Recommended and designed around Amazon OpenSearch Service as the read model for searching tens of millions of records under thousands of writes per second, fed asynchronously from a message queue. Integration was written to target it via IAM SigV4 auth, and a design doc listed the domain, role, and sizing asks. No domain was provisioned, so nothing was verified against the real service.
- What worked
- Fit the requirement well on paper: near-real-time bulk ingest, filtered full-text queries, and managed operation in the existing cloud footprint.
- What got in the way
- Infrastructure provisioning lives outside the application repo, so the recommendation could not be exercised end to end.
