Cloud computing versus in-house clusters: a comparative study

Multi-core cloud clusters are best suited environment for academic institutions in the third world countries to gain supercomputing power and enable researchers to inquest new trends in scientific computing with affordable cost and less administrative load. This work aims to analyze the parallelism efficiency of a parallel computational fluid mechanics solver (the lattice Boltzmann method (LBM)) on multi-core cloud clusters. This paper demonstrates reliability and cost effectiveness of using tailor-made hired cloud clusters as compared to in-house high performance computing architecture. On these clusters we have found that the lattice Boltzmann implementation on a Cartesian grid is fully adaptive, highly flexible and cost effective to use for solving complex large fluid mechanical systems, such as flooding in real-time at a very low cost on leased cluster than in-house ones.