Advice on developing Parallelized / Distributed Application

[WMK]

I am a total newbie to parallel and distributed computing and I need
some advice from anybody kind enough to guide me on ways to solve my
problem. As part of scientific research at my university, I frequently
have to do complex thermochemical equilibrium calculations which could
be quite resource intensive and time consuming even on the latest
generation of standalone workstations. For the purpose of
thermochemical calculations, I use a proprietory state of the art but
commercial programming library in the form a .dll that allows calls to
a programming library which is coded in FORTRAN but supports function
calling via C and C++. Therefore I dont have access to the actual
source code of this library so I cant parallelize the actual algorithm
code.

After learning about distributed and parallel computing, I am
contemplating on implementing either a MPI implementation or a
distributed implementation for use in an office environment
Heterogenous cluster . Since I cannot optimize the actual algorithms,
MPI wouldnt be of much help. Distributed computing with RPC might be
an option where in I can distribute the entire computation over a
heterogenous networked cluster of office machines.

Can someone please guide me whether such an implementation is possible
or suggest me any alternate methods whereby I can make use of idle lab
machines to speedup my calculations.

Note: There is currently no possibility for me to use any other
programming library be it opensource or commercial. Therefore anything
that needs to be done by me has to be using the library I mentioned
earlier in this post.

Thanks in advance for taking out your time and answering to this post.

WMK

[Xabi]

Post by WMK
I am a total newbie to parallel and distributed computing and I need
some advice from anybody kind enough to guide me on ways to solve my
problem. As part of scientific research at my university, I frequently
have to do complex thermochemical equilibrium calculations which could
be quite resource intensive and time consuming even on the latest
generation of standalone workstations. For the purpose of
thermochemical calculations, I use a proprietory state of the art but
commercial programming library in the form a .dll that allows calls to
a programming library which is coded in FORTRAN but supports function
calling via C and C++. Therefore I dont have access to the actual
source code of this library so I cant parallelize the actual algorithm
code.
After learning about distributed and parallel computing, I am
contemplating on implementing either a MPI implementation or a
distributed implementation for use in an office environment
Heterogenous cluster . Since I cannot optimize the actual algorithms,
MPI wouldnt be of much help. Distributed computing with RPC might be
an option where in I can distribute the entire computation over a
heterogenous networked cluster of office machines.
Can someone please guide me whether such an implementation is possible
or suggest me any alternate methods whereby I can make use of idle lab
machines to speedup my calculations.
Note: There is currently no possibility for me to use any other
programming library be it opensource or commercial. Therefore anything
that needs to be done by me has to be using the library I mentioned
earlier in this post.
Thanks in advance for taking out your time and answering to this post.
WMK

Well, I'm not an expert but I think that since you can access the
library and use C/C++ functions, you could split the entire process
into smaller programs and run them in different machines, you can then
code a program to control wether one of those small programs finnish
and pass the results to other and so. Since i'm not into chemistry I
cannot tell you if the algorithm can be splitted in any other way.