aws-parallelcluster-relion

set up example for Relion on AWS ParallelCluster

• Validated conditions
  • AWS ParallelCluster 3.1.3
  • Relion 3.1.3
  • us-east-1 region

Architecture Overview

• ⚠️ This tutorial environment does not include FSx for Lustre / S3 integration

Tutorial

1. Launch Cluster

On AWS CloudShell, run following commands for cluster creation. The create_relion_cluster.sh script use us-east-1 region in default. You could change AWS_REGION parameter in the script.

git clone https://github.com/DaisukeMiyamoto/aws-parallelcluster-relion
cd aws-parallelcluster-relion/01_setup_cluster
./create_relion_cluster.sh
source ~/.bashrc

You could check cluster status with following command. Cluster creation usually needs 5 to 10 minutes. If the clusterStatus become CREATE_COMPLETE from CREATE_IN_PROGRESS, go to next step.

pcluster list-clusters

2. Run Relion with GUI via NICE-DCV

2.1 Login ParallelCluster via NICE-DCV

Run following command and open published URL to access NICE-DCV remote desktop environment.

pcluster dcv-connect --cluster-name ${PCLUSTER_CLUSTER_NAME} --key-path ~/.ssh/${SSH_KEY}

• If it is your first time to access the cluster, a message like The authenticity of host '123.123.123.123 (123.123.123.123)' can't be established. could be shown. You need to type yes and hitting enter.
• On opening the NICE-DCV URL with browser, you will face Your connection is not private warning e.g. image. You need to click Advanced and Proceed to <IP> (unsafe) to open NICE-DCV desktop.

2.2 Compile and Install Relion

On NICE-DCV desktop, open terminal from Activities button on top-left corner. Run following commands on the terminal for installing Relion.

cd /shared
git clone https://github.com/DaisukeMiyamoto/aws-parallelcluster-relion
cd aws-parallelcluster-relion/02_relion_gui
chmod +x ./setup_relion_v31_gui.sh
./setup_relion_v31_gui.sh
source ~/.bashrc

• ⚠️ This compilation settings are not optimized. You need to customize for appropriate benchmarkings.

2.3 Use public data for demonstration

Run following commands to download datasets and launch Relion. This dataset include all completed steps and intermediate files. You could run any process without input parameter settings. Compilation needs few minutes.

• ⚠️ This step downloads large amount of data (7GB) from Osaka University. Please read guides for appropriate use.
  • http://www.protein.osaka-u.ac.jp/rcsfp/databases/members/kawabata/EMtutorial.html

cd /shared
wget http://www.protein.osaka-u.ac.jp/rcsfp/databases/members/kawabata/EMtutorial/EMPIAR-10248/EMPIAR-10248_tutorial_precalculated_results.tar.gz
tar xvf EMPIAR-10248_tutorial_precalculated_results.tar.gz
cd EMPIAR-10248_tutorial
relion &

2.4 Run Class2D on batch job

You could submit a job to job scheduler with Running tab settings. This tutorial shows an example about running Class2D process with batch job way.

1. Click 007:Class2D/ManuManuPick/ in Finished jobs window to re-load previous settings.
2. Click 2D classification in top-left corner to create a new settings.
3. Open Compute tab and set following parameters:
   • Use GPU acceleration?: No
4. Open Running tab and set following parameters:

• [Running] Tab
  • Number of MPI procs: 1
  • Number of threads: 16
  • Submit to queue?: Yes
  • Queue name: c6i-4xlarge
  • Queue submit command: sbatch
  • Standard submission script: /shared/aws-parallelcluster-relion/02_relion_gui/gui_batch_template_slurm.sh

5. Click Run! button to submit a new job.
   • After job submission, you could find new EC2 instance is launching on EC2 management console.
   • You could also find your job status with squeue command. Usually, launching instance needs almost 4 minutes.
   • When the job started, ST (STATUS) become R (RUNNING) from CF in squeue command. You could also check the job progress with output box on Relion ( image ).
6. This job takes arround 10 minutes.

• This figure show how relion job submission works with GUI and template files. Relion GUI replace parameters (e.g. XXXmpinodesXXX ) with GUI settings.

benchmark Relion with compile optimization

TBU

3.1 Download sample benchmark files

cd /shared/aws-parallelcluster-relion/03_relion_benchmarks
chmod +x 03_download_benchmarks.sh
./03_download_benchmarks.sh

3.2 benchmark for Intel/AMD CPUs

chmod +x 01_setup_dependencies.sh
./01_setup_dependencies.sh

3.3 benchmark for GPUs

3.4 benchmark for Graviton(Arm64) CPUs

Clean up Environments

TBU

On the terminal on CloudRun, run following command to delete cluster. This step delete all data in the cluster including FSx for Lustre.

pcluster delete-cluster -n ${PCLUSTER_CLUSTER_NAME}

Reference

• AWS ParallelCluster3 Tutorial