Speaker
Description
Thanks to the advent of recent orchestration techniques, it is possible to easily deploy computational clusters in a cloud environment. This project consists of orchestrating a SLURM batch system on top of a Kubernetes cluster instantiated on Cloud@CNAF infrastructure.
The proposed physical infrastructure is based on OpenStack, on hypervisors equipped with low latency interconnection and GP-GPUs. This configuration will enable dynamic scaling of the K8S cluster, according to the application needs.
In the presentation we will show how this cluster can be exploited to run SLURM-containerized real-life workflows taken from the bioinformatics domain.
We will also evaluate the performances obtained, comparing them with those measured in a baremetal HPC cluster.
