GENREL05-BP03 Verify that agent capabilities are available across all regions of availability
Agents require supporting infrastructure to service requests from foundation models. Using agents across a region of availability requires the supporting infrastructure to be available in that region.
Desired outcome: When implemented, this best practice improves the reliability of your generative AI workload by verifying that agents have access to the appropriate supporting infrastructure such as APIs or functions, so they may service a wider region of availability.
Benefits of establishing this best practice: Scale horizontally to increase aggregate workload availability - Data replication across a region of availability horizontally scales data access infrastructure, enabling foundation models to consistently service inference requests across a region of availability.
Level of risk exposed if this best practice is not established: Medium
Implementation guidance
Agents for Amazon Bedrock can be made available across regions, so long as the models and supporting infrastructure exist in the desired regions. Amazon Bedrock Agents make API calls on behalf of a user. Once deployed to a new region, these agents must have access to the same or regionally-equivalent API. Consider deploying your APIs across multiple regions behind a CloudFront distribution with latency-based routing. When possible, leverage Amazon Route 53 with latency-based routing to direct traffic within your VPC (and on the Amazon backbone) rather than taking private traffic public to route to an internal service. If your agent is not making calls to a foundation model using a cross-region inference profile, be sure to configure model access in all required regions.
Implementation steps
-
Deploy supporting agent infrastructure in the primary region or Availability Zone.
-
Deploy supporting agent infrastructure in the secondary region or Availability Zone.
-
Configure latency-based routing or a similar routing protocol which will distribute your load accordingly.
Resources
Related practices:
Related guides, videos, and documentation:
Related examples:
