1 Migrating HPC applications to the Cloud: a use case of weather forecast André Monteiro **1 , Cláudio Teixeira 1 , Joaquim Sousa Pinto 1 1 Institute of Electronics and Telematics Engineering of Aveiro (IEETA), University of Aveiro, 3810- 193 Aveiro, Portugal Email: {andremonteiro, claudio, jsp}@ua.pt Abstract. On the last few years, Cloud Computing has grown and matured in a way that managed to compete with clusters and grids for scientific applications supported. The use of this technology has allowed minor projects to allocate proper computational resources to their demands, with lower budgets. On the other hand many conventional users keep ignoring the Cloud potential – fast deployment, pay- per-use, elasticity, adaptability, among others. Our objective was to prove the feaseability of scientific applications and delineate an outline work that researchers could follow to take advantage of the Cloud infrastructure. This work describes the migration of a weather forecast application to the Cloud, demonstrating a system contextualization, application analysis, virtualization and deployment. We compare performance and combined costs of the several types of Clouds with the original solution and also future solutions. 1 Introduction Scientific computing applications have historically been supported by datacenters ’ clusters or grids, but Cloud Computing (CC) evolved and matured in a way that can compete with these types of resources. CC is becoming extremely suitable for computing scientific applications as they advertise providing virtually unlimited amount of computing resources on demand and nearly in real-time [1] and high proficiency all along [2, 3]. For small organizations this paradigm change has been especially encouraging, as they do not easily access computational resources adequate to their needs. Another aspect is the fast resource and applications deployment. The Cloud scales in and out quickly on demand, showing no constraints on hardware deploying, configuration and maintaining. A dynamic shared computing infrastructure offers resources available on demand on pre-settled basis and is much more cost effective for a university to operate than public providers [4]. Migrating scientific applications to the Cloud is not a conventional procedure. Running systems need to be virtualized in order to create templates, operating systems need tuning with the new infrastructure and software requires adaptation to Cloud specifications. CC offers many potential benefits to researchers and scientists. However, many traditional resources’ users seem to disregard the potential of the Cloud, based on the existing disavantadges – and there are a few - and inherent different deployment model. The maturing of the Cloud on the High-Performance Computing (HPC) field seems to be an excellent opportunity to take advantage of its intrinsic characteristics as elasticity and scalability, but also pay-per- ** e-mail of corresponding author: andremonteiro@ua.pt