AWS ParallelCluster 3.8 is now generally available. Key features in this release are support for Rocky Linux 8 and Amazon EC2 Capacity Blocks for ML, allowing you to reserve highly sought-after GPU instances on a future date to support your short duration machine learning (ML) workloads. Other important features in this release include:
Scaling strategies to help you meet your scaling and cost requirements, ranging from strict all-or-nothing, that avoid launching idle instances, to best-effort, which prioritizes throughput scaling over avoiding idle instances.
Support for mounting user-defined file systems in AWS to the /home directory of the head node, helping you simplify cluster storage management.
Custom Munge key to help you simplify integration of additional cluster resources that require the Scheduler Munge key.
Source:: Amazon AWS