Software Defined Data Center for High Performance Computing Applications

Abstract

In recent years, traditional data centers have been used to host high-performance computing infrastructure, such as HPC clusters, addressing specific requirements for different research and scientific projects. Computes, storage, network, and security infrastructure is usually from heterogenous manufacturers and has multiple management interfaces. This process implies a higher demand on data center administrators to manually attend the problems or particular configurations that require each of the HPC applications, which can affect its performance, together with the fact that it can present a poor resource allocation on Data Center. Software-Defined Data Centers (SDDC) have emerged as a solution for automating the management and self-provisioning of computing, storage, network, and security resources dynamically according to the software-defined policies for each of the applications running in the SDDC. With these paradigms in mind, this work aims to answer whether HPC applications can benefit in their performance using the advantages that SDDC offers to other applications such as business or enterprise. The results of this article are (i) we identify SDDC main components. (ii) we present an experimental approach to use SDDC network component for High-Performance MPI-based applications.