Automating Benchpark with Packer

This project automates the creation of an Amazon Machine Image (AMI) that includes everything needed to run Benchpark inside a Jupyter Notebook environment. It uses Packer to provision an Amazon Linux EC2 instance, installs Docker, builds a custom Docker image with Benchpark pre-configured, and sets up a systemd service to run Jupyter automatically on boot.

Current Status & Known Issues

1. OpenMPI Installation Issue: The current build has an issue with OpenMPI not being installed properly. This affects the functionality of MPI-dependent commands within Benchpark.

Next Steps for Future Development:

1. Investigate and fix the OpenMPI installation in the Docker image build process
2. Verify MPI functionality works correctly after installation
3. Test affected benchmarks to ensure they run properly

Prerequisites

• AWS CLI
• Packer
• AWS credentials configured via aws configure or environment variables

Build the AMI

Step 1: Configure AWS credentials

1. Copy and Paste your unique AWS Credentials into the terminal

Step 2: Build the AMI With Packer

1. Run the following command from the directory containing your benchpark.pkr.hcl and Dockerfile

   packer build benchpark.pkr.hcl
   

   If your build is successful you will have output that looks something like

    ==> Builds finished. The artifacts of successful builds are:
    --> caliper-benchpark.amazon-ebs.amazonlinux: AMIs were created:
    us-east-1: ami-0995bf89a8c655414