Salesforce Used SageMaker Inference Component SchedulingConfig to Hit Multi-AZ HA for Model Serving
AWS Machine Learning Blog·low signal
AWS details how Salesforce distributed model copies across Availability Zones using the SchedulingConfig parameter on SageMaker AI Inference Components, meeting an internal Multi-AZ high-availability requirement without running separate endpoints per zone. The relevant detail for anyone on SageMaker is that AZ placement is now a config parameter on the inference component rather than something you engineer around at the endpoint layer. Single-source vendor post.
AWS Machine Learning Bloglow signal