Top 500 trends

Florent Calvayrac Florent.Calvayrac at univ-lemans.fr
Wed Nov 27 04:26:56 PST 2002


Robert G. Brown wrote:

> 
>   a) Buy the systems all at once in the beginning.  Let's call the speed
> of each system in year 1 "1".  This buys you 50 units of computing.
> We'll assume that the probability of system failure is 10% per year, but
> that the first year is under warranty.  Thus
> 
>   (end of) year     work done in year      cumulative work
> ==========================================================
>         1                  50                    50
>         2                  45                    95
>         3                  40                   135
>         4                  35                   170
>         5                  30                   200
> 
> or you can:
> 
>   b) Buy $10K worth of systems a year, but each NEW year the systems you
> buy double in speed.  Thus
> 
>   (end of) year     work done in year      cumulative work
> ==========================================================
>         1                  10                    10
>         2            9+20= 29                    39
>         3         8+18+40= 66                   105
>         4      7+16+36+80=139                   244
>         5  6+14+32+72+160=284                   528
>


OK, but you forget that (at least here in France) you can't just
order a cluster, but have to go through a very long and painful
tender procedure. Doing it once every five years is more than enough.
Besides, if you don't waste 100% of the budget the first year
you get it after writing detailed proposals, colleagues tend to
"steal" it arguing that if you do not spend money, you do not work,
and it can even be affected to things paid for by the students and
the state  trough an erosion  process (10% per annum for each budget) 
supposed to pay for accountants and other common ressources(fluids, etc).

There is also a lot of time spent in your idea adapting to new hardware
(drivers, etc) unless there is a competent, dedicated person caring for 
it ; the salary of whom does not come into your computations, making
large computational facilities more relevant.


So when building clusters, one learns about parallelism, systems,
networks, air conditioning, cooling, power distribution....dimensand
a mass of administrative and legal trivia about european competition
rules, tenders, etc....





-- 
Florent Calvayrac                          | Tel : 02 43 83 26 26
Laboratoire de Physique de l'Etat Condense | Fax : 02 43 83 35 18
UMR-CNRS 6087         | http://www.univ-lemans.fr/~fcalvay
Universite du Maine-Faculte des Sciences   |
72085 Le Mans Cedex 9




More information about the Beowulf mailing list