Seeking Best Practices for Multi-Region Load Balancing with Azure OpenAI Assistants API

Question

We would be grateful for any advice that could help us resolve these issues and achieve efficient multi-region operation.We are considering using Azure OpenAI's Assistants API in a multi-region environment for load balancing purposes. We would appreciate any best practices for implementing a solution that meets the following requirements:

[Requirements]

• Users will not create Assistants.
• Files to be used by Assistants will be pre-managed in our own database.
• When users need to use a file, they will select it from a list of files managed in our database.
• We want to keep the number of files in Azure OpenAI's storage to a minimum to avoid duplication.

The following are the strategies we have considered but found insufficient:

1. Uploading the files uploaded by users to all regions
   1. File IDs differ by region in Azure OpenAI, making management difficult.
   2. Storage usage fees may apply in Azure OpenAI.
2. Uploading files to the region being used at the time of Assistants execution
   1. Uploading large files can be time-consuming.
   2. There is a risk of uploading the same file multiple times or across multiple regions.
3. Fixing the region available to each user
   1. There is a possibility of uneven access load on a particular region.

We would be grateful for any advice that could help us resolve these issues and achieve efficient multi-region operation.