Application of cloud computing in astrophysics: the case of Amazon Web Services

Astrophysics entered a new era where Big Data technologies and their related analysis paradigms are mandatory as off the-shelf tools for everyday duties. Although large HPC facilities and Grid Computing are offered for extremely intensive simulations to the community (for example in General Relativity, Magneto Hydrodynamics or Cosmological simulations) a number of medium sized problem, especially in the field of astrophysical numerical codes, are hard to be solved on premises and cannot approached with a proper computational architecture for large HPC. We present in this paper a case study application of computational power offered as unsold capacity on Amazon Elastic Cloud Computing (EC2). We concentrate here on two cases scenarios. In the first one we explore the cloud computing architecture for heavily parallelized CUDA based simulations for high resolution echelle spectrograph, while in the second one we report on architecture based on EC2 + S3 coupled with Docker for astrophysical simulations of Ultra Luminous X-rays Sources (ULXs) in nearby galaxies. Both architectures allowed us to decrease by order of magnitude the required CPU times by optimizing the overall cost.