ParallelCluster or Parallel Computing Service - how to choose

Brendan Bouffler
Date : October 9, 2025
Categories : AI/ML ,AWS Batch ,AWS ParallelCluster ,Amazon NICE DCV ,Elastic Fabric Adapter ,Life Sciences
Tags : ai ml ,aws ,batch ,cpus ,dcv ,ec2 ,efa ,gpus ,hpc ,high performance computing ,lustre ,mpi ,nccl ,parallel cluster ,schedulers ,storage ,autoscaling ,aws batch ,bioinformatics ,cloud computing ,elastic ,elastic fabric adapter ,hpc instances ,infiniband ,job scheduling ,scientific computing ,supercomputing ,technical computing ,tightly coupled ,virtualization ,vizualization ,techshorts

AWS Parallel Cluster and AWS Parallel Computing Service are not the same thing.

Today we dive into some of the quirky, pizza-fueled history of ParallelCluster, tracing its roots back to an open-source experiment, and explain how evolution produced PCS: a fully managed service that takes the heavy lifting out of building and running HPC clusters.

From cable ties to elastic queues, we explore the tradeoffs between control and convenience in cloud-based high performance computing.

If you have ideas for technical topics you’d like to see us cover in a future show, let us know by emailing us at ask-hpc@amazon.com.

Chapters

00:00 - Start 00:07 - ParallelCluster and Parallel Computing Service are not the same thing 01:02 - Then came ParallelCluster 01:43 - Finally, AWS Parallel Computing Service 01:56 - ParallelCluster isn’t a service 03:25 - Owmomg a cluster is a chore 04:43 - What’s different about PCS? 05:15 - Lots of node types, and file systems, too 05:22 - PCS managers Slurm for you 05:42 - New features just happen - like magic 07:46 - You have choices 08:25 - How to get started