Skip to content

Latest commit

 

History

History
87 lines (56 loc) · 4.43 KB

README.md

File metadata and controls

87 lines (56 loc) · 4.43 KB
Raw

HPC@Cloud Toolkit

This repository contains a dockerized environment and source code for the HPC@Cloud toolkit, comprised of a command-line interface for managing cloud infrastructure for HPC applications.

Contributing

Setting up a development environment

Although HPC@Cloud can be executed on Windows using Docker, for developing it's recommended to use a Linux distro or MacOS.

  • Install git
  • Install docker
  • Install terraform
  • Install aws cli, if you plan to use AWS
  • Install python version 3.11 or higher (ideally, install pyenv and setup a virtual environment for HPC@Cloud)

First time setup

  1. Run Docker and launch the dev containers with the following:
make init
  1. You then can install HPC@Cloud and its dependencies with the command below. You may need to install libpq-dev or other packages to be able to compile psycopg2:
make install
  1. If you plan to use AWS, generate credentials (ACCESS and SECRET keys) and run the following to create a default profile:
aws configure

It should create the file ~/.aws/configure with the default profile.

You are now ready to use HPC@Cloud!

The cluster_config.yaml file

To create a cluster configuration, copy the cluster_config.example.yaml file and rename it to cluster_config.yaml. Edit the file as you like, and then run from the command-line:

hpcac create-cluster

After the execution completes, the cluster will be available over SSH connection.

To destroy your cluster, run:

hpcac destroy-cluster

The tasks_config.yaml file and the my_files directory

Place all the files you want to transfer to the cluster inside the my_files folder. To setup a task to be executed in the cluster, copy the tasks_config.example.yaml file and rename it to tasks_config.yaml. Edit the file as you like, and then run from the command-line:

hpcac run-tasks

Publications

Please, find bellow the list of publications related to the HPC@Cloud Toolkit. If you need to cite HPC@Cloud Toolkit, please reference HPC@Cloud: A Provider-Agnostic Software Framework for Enabling HPC in Public Cloud Platforms for a general presentation.

  • Vanderlei Munhoz, Márcio Castro, Odorico Mendizabal. Strategies for Fault-Tolerant Tightly-coupled HPC Workloads Running on Low-Budget Spot Cloud Infrastructures. International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD). Bordeaux, France: IEEE Computer Society, 2022. [link] [bib]

  • Vanderlei Munhoz, Márcio Castro. Enabling the execution of HPC applications on public clouds with HPC@Cloud toolkit. Concurrency and Computation Practice and Experience (CCPE), 2023. [link]

  • Vanderlei Munhoz, Márcio Castro. HPC@Cloud: A Provider-Agnostic Software Framework for Enabling HPC in Public Cloud Platforms. Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD). Florianópolis, Brazil: SBC, 2022. [link] [bib]

  • Daniel Cordeiro, Emilio Francesquini, Marcos Amaris, Márcio Castro, Alexandro Baldassin, João Vicente Lima. Green Cloud Computing: Challenges and Opportunities. Simpósio Brasileiro de Sistemas de Informação (SBSI). Maceió, Brazil: SBC, 2023. [link]

  • Livia Ferrão, Vanderlei Munhoz, Márcio Castro. Análise do Sobrecusto de Utilização de Contêineres para Execução de Aplicações de HPC na Nuvem. Escola Regional de Alto Desempenho da Região Sul (ERAD/RS). Porto Alegre, Brazil: SBC, 2023. [link]

  • Luiz Fernando Althoff, Vanderlei Munhoz, Márcio Castro. Análise de Viabilidade do Perfilamento de Aplicações de HPC Baseada em Contadores de Hardware na AWS. Escola Regional de Alto Desempenho da Região Sul (ERAD/RS). Porto Alegre, Brazil: SBC, 2023. [link]