AWS parallelcluster documentation change
Summary
Added documentation about default EFA network configuration and customization options in AWS ParallelCluster 3.15.0, including warnings about unsupported configurations
Security assessment
This change documents new EFA network configuration features in ParallelCluster 3.15.0. While network configuration can have security implications, there is no evidence this addresses a specific security vulnerability. The change is primarily about performance optimization (maximizing EFA bandwidth) and functionality documentation.
Diff
diff --git a/parallelcluster/latest/ug/efa-v3.md b/parallelcluster/latest/ug/efa-v3.md index e20029475..4e1f4d2ff 100644 --- a//parallelcluster/latest/ug/efa-v3.md +++ b//parallelcluster/latest/ug/efa-v3.md @@ -4,0 +5,2 @@ +Default EFA network configurationCustomizing EFA network interfaces + @@ -15,2 +16,0 @@ We recommend that you run your EFA-enabled instances in a placement group. This -For more information, see [Elastic Fabric Adapter](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa.html) in the _Amazon EC2 User Guide_ and [Scale HPC workloads with elastic fabric adapter and AWS ParallelCluster](https://aws.amazon.com/blogs/opensource/scale-hpc-workloads-elastic-fabric-adapter-and-aws-parallelcluster/) in the _AWS Open Source Blog_. - @@ -24,0 +25,18 @@ By default, Ubuntu distributions enable ptrace (process trace) protection. ptrac +## Default EFA network configuration + +Starting in AWS ParallelCluster 3.15.0, when EFA is enabled, AWS ParallelCluster automatically configures EFA-only network interfaces to separate EFA traffic from IP traffic. This maximizes EFA bandwidth while minimizing IP address consumption. AWS ParallelCluster determines the optimal configuration based on the capabilities of the instance type. + +This default configuration is recommended for most workloads, including tightly-coupled HPC and distributed AI/ML training. + +## Customizing EFA network interfaces + +If your workload requires a different network configuration, such as maximizing ENA bandwidth on secondary network cards or configuring a subset of available network cards, you can override the default settings using the [SlurmQueues](./Scheduling-v3.html#Scheduling-v3-SlurmQueues) / [ComputeResources](./Scheduling-v3.html#Scheduling-v3-SlurmQueues-ComputeResources) / [LaunchTemplateOverrides](./Scheduling-v3.html#yaml-Scheduling-SlurmQueues-ComputeResources-LaunchTemplateOverrides) parameter. This replaces the entire network interface configuration of the compute nodes with the configuration defined in your launch template. + +For a step-by-step walkthrough, see [Customize compute node network interfaces with launch template overrides](./tutorial-network-customization-v3.html). + +###### Warning + +If you configure network interfaces in a way that is not supported by the instance type, instances will fail to launch. To verify the supported network configurations for your instance type, see [DescribeInstanceTypes](https://docs.aws.amazon.com/AWSEC2/latest/APIReference/API_DescribeInstanceTypes.html) in the _Amazon EC2 API Reference_. + +For more information, see [Elastic Fabric Adapter](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa.html) in the _Amazon EC2 User Guide_ and [Scale HPC workloads with elastic fabric adapter and AWS ParallelCluster](https://aws.amazon.com/blogs/opensource/scale-hpc-workloads-elastic-fabric-adapter-and-aws-parallelcluster/) in the _AWS Open Source Blog_. +