AWS adds AZ-placement controls for SageMaker inference components

AWS says its new SchedulingConfig controls for SageMaker AI Inference Components let customers set how model copies are distributed across instances and Availability Zones. Salesforce used the capability with balanced AZ placement and a SPREAD strategy for Agentforce deployments, AWS says, to meet its requirement that production models support two zones while retaining shared-GPU hosting. The controls can preserve configured placement during scaling and updates, though AWS says capacity constraints can still result in only partially balanced deployment unless customers reserve capacity in each zone.