AWS for HPC

Computational environment for scientific computing (focusing on computational chemistry) on AWS EC2 platform.

This repository provides information about high-performance computing (HPC) and cluster management tools on AWS Cloud Cluster services. I compiled my notes so that you can learn how to leverage AWS services related to cloud clusters:

• Amazon Elastic Cloud Cluster (EC2)
• Amazon ParallelCluster
• Amazon CloudFormation
• Amazon EC2 Fleet
• Amazon Elastic Block Store (EBS)
• Amazon Elastic File System (EFS)
• Amazon Simple Storage Service (S3)
• Amazon AutoScaling

Note that these notes should be read from top to bottom.

Getting started and quickly launching an AWS EC2 instance

• What is AWS EC2?
• Creating your first compute node on AWS EC2
• Youtube: Instances In AWS | How To Create AWS EC2 Instance | AWS EC2 Tutorial | Edureka and complete video list is here.

Initial creation and setup of AWS EC2 instance

1. Go to https://aws.amazon.com/ec2/ and sign in to the AWS EC2 console with your Account ID, IAM username, and password
2. Go to Service and click EC2 dashboard and select Launch instance
3. Select an Amazon Machine Image (AMI)
   • Choose a Linux distribution (deliver method by Amazon Machine Image (AMI))
     • Red Hat 8 with HVM
     • Red Hat 7.2 with HVM
     • CentOS 7 with HVM
       • Setting up CentOS 7
       • Installing basic development tools
     • Ubuntu 19.04
     • Ubuntu 18.04 LTS
     • Ubuntu 16.04 LTS
4. Select an instance type
5. Configure instance details (optional)
6. Add and configure storage (optional)
7. Add tag (optional)
8. Set a security group (optional)
9. Launch instance
10. Create a new key pair or choose an existing key pair

Let's keep in mind downloading a newly created key pair (.pem) before launching an instance, as you will not be able to download it again.

Creating AMI image (optional)

• Create AMI image
• Launch a new instance using the created AMI image
• Launch a new parallel cluster using the created AMI image

Creating cloud cluster

• Installing AWS CLI
• Installing AWS ParallelCluster
• Installing AWS CloudFormation Helper Scripts
• Creating virtual parallel cluster
• Creating virtual parallel cluster with user-adjusted configuration
• Resizing worker nodes

Getting started parallel cluster

• First landing on the cluster
• Measuring computer power using LINPACK (benchmarking) (optional)
• Evaluating the performance of a parallel file system using IOR (optional)

Using job scheduler

• SGE
  • SGE job scheduling
  • Basic SGE command usage
  • Preparing and submitting the job on the cluster
  • Intel job script for running NWChem
• AWSBATCH
• SLURM
  • Common command and parallel code
  • Script for running with distributed memory
  • Script for running with hybrid memory
• TORQUE

Installing development tools

• GNU and GCC compilers
• Intel Compiler
• Intel Math Kernel Library (MKL)
• ATLAS
• BLAS
• LAPACK
• ScaLAPACK

Installing message passing interface (MPI) library

• Intel MPI
• OpenMPI
• MPICH
• MVAPICH2

Installing GPU-accelerated library

• NVIDIA driver
• CUDA toolkit

Installing computational chemistry software

• NWChem
  • Ubuntu
    • Installing with OpenMPI
    • Installing with OpenMPI and enabling CUDA
    • Installing with OpenMPI and enabling CUDA on pre-configured AWS Deep Learning AMI (Ubuntu 18.04)
  • Red Hat / CentOS
    • Installing with OpenMPI and enabling CUDA
    • Installing with Intel MPI and enabling CUDA
    • Installing with MPICH and enabling CUDA
    • Installing with MVAPICH2 and enabling CUDA
  • Note on NWChem compilation
• ORCA
  • OpenMPI
• GAMESS (US)
• PSI4
• OpenMolcas
• CP2K
• Gromacs
• LAMMPS
• VASP
• Quantum ESPRESSO

Useful notes

• Useful AWS CLI commands
• Copying file and folder between local computer and AWS instance

More to read

If you are interested in the AWS cloud cluster service, I recommend reading or visiting the following websites, blogs, and forums.

• https://aws-parallelcluster.readthedocs.io/en/latest/
• https://github.com/aws/aws-parallelcluster

Contact

Rangsiman Ketkaew - [email protected]