Ab initio quantum chemistry calculations demand enormous memory and computational requirements. So it makes them perfect applications to run on supercomputers. In order to optimize the cost efficiency of running on many nodes, we have engaged in the development of effective ways to run these very demanding codes on clusters of computers, using both AIX and Linux. Given an efficient method for inter-node communications, clusters of computers or workstations can deliver the power of supercomputers at a fraction of the cost. On the other hand, clusters have low bandwidth and high latency that makes quantum chemistry code that scale well on supercomputers scale less well on clusters.

       The purpose of my thesis work is to develop/support new perspective parallelization tools that will help to overcome the disadvantages of using clusters. Our new methods will ultimately greatly improve the scalability of computational chemistry codes on clusters. Currently, we are developing a library that emulates a shared memory approach on distributed memory systems via asynchronous one sided operations. Most importantly, we are developing a parallelization of quantum chemistry and other computational chemistry codes, most notably those in GAMESS, in such a way that they can take advantage of this new model. We have already developed a successful distributed parallelized Self Consistent Field (SCF) code. The next step will be to parallelize the very important and highly demanding Multi Configurational Self Consistent Field code. The third part of thesis work is to run large calculations of industrially important reactions that can take advantage from such code on clusters of workstations.