a computer".
" 
" By your definition, my Mac Pro is not "a computer", it's several  
" combined together.

mainframes have been this way since at least the '60s - the 'real'
computer and a bunch of specialized 'attached processors'.  we still
call them 'a computer'.

" I agree that a cluster of PCs is a "bunch of computers", but I wonder  
" why we are willing to draw the line there when a lot of things we call  
" "a computer" are also made up of 2 or more computers.

most of which aren't [re]programmable and are for specialized
functions.

" You have to admit, the lines are blurred a bit...
" 
" I think the big issue here is one of convenience, not technical  
" accuracy.

for me the lines are general purpose vs. specialized system
architecture and loose vs. tight clustering/networking*.  as far as i'm
concerned, this 'supercomputer' sounds like a herd of pcs running
whatever, with some handwaving program that distributes a job over
them [mostly] invisibly to the operator.

from your definition, perhaps all the pcs running that nasa
screensaver are part of a single supercomputer.

" We like for things to be clarified and classified, fitting into neat  
" little boxes, and it's actually quite rare for that to be true.
" 
" We love our illusions.

i recall one of my early sysadm gurus saying how he got nethack to run
on an ibm storage processor so it wouldn't show up in the process
table...  i guess that particular ap wasn't quite as specialized as
ibm intended it to be.

* a tight cluster would be one that shares memory, i/o, etc.  a loose
  cluster is one where each cpu has its own i/o and memory, and could
  easily be run as a standalone system.
________________________________________________________________________
Andrew Hay                                  the genius nature
internet rambler                            is to see what all have seen
adh at an.bradford.ma.us                       and think what none thought