The old fashioned parallel and proprietary solutions suffer of ``legacy''. Legacy is a term, associated to hardware and software component, which gives a thermometer of how costs versus system maintenance and changes increase or decrease.
A simple principle on which ParMa2 is based is to reduce to minimum maintenance and start-up costs. This can be achieved because we does not matter about hardware technology evolution. We do not build complex proprietary hardware solutions (old in the same time they are built); in fact we let the big hardware company build more and more efficient network or processors and we simply buy their best ``brick''.
This concept is also applicable to software. The presence of operating systems like Linux (a better UNIX clone), which have been Open Sourced and in continuous evolution, guarantees us to have always the upgraded version of its kernel, concentrating ours effort in writing applications more than building hardware or operating systems.
We also use a standard parallel library (MPI: Message Passing Interface) by which we can write and port a lot of distributed application. MPI is not alone, there are a lt of other standard by which we can write distributed applications. In fact there are some ParMa2 activities based on CORBA (Common Object Request Broker Architecture) a distributed object oriented standard; other solution could use shared memory approach (SGI, Sun, Sequent) but we are simply not polarized in those direction.
With these simple concepts we have built a system with a cost/performance
ratio which is comparable to other proprietary solution (SGI, SP2), which
are good static system in a dynamic world.