AWS wellarchitected documentation change
Summary
Added guidance for pairing Savings Plans with generative AI endpoints and managing reserved capacity
Security assessment
Content updates provide cost optimization strategies for generative AI inference workloads. Recommendations about separating environments are for cost containment, not security isolation. No security controls, vulnerabilities, or security features are discussed.
Diff
diff --git a/wellarchitected/latest/financial-services-industry-lens/fsicost08.md b/wellarchitected/latest/financial-services-industry-lens/fsicost08.md index a0f331574..79fadb89d 100644 --- a//wellarchitected/latest/financial-services-industry-lens/fsicost08.md +++ b//wellarchitected/latest/financial-services-industry-lens/fsicost08.md @@ -17 +17,3 @@ Financial systems usually have a predicted usage pattern. Sign up for a compute - +For steady inference and model-serving workloads, pair Savings Plans with provisioned throughput or concurrency settings on managed generative AI endpoints (for example, Amazon Bedrock or Amazon SageMaker AI Endpoint). Avoid over-commitment by separating development or test environments from production accounts and verify that only sustained production traffic uses reserved compute capacity. Review plan coverage quarterly as model architectures, token volumes, and context sizes evolve. + +