Elisabetta Di Nitto, Jesús Gorroñogoitia, Indika Kumara, Georgios Meditskos, Dragan Radolović, Karthee Sivalingam, Román Sosa González

2020 22nd International Symposium on Symbolic and Numeric Algorithms for Scientific Computing (SYNASC)

Abstract: Complex applications, which include microservices, computationally intensive batch jobs, and sophisticated interaction with the external environment, demand heterogeneous computational infrastructures that range from cloud to HPC and edge computing. In this context, a crucial problem is to facilitate the work of DevOps teams in i. the conception of the right operational architecture for the application, ii. its transformation into infrastructural code that automates its deployment, taking into account the peculiarities of each of the diverse infrastructures involved in this, and iii. its operation. The SODALITE framework aims at addressing this scenario. This paper presents the main features offered by the first version of the framework, currently focusing on managing cloud and HPC clusters, and shows them in practice through a relevant case study.

Keywords: Scientific computingHeuristic algorithmsFocusingComputer architectureDynamic schedulingNumerical modelsEdge computing