AI/ML Infrastructure Configuration in EKS

AI/ML Infrastructure Configuration in EKS An architecture diagram generated by Archify. EKS Cluster · AWS Cloud › EKS workloads and integrations EKS Cluster AWS Services · SageMaker · ECR · CloudWatch · AWS Cloud AWS Services SageMaker · ECR · CloudWatch Training Node Group · p4d.24xlarge · p3.16xlarge · g5.48xlarge · AWS Cloud › EKS workloads and integrations › Node Groups Training Node Group p4d.24xlarge · p3.16xlarge · g5.48xlarge Inference Node Group · CPU / GPU / Neuron runtime · AWS Cloud › EKS workloads and integrations › Node Groups Inference Node Group CPU / GPU / Neuron runtime CPU Node Group · c6i.32xlarge · r6i.32xlarge · AWS Cloud › EKS workloads and integrations › Node Groups CPU Node Group c6i.32xlarge · r6i.32xlarge Storage · EBS · EFS · FSx for Lustre · S3 · AWS Cloud › EKS workloads and integrations Storage EBS · EFS · FSx for Lustre · S3 Networking · VPC CNI · ENA/EFA · Placement Group · AWS Cloud › EKS workloads and integrations Networking VPC CNI · ENA/EFA · Placement Group integrates datasets · model artifacts distributed training AWS Cloud EKS workloads and integrations Node Groups Legend Node Groups Storage EKS Cluster Networking AWS Services

Illustrative configuration

  • • Instance examples require regional/quota/software checks.
  • • Shared storage and data permissions are not automatic.