Skip to content

Flows Bedrock: Model Cross-Region / On-Demand

0

Hi All,

I am creating a flow in Amazon Bedrock that requires a 'KnowledgeBase' node. If I create the flow using the Amazon Bedrock / FLOWS option, the node is set to On-DEMAND mode by default. When I test the node, a cross-region error occurs. However, if I create the same flow through Amazon SageMaker / Domains / Projects / Build Flow, the node is available with the Cross-Region option, and no error is generated. My question is, why are the features different between the two AWS services?

Thanks a lot

1 Answer
0

The difference you're experiencing between creating a flow in Amazon Bedrock and Amazon SageMaker is related to the recent introduction of cross-region inference support in Amazon Bedrock.

When you create a flow directly in Amazon Bedrock, it's likely using an older or default configuration that doesn't automatically include cross-region inference capabilities. This is why you're encountering the on-demand mode by default and experiencing a cross-region error when testing the KnowledgeBase node.

On the other hand, when you create the flow through Amazon SageMaker Domains/Projects/Build Flow, it's probably using a more up-to-date configuration that includes the cross-region inference option. This newer setup allows for better management of traffic bursts and higher throughput by utilizing compute across different AWS Regions.

To resolve the issue in Amazon Bedrock, you need to use an inference profile that supports cross-region inference. Here's what you can do:

  1. Create or select an inference profile in the Amazon Bedrock console that includes the model you're using and supports cross-region inference.
  2. In your Flow Bedrock setup, instead of using the default on-demand mode, use the ID or ARN of the inference profile you created.

Using an inference profile with cross-region support allows you to:

  • Manage traffic bursts more effectively
  • Achieve higher throughput limits (up to 2x the in-region quotas)
  • Enhance resilience during peak demand periods
  • Comply with data residency requirements by selecting from pre-defined region sets

It's worth noting that Amazon Bedrock is continually evolving, and the discrepancy you're seeing might be due to different release cycles or configurations between the services. As Amazon Bedrock further integrates cross-region inference capabilities, you may see this difference disappear in future updates.
Sources
Flow Bedrock: Error Knowledge Bases | AWS re:Post
Getting started with cross-region inference in Amazon Bedrock | AWS Machine Learning Blog
Amazon Bedrock now supports cross-region inference - AWS

answered 2 years ago

EXPERT

reviewed 2 years ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.