Online / 6 & 7 February 2021

schedule

Hitchhiker's guide from traditional HPC Cluster to containerized ensemble run at scale

How to lift GROMACS into a cloudy SLURM cluster and evolve to run GROMACS globally using containers.


In this sessions we are presenting several approaches to migrate from traditional HPC to cloud-native, containerized HPC using an ensemble run of the molecular dynamics code GROMACS as an example. The session will show how containerization via software management is coming to the rescue and how a palatable journey might look like.

In this sessions we are presenting several approaches to migrate from traditional HPC to cloud-native, containerized HPC using an ensemble run of the molecular dynamics code GROMACS as an example. The session will show how containerization via software management is coming to the rescue and how a palatable journey might look like. After an overview of the domain the talk will walk through how a traditional setup on-prem was lifted into the cloud by simplifying software management with SPACK and cluster lifecycle management with AWS ParallelCluster and discusses the differences and challenges introduced by ephemeral, on-demand compute clusters. Next, we will describe how the independent nature of ensemble runs fancy extrem parallel execution and how containerized schedulers are fulfilling this need, while introducing the burden of containerization with its own challenges. The journey described in the session will describe the challenges and oppertunites using software management software, traditional HPC and cloud-native schedulers as well as containerization as a range of options the audience can pick from, depending on their demands.

Speakers

Photo of Carsten Kutzner Carsten Kutzner

Links